Name

ngx_http_lua_module - Embed the power of Lua into Nginx HTTP Servers.

This module is not distributed with the Nginx source. See the installation instructions.

Table of Contents

Status

Production ready.

Version

This document describes ngx_lua v0.10.13 released on 22 April 2018.

Synopsis

  1. # set search paths for pure Lua external libraries (';;' is the default path):
  2. lua_package_path '/foo/bar/?.lua;/blah/?.lua;;';
  3. # set search paths for Lua external libraries written in C (can also use ';;'):
  4. lua_package_cpath '/bar/baz/?.so;/blah/blah/?.so;;';
  5. server {
  6. location /lua_content {
  7. # MIME type determined by default_type:
  8. default_type 'text/plain';
  9. content_by_lua_block {
  10. ngx.say('Hello,world!')
  11. }
  12. }
  13. location /nginx_var {
  14. # MIME type determined by default_type:
  15. default_type 'text/plain';
  16. # try access /nginx_var?a=hello,world
  17. content_by_lua_block {
  18. ngx.say(ngx.var.arg_a)
  19. }
  20. }
  21. location = /request_body {
  22. client_max_body_size 50k;
  23. client_body_buffer_size 50k;
  24. content_by_lua_block {
  25. ngx.req.read_body() -- explicitly read the req body
  26. local data = ngx.req.get_body_data()
  27. if data then
  28. ngx.say("body data:")
  29. ngx.print(data)
  30. return
  31. end
  32. -- body may get buffered in a temp file:
  33. local file = ngx.req.get_body_file()
  34. if file then
  35. ngx.say("body is in file ", file)
  36. else
  37. ngx.say("no body found")
  38. end
  39. }
  40. }
  41. # transparent non-blocking I/O in Lua via subrequests
  42. # (well, a better way is to use cosockets)
  43. location = /lua {
  44. # MIME type determined by default_type:
  45. default_type 'text/plain';
  46. content_by_lua_block {
  47. local res = ngx.location.capture("/some_other_location")
  48. if res then
  49. ngx.say("status: ", res.status)
  50. ngx.say("body:")
  51. ngx.print(res.body)
  52. end
  53. }
  54. }
  55. location = /foo {
  56. rewrite_by_lua_block {
  57. res = ngx.location.capture("/memc",
  58. { args = { cmd = "incr", key = ngx.var.uri } }
  59. )
  60. }
  61. proxy_pass http://blah.blah.com;
  62. }
  63. location = /mixed {
  64. rewrite_by_lua_file /path/to/rewrite.lua;
  65. access_by_lua_file /path/to/access.lua;
  66. content_by_lua_file /path/to/content.lua;
  67. }
  68. # use nginx var in code path
  69. # CAUTION: contents in nginx var must be carefully filtered,
  70. # otherwise there'll be great security risk!
  71. location ~ ^/app/([-_a-zA-Z0-9/]+) {
  72. set $path $1;
  73. content_by_lua_file /path/to/lua/app/root/$path.lua;
  74. }
  75. location / {
  76. client_max_body_size 100k;
  77. client_body_buffer_size 100k;
  78. access_by_lua_block {
  79. -- check the client IP address is in our black list
  80. if ngx.var.remote_addr == "132.5.72.3" then
  81. ngx.exit(ngx.HTTP_FORBIDDEN)
  82. end
  83. -- check if the URI contains bad words
  84. if ngx.var.uri and
  85. string.match(ngx.var.request_body, "evil")
  86. then
  87. return ngx.redirect("/terms_of_use.html")
  88. end
  89. -- tests passed
  90. }
  91. # proxy_pass/fastcgi_pass/etc settings
  92. }
  93. }

Back to TOC

Description

This module embeds Lua, via LuaJIT 2.0/2.1, into Nginx and by leveraging Nginx’s subrequests, allows the integration of the powerful Lua threads (Lua coroutines) into the Nginx event model.

Unlike Apache’s mod_lua and Lighttpd’s mod_magnet, Lua code executed using this module can be 100% non-blocking on network traffic as long as the Nginx API for Lua provided by this module is used to handle
requests to upstream services such as MySQL, PostgreSQL, Memcached, Redis, or upstream HTTP web services.

At least the following Lua libraries and Nginx modules can be used with this ngx_lua module:

Almost all the Nginx modules can be used with this ngx_lua module by means of ngx.location.capture or ngx.location.capture_multi but it is recommended to use those lua-resty-* libraries instead of creating subrequests to access the Nginx upstream modules because the former is usually much more flexible and memory-efficient.

The Lua interpreter or LuaJIT instance is shared across all the requests in a single nginx worker process but request contexts are segregated using lightweight Lua coroutines.

Loaded Lua modules persist in the nginx worker process level resulting in a small memory footprint in Lua even when under heavy loads.

This module is plugged into NGINX’s “http” subsystem so it can only speaks downstream communication protocols in the HTTP family (HTTP 0.9/1.0/1.1/2.0, WebSockets, and etc).
If you want to do generic TCP communications with the downstream clients, then you should use the ngx_stream_lua module instead
which has a compatible Lua API.

Back to TOC

Typical Uses

Just to name a few:

  • Mashup’ing and processing outputs of various nginx upstream outputs (proxy, drizzle, postgres, redis, memcached, and etc) in Lua,
  • doing arbitrarily complex access control and security checks in Lua before requests actually reach the upstream backends,
  • manipulating response headers in an arbitrary way (by Lua)
  • fetching backend information from external storage backends (like redis, memcached, mysql, postgresql) and use that information to choose which upstream backend to access on-the-fly,
  • coding up arbitrarily complex web applications in a content handler using synchronous but still non-blocking access to the database backends and other storage,
  • doing very complex URL dispatch in Lua at rewrite phase,
  • using Lua to implement advanced caching mechanism for Nginx’s subrequests and arbitrary locations.

The possibilities are unlimited as the module allows bringing together various elements within Nginx as well as exposing the power of the Lua language to the user. The module provides the full flexibility of scripting while offering performance levels comparable with native C language programs both in terms of CPU time as well as memory footprint. This is particularly the case when LuaJIT 2.x is enabled.

Other scripting language implementations typically struggle to match this performance level.

The Lua state (Lua VM instance) is shared across all the requests handled by a single nginx worker process to minimize memory use.

Back to TOC

Nginx Compatibility

The latest version of this module is compatible with the following versions of Nginx:

  • 1.13.x (last tested: 1.13.6)
  • 1.12.x
  • 1.11.x (last tested: 1.11.2)
  • 1.10.x
  • 1.9.x (last tested: 1.9.15)
  • 1.8.x
  • 1.7.x (last tested: 1.7.10)
  • 1.6.x

Nginx cores older than 1.6.0 (exclusive) are not supported.

Back to TOC

Installation

It is highly recommended to use OpenResty releases which integrate Nginx, ngx_lua, OpenResty’s LuaJIT 2.1 branch version, as well as other powerful companion Nginx modules and Lua libraries. It is discouraged to build this module with nginx yourself since it is tricky to set up exactly right. Also, the stock nginx cores have various limitations and long standing bugs that can make some of this modules’ features become disabled, not work properly, or run slower. The same applies to LuaJIT as well. OpenResty includes its own version of LuaJIT which gets specifically optimized and enhanced for the OpenResty environment.

Alternatively, ngx_lua can be manually compiled into Nginx:

  1. LuaJIT can be downloaded from the latest release of OpenResty’s LuaJIT branch version. The official LuaJIT 2.0 and 2.1 releases are also supported, although the performance will be significantly lower in many cases.
  2. Download the latest version of the ngx_devel_kit (NDK) module HERE.
  3. Download the latest version of ngx_lua HERE.
  4. Download the latest version of Nginx HERE (See Nginx Compatibility)

Build the source with this module:

  1. wget 'http://nginx.org/download/nginx-1.13.6.tar.gz'
  2. tar -xzvf nginx-1.13.6.tar.gz
  3. cd nginx-1.13.6/
  4. # tell nginx's build system where to find LuaJIT 2.0:
  5. export LUAJIT_LIB=/path/to/luajit/lib
  6. export LUAJIT_INC=/path/to/luajit/include/luajit-2.0
  7. # tell nginx's build system where to find LuaJIT 2.1:
  8. export LUAJIT_LIB=/path/to/luajit/lib
  9. export LUAJIT_INC=/path/to/luajit/include/luajit-2.1
  10. # or tell where to find Lua if using Lua instead:
  11. #export LUA_LIB=/path/to/lua/lib
  12. #export LUA_INC=/path/to/lua/include
  13. # Here we assume Nginx is to be installed under /opt/nginx/.
  14. ./configure --prefix=/opt/nginx \
  15. --with-ld-opt="-Wl,-rpath,/path/to/luajit-or-lua/lib" \
  16. --add-module=/path/to/ngx_devel_kit \
  17. --add-module=/path/to/lua-nginx-module
  18. # Note that you may also want to add `./configure` options which are used in your
  19. # current nginx build.
  20. # You can get usually those options using command nginx -V
  21. # you can change the parallism number 2 below to fit the number of spare CPU cores in your
  22. # machine.
  23. make -j2
  24. make install

Back to TOC

Building as a dynamic module

Starting from NGINX 1.9.11, you can also compile this module as a dynamic module, by using the --add-dynamic-module=PATH option instead of --add-module=PATH on the
./configure command line above. And then you can explicitly load the module in your nginx.conf via the load_module
directive, for example,

  1. load_module /path/to/modules/ndk_http_module.so; # assuming NDK is built as a dynamic module too
  2. load_module /path/to/modules/ngx_http_lua_module.so;

Back to TOC

C Macro Configurations

While building this module either via OpenResty or with the NGINX core, you can define the following C macros via the C compiler options:

  • NGX_LUA_USE_ASSERT
    When defined, will enable assertions in the ngx_lua C code base. Recommended for debugging or testing builds. It can introduce some (small) runtime overhead when enabled. This macro was first introduced in the v0.9.10 release.
  • NGX_LUA_ABORT_AT_PANIC
    When the Lua/LuaJIT VM panics, ngx_lua will instruct the current nginx worker process to quit gracefully by default. By specifying this C macro, ngx_lua will abort the current nginx worker process (which usually result in a core dump file) immediately. This option is useful for debugging VM panics. This option was first introduced in the v0.9.8 release.
  • NGX_LUA_NO_FFI_API
    Excludes pure C API functions for FFI-based Lua API for NGINX (as required by lua-resty-core, for example). Enabling this macro can make the resulting binary code size smaller.

To enable one or more of these macros, just pass extra C compiler options to the ./configure script of either NGINX or OpenResty. For instance,

  1. ./configure --with-cc-opt="-DNGX_LUA_USE_ASSERT -DNGX_LUA_ABORT_AT_PANIC"

Back to TOC

Community

Back to TOC

English Mailing List

The openresty-en mailing list is for English speakers.

Back to TOC

Chinese Mailing List

The openresty mailing list is for Chinese speakers.

Back to TOC

Code Repository

The code repository of this project is hosted on github at openresty/lua-nginx-module.

Back to TOC

Bugs and Patches

Please submit bug reports, wishlists, or patches by

  1. creating a ticket on the GitHub Issue Tracker,
  2. or posting to the OpenResty community.

Back to TOC

Lua/LuaJIT bytecode support

As from the v0.5.0rc32 release, all *_by_lua_file configure directives (such as content_by_lua_file) support loading Lua 5.1 and LuaJIT 2.0/2.1 raw bytecode files directly.

Please note that the bytecode format used by LuaJIT 2.0/2.1 is not compatible with that used by the standard Lua 5.1 interpreter. So if you are using LuaJIT 2.0/2.1 with ngx_lua, LuaJIT compatible bytecode files must be generated as shown:

  1. /path/to/luajit/bin/luajit -b /path/to/input_file.lua /path/to/output_file.ljbc

The -bg option can be used to include debug information in the LuaJIT bytecode file:

  1. /path/to/luajit/bin/luajit -bg /path/to/input_file.lua /path/to/output_file.ljbc

Please refer to the official LuaJIT documentation on the -b option for more details:

http://luajit.org/running.html#opt_b

Also, the bytecode files generated by LuaJIT 2.1 is not compatible with LuaJIT 2.0, and vice versa. The support for LuaJIT 2.1 bytecode was first added in ngx_lua v0.9.3.

Attempts to load standard Lua 5.1 bytecode files into ngx_lua instances linked to LuaJIT 2.0/2.1 or vice versa, will result in an error message, such as that below, being logged into the Nginx error.log file:

  1. [error] 13909#0: *1 failed to load Lua inlined code: bad byte-code header in /path/to/test_file.luac

Loading bytecode files via the Lua primitives like require and dofile should always work as expected.

Back to TOC

System Environment Variable Support

If you want to access the system environment variable, say, foo, in Lua via the standard Lua API os.getenv, then you should also list this environment variable name in your nginx.conf file via the env directive. For example,

  1. env foo;

Back to TOC

HTTP 1.0 support

The HTTP 1.0 protocol does not support chunked output and requires an explicit Content-Length header when the response body is not empty in order to support the HTTP 1.0 keep-alive.
So when a HTTP 1.0 request is made and the lua_http10_buffering directive is turned on, ngx_lua will buffer the
output of ngx.say and ngx.print calls and also postpone sending response headers until all the response body output is received.
At that time ngx_lua can calculate the total length of the body and construct a proper Content-Length header to return to the HTTP 1.0 client.
If the Content-Length response header is set in the running Lua code, however, this buffering will be disabled even if the lua_http10_buffering directive is turned on.

For large streaming output responses, it is important to disable the lua_http10_buffering directive to minimise memory usage.

Note that common HTTP benchmark tools such as ab and http_load issue HTTP 1.0 requests by default.
To force curl to send HTTP 1.0 requests, use the -0 option.

Back to TOC

Statically Linking Pure Lua Modules

When LuaJIT 2.x is used, it is possible to statically link the bytecode of pure Lua modules into the Nginx executable.

Basically you use the luajit executable to compile .lua Lua module files to .o object files containing the exported bytecode data, and then link the .o files directly in your Nginx build.

Below is a trivial example to demonstrate this. Consider that we have the following .lua file named foo.lua:

  1. -- foo.lua
  2. local _M = {}
  3. function _M.go()
  4. print("Hello from foo")
  5. end
  6. return _M

And then we compile this .lua file to foo.o file:

  1. /path/to/luajit/bin/luajit -bg foo.lua foo.o

What matters here is the name of the .lua file, which determines how you use this module later on the Lua land. The file name foo.o does not matter at all except the .o file extension (which tells luajit what output format is used). If you want to strip the Lua debug information from the resulting bytecode, you can just specify the -b option above instead of -bg.

Then when building Nginx or OpenResty, pass the --with-ld-opt="foo.o" option to the ./configure script:

  1. ./configure --with-ld-opt="/path/to/foo.o" ...

Finally, you can just do the following in any Lua code run by ngx_lua:

  1. local foo = require "foo"
  2. foo.go()

And this piece of code no longer depends on the external foo.lua file any more because it has already been compiled into the nginx executable.

If you want to use dot in the Lua module name when calling require, as in

  1. local foo = require "resty.foo"

then you need to rename the foo.lua file to resty_foo.lua before compiling it down to a .o file with the luajit command-line utility.

It is important to use exactly the same version of LuaJIT when compiling .lua files to .o files as building nginx + ngx_lua. This is because the LuaJIT bytecode format may be incompatible between different LuaJIT versions. When the bytecode format is incompatible, you will see a Lua runtime error saying that the Lua module is not found.

When you have multiple .lua files to compile and link, then just specify their .o files at the same time in the value of the --with-ld-opt option. For instance,

  1. ./configure --with-ld-opt="/path/to/foo.o /path/to/bar.o" ...

If you have just too many .o files, then it might not be feasible to name them all in a single command. In this case, you can build a static library (or archive) for your .o files, as in

  1. ar rcus libmyluafiles.a *.o

then you can link the myluafiles archive as a whole to your nginx executable:

  1. ./configure \
  2. --with-ld-opt="-L/path/to/lib -Wl,--whole-archive -lmyluafiles -Wl,--no-whole-archive"

where /path/to/lib is the path of the directory containing the libmyluafiles.a file. It should be noted that the linker option --whole-archive is required here because otherwise our archive will be skipped because no symbols in our archive are mentioned in the main parts of the nginx executable.

Back to TOC

Data Sharing within an Nginx Worker

To globally share data among all the requests handled by the same nginx worker process, encapsulate the shared data into a Lua module, use the Lua require builtin to import the module, and then manipulate the shared data in Lua. This works because required Lua modules are loaded only once and all coroutines will share the same copy of the module (both its code and data). Note however that Lua global variables (note, not module-level variables) WILL NOT persist between requests because of the one-coroutine-per-request isolation design.

Here is a complete small example:

  1. -- mydata.lua
  2. local _M = {}
  3. local data = {
  4. dog = 3,
  5. cat = 4,
  6. pig = 5,
  7. }
  8. function _M.get_age(name)
  9. return data[name]
  10. end
  11. return _M

and then accessing it from nginx.conf:

  1. location /lua {
  2. content_by_lua_block {
  3. local mydata = require "mydata"
  4. ngx.say(mydata.get_age("dog"))
  5. }
  6. }

The mydata module in this example will only be loaded and run on the first request to the location /lua,
and all subsequent requests to the same nginx worker process will use the reloaded instance of the
module as well as the same copy of the data in it, until a HUP signal is sent to the Nginx master process to force a reload.
This data sharing technique is essential for high performance Lua applications based on this module.

Note that this data sharing is on a per-worker basis and not on a per-server basis. That is, when there are multiple nginx worker processes under an Nginx master, data sharing cannot cross the process boundary between these workers.

It is usually recommended to share read-only data this way. You can also share changeable data among all the concurrent requests of each nginx worker process as
long as there is no nonblocking I/O operations (including ngx.sleep)
in the middle of your calculations. As long as you do not give the
control back to the nginx event loop and ngx_lua’s light thread
scheduler (even implicitly), there can never be any race conditions in
between. For this reason, always be very careful when you want to share changeable data on the
worker level. Buggy optimizations can easily lead to hard-to-debug
race conditions under load.

If server-wide data sharing is required, then use one or more of the following approaches:

  1. Use the ngx.shared.DICT API provided by this module.
  2. Use only a single nginx worker and a single server (this is however not recommended when there is a multi core CPU or multiple CPUs in a single machine).
  3. Use data storage mechanisms such as memcached, redis, MySQL or PostgreSQL. The OpenResty bundle associated with this module comes with a set of companion Nginx modules and Lua libraries that provide interfaces with these data storage mechanisms.

Back to TOC

Known Issues

Back to TOC

TCP socket connect operation issues

The tcpsock:connect method may indicate success despite connection failures such as with Connection Refused errors.

However, later attempts to manipulate the cosocket object will fail and return the actual error status message generated by the failed connect operation.

This issue is due to limitations in the Nginx event model and only appears to affect Mac OS X.

Back to TOC

Lua Coroutine Yielding/Resuming

  • Because Lua’s dofile and require builtins are currently implemented as C functions in LuaJIT 2.0/2.1, if the Lua file being loaded by dofile or require invokes ngx.location.capture*, ngx.exec, ngx.exit, or other API functions requiring yielding in the top-level scope of the Lua file, then the Lua error “attempt to yield across C-call boundary” will be raised. To avoid this, put these calls requiring yielding into your own Lua functions in the Lua file instead of the top-level scope of the file.

Back to TOC

Lua Variable Scope

Care must be taken when importing modules and this form should be used:

  1. local xxx = require('xxx')

instead of the old deprecated form:

  1. require('xxx')

Here is the reason: by design, the global environment has exactly the same lifetime as the Nginx request handler associated with it. Each request handler has its own set of Lua global variables and that is the idea of request isolation. The Lua module is actually loaded by the first Nginx request handler and is cached by the require() built-in in the package.loaded table for later reference, and the module() builtin used by some Lua modules has the side effect of setting a global variable to the loaded module table. But this global variable will be cleared at the end of the request handler, and every subsequent request handler all has its own (clean) global environment. So one will get Lua exception for accessing the nil value.

The use of Lua global variables is a generally inadvisable in the ngx_lua context as:

  1. the misuse of Lua globals has detrimental side effects on concurrent requests when such variables should instead be local in scope,
  2. Lua global variables require Lua table look-ups in the global environment which is computationally expensive, and
  3. some Lua global variable references may include typing errors which make such difficult to debug.

It is therefore highly recommended to always declare such within an appropriate local scope instead.

  1. -- Avoid
  2. foo = 123
  3. -- Recommended
  4. local foo = 123
  5. -- Avoid
  6. function foo() return 123 end
  7. -- Recommended
  8. local function foo() return 123 end

To find all instances of Lua global variables in your Lua code, run the lua-releng tool across all .lua source files:

  1. $ lua-releng
  2. Checking use of Lua global variables in file lib/foo/bar.lua ...
  3. 1 [1489] SETGLOBAL 7 -1 ; contains
  4. 55 [1506] GETGLOBAL 7 -3 ; setvar
  5. 3 [1545] GETGLOBAL 3 -4 ; varexpand

The output says that the line 1489 of file lib/foo/bar.lua writes to a global variable named contains, the line 1506 reads from the global variable setvar, and line 1545 reads the global varexpand.

This tool will guarantee that local variables in the Lua module functions are all declared with the local keyword, otherwise a runtime exception will be thrown. It prevents undesirable race conditions while accessing such variables. See Data Sharing within an Nginx Worker for the reasons behind this.

Back to TOC

Locations Configured by Subrequest Directives of Other Modules

The ngx.location.capture and ngx.location.capture_multi directives cannot capture locations that include the add_before_body, add_after_body, auth_request, echo_location, echo_location_async, echo_subrequest, or echo_subrequest_async directives.

  1. location /foo {
  2. content_by_lua_block {
  3. res = ngx.location.capture("/bar")
  4. }
  5. }
  6. location /bar {
  7. echo_location /blah;
  8. }
  9. location /blah {
  10. echo "Success!";
  11. }
  1. $ curl -i http://example.com/foo

will not work as expected.

Back to TOC

Cosockets Not Available Everywhere

Due to internal limitations in the nginx core, the cosocket API is disabled in the following contexts: set_by_lua*, log_by_lua*, header_filter_by_lua*, and body_filter_by_lua.

The cosockets are currently also disabled in the init_by_lua* and init_worker_by_lua* directive contexts but we may add support for these contexts in the future because there is no limitation in the nginx core (or the limitation might be worked around).

There exists a work-around, however, when the original context does not need to wait for the cosocket results. That is, creating a zero-delay timer via the ngx.timer.at API and do the cosocket results in the timer handler, which runs asynchronously as to the original context creating the timer.

Back to TOC

Special Escaping Sequences

NOTE Following the v0.9.17 release, this pitfall can be avoided by using the *_by_lua_block {} configuration directives.

PCRE sequences such as \d, \s, or \w, require special attention because in string literals, the backslash character, \, is stripped out by both the Lua language parser and by the nginx config file parser before processing if not within a *_by_lua_block {} directive. So the following snippet will not work as expected:

  1. # nginx.conf
  2. ? location /test {
  3. ? content_by_lua '
  4. ? local regex = "\d+" -- THIS IS WRONG OUTSIDE OF A *_by_lua_block DIRECTIVE
  5. ? local m = ngx.re.match("hello, 1234", regex)
  6. ? if m then ngx.say(m[0]) else ngx.say("not matched!") end
  7. ? ';
  8. ? }
  9. # evaluates to "not matched!"

To avoid this, double escape the backslash:

  1. # nginx.conf
  2. location /test {
  3. content_by_lua '
  4. local regex = "\\\\d+"
  5. local m = ngx.re.match("hello, 1234", regex)
  6. if m then ngx.say(m[0]) else ngx.say("not matched!") end
  7. ';
  8. }
  9. # evaluates to "1234"

Here, \\\\d+ is stripped down to \\d+ by the Nginx config file parser and this is further stripped down to \d+ by the Lua language parser before running.

Alternatively, the regex pattern can be presented as a long-bracketed Lua string literal by encasing it in “long brackets”, [[...]], in which case backslashes have to only be escaped once for the Nginx config file parser.

  1. # nginx.conf
  2. location /test {
  3. content_by_lua '
  4. local regex = [[\\d+]]
  5. local m = ngx.re.match("hello, 1234", regex)
  6. if m then ngx.say(m[0]) else ngx.say("not matched!") end
  7. ';
  8. }
  9. # evaluates to "1234"

Here, [[\\d+]] is stripped down to [[\d+]] by the Nginx config file parser and this is processed correctly.

Note that a longer from of the long bracket, [=[...]=], may be required if the regex pattern contains [...] sequences.
The [=[...]=] form may be used as the default form if desired.

  1. # nginx.conf
  2. location /test {
  3. content_by_lua '
  4. local regex = [=[[0-9]+]=]
  5. local m = ngx.re.match("hello, 1234", regex)
  6. if m then ngx.say(m[0]) else ngx.say("not matched!") end
  7. ';
  8. }
  9. # evaluates to "1234"

An alternative approach to escaping PCRE sequences is to ensure that Lua code is placed in external script files and executed using the various *_by_lua_file directives.
With this approach, the backslashes are only stripped by the Lua language parser and therefore only need to be escaped once each.

  1. -- test.lua
  2. local regex = "\\d+"
  3. local m = ngx.re.match("hello, 1234", regex)
  4. if m then ngx.say(m[0]) else ngx.say("not matched!") end
  5. -- evaluates to "1234"

Within external script files, PCRE sequences presented as long-bracketed Lua string literals do not require modification.

  1. -- test.lua
  2. local regex = [[\d+]]
  3. local m = ngx.re.match("hello, 1234", regex)
  4. if m then ngx.say(m[0]) else ngx.say("not matched!") end
  5. -- evaluates to "1234"

As noted earlier, PCRE sequences presented within *_by_lua_block {} directives (available following the v0.9.17 release) do not require modification.

  1. # nginx.conf
  2. location /test {
  3. content_by_lua_block {
  4. local regex = [[\d+]]
  5. local m = ngx.re.match("hello, 1234", regex)
  6. if m then ngx.say(m[0]) else ngx.say("not matched!") end
  7. }
  8. }
  9. # evaluates to "1234"

Back to TOC

Mixing with SSI Not Supported

Mixing SSI with ngx_lua in the same Nginx request is not supported at all. Just use ngx_lua exclusively. Everything you can do with SSI can be done atop ngx_lua anyway and it can be more efficient when using ngx_lua.

Back to TOC

SPDY Mode Not Fully Supported

Certain Lua APIs provided by ngx_lua do not work in Nginx’s SPDY mode yet: ngx.location.capture, ngx.location.capture_multi, and ngx.req.socket.

Back to TOC

Missing data on short circuited requests

Nginx may terminate a request early with (at least):

  • 400 (Bad Request)
  • 405 (Not Allowed)
  • 408 (Request Timeout)
  • 413 (Request Entity Too Large)
  • 414 (Request URI Too Large)
  • 494 (Request Headers Too Large)
  • 499 (Client Closed Request)
  • 500 (Internal Server Error)
  • 501 (Not Implemented)

This means that phases that normally run are skipped, such as the rewrite or
access phase. This also means that later phases that are run regardless, e.g.
log_by_lua, will not have access to information that is normally set in those
phases.

Back to TOC

TODO

  • cosocket: implement LuaSocket’s unconnected UDP API.
  • port this module to the “datagram” subsystem of NGINX for implementing general UDP servers instead of HTTP
    servers in Lua. For example,

    1. datagram {
    2. server {
    3. listen 1953;
    4. handler_by_lua_block {
    5. -- custom Lua code implementing the special UDP server...
    6. }
    7. }
    8. }
  • shm: implement a “shared queue API” to complement the existing shared dict API.
  • cosocket: add support in the context of init_by_lua*.
  • cosocket: implement the bind() method for stream-typed cosockets.
  • cosocket: pool-based backend concurrency level control: implement automatic connect queueing when the backend concurrency exceeds its connection pool limit.
  • cosocket: review and merge aviramc’s patch for adding the bsdrecv method.
  • add new API function ngx.resp.add_header to emulate the standard add_header config directive.
  • review and apply vadim-pavlov’s patch for ngx.location.capture‘s extra_headers option
  • use ngx_hash_t to optimize the built-in header look-up process for ngx.req.set_header, ngx.header.HEADER, and etc.
  • add configure options for different strategies of handling the cosocket connection exceeding in the pools.
  • add directives to run Lua codes when nginx stops.
  • add ignore_resp_headers, ignore_resp_body, and ignore_resp options to ngx.location.capture and ngx.location.capture_multi methods, to allow micro performance tuning on the user side.
  • add automatic Lua code time slicing support by yielding and resuming the Lua VM actively via Lua’s debug hooks.
  • add stat mode similar to mod_lua.
  • cosocket: add client SSL certificate support.

Back to TOC

Changes

The changes made in every release of this module are listed in the change logs of the OpenResty bundle:

http://openresty.org/#Changes

Back to TOC

Test Suite

The following dependencies are required to run the test suite:

The order in which these modules are added during configuration is important because the position of any filter module in the
filtering chain determines the final output, for example. The correct adding order is shown above.

  • 3rd-party Lua libraries:

  • Applications:

    • mysql: create database ‘ngx_test’, grant all privileges to user ‘ngx_test’, password is ‘ngx_test’
    • memcached: listening on the default port, 11211.
    • redis: listening on the default port, 6379.

See also the developer build script for more details on setting up the testing environment.

To run the whole test suite in the default testing mode:

  1. cd /path/to/lua-nginx-module
  2. export PATH=/path/to/your/nginx/sbin:$PATH
  3. prove -I/path/to/test-nginx/lib -r t

To run specific test files:

  1. cd /path/to/lua-nginx-module
  2. export PATH=/path/to/your/nginx/sbin:$PATH
  3. prove -I/path/to/test-nginx/lib t/002-content.t t/003-errors.t

To run a specific test block in a particular test file, add the line --- ONLY to the test block you want to run, and then use the prove utility to run that .t file.

There are also various testing modes based on mockeagain, valgrind, and etc. Refer to the Test::Nginx documentation for more details for various advanced testing modes. See also the test reports for the Nginx test cluster running on Amazon EC2: http://qa.openresty.org.

Back to TOC

Copyright and License

This module is licensed under the BSD license.

Copyright (C) 2009-2017, by Xiaozhe Wang (chaoslawful) chaoslawful@gmail.com.

Copyright (C) 2009-2018, by Yichun “agentzh” Zhang (章亦春) agentzh@gmail.com, OpenResty Inc.

All rights reserved.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:

  • Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.

  • Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution.

THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS “AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

Back to TOC

See Also

Back to TOC

Directives

The basic building blocks of scripting Nginx with Lua are directives. Directives are used to specify when the user Lua code is run and
how the result will be used. Below is a diagram showing the order in which directives are executed.

Lua Nginx Modules Directives

Back to TOC

lua_capture_error_log

syntax: lua_capture_error_log size

default: none

context: http

Enables a buffer of the specified size for capturing all the nginx error log message data (not just those produced
by this module or the nginx http subsystem, but everything) without touching files or disks.

You can use units like k and m in the size value, as in

  1. lua_capture_error_log 100k;

As a rule of thumb, a 4KB buffer can usually hold about 20 typical error log messages. So do the maths!

This buffer never grows. If it is full, new error log messages will replace the oldest ones in the buffer.

The size of the buffer must be bigger than the maximum length of a single error log message (which is 4K in OpenResty and 2K in stock NGINX).

You can read the messages in the buffer on the Lua land via the
get_logs()
function of the
ngx.errlog
module of the lua-resty-core
library. This Lua API function will return the captured error log messages and
also remove these already read from the global capturing buffer, making room
for any new error log data. For this reason, the user should not configure this
buffer to be too big if the user read the buffered error log data fast enough.

Note that the log level specified in the standard error_log directive
does have effect on this capturing facility. It only captures log
messages of a level no lower than the specified log level in the error_log directive.
The user can still choose to set an even higher filtering log level on the fly via the Lua API function
errlog.set_filter_level.
So it is more flexible than the static error_log directive.

It is worth noting that there is no way to capture the debugging logs
without building OpenResty or NGINX with the ./configure
option --with-debug. And enabling debugging logs is
strongly discouraged in production builds due to high overhead.

This directive was first introduced in the v0.10.9 release.

Back to TOC

lua_use_default_type

syntax: lua_use_default_type on | off

default: lua_use_default_type on

context: http, server, location, location if

Specifies whether to use the MIME type specified by the default_type directive for the default value of the Content-Type response header. Deactivate this directive if a default Content-Type response header for Lua request handlers is not desired.

This directive is turned on by default.

This directive was first introduced in the v0.9.1 release.

Back to TOC

lua_malloc_trim

syntax: lua_malloc_trim <request-count>

default: lua_malloc_trim 1000

context: http

Asks the underlying libc runtime library to release its cached free memory back to the operating system every
N requests processed by the NGINX core. By default, N is 1000. You can configure the request count
by using your own numbers. Smaller numbers mean more frequent releases, which may introduce higher CPU time consumption and
smaller memory footprint while larger numbers usually lead to less CPU time overhead and relatively larger memory footprint.
Just tune the number for your own use cases.

Configuring the argument to 0 essentially turns off the periodical memory trimming altogether.

  1. lua_malloc_trim 0; # turn off trimming completely

The current implementation uses an NGINX log phase handler to do the request counting. So the appearance of the
log_subrequest on directives in nginx.conf
may make the counting faster when subrequests are involved. By default, only “main requests” count.

Note that this directive does not affect the memory allocated by LuaJIT’s own allocator based on the mmap
system call.

This directive was first introduced in the v0.10.7 release.

Back to TOC

lua_code_cache

syntax: lua_code_cache on | off

default: lua_code_cache on

context: http, server, location, location if

Enables or disables the Lua code cache for Lua code in *_by_lua_file directives (like set_by_lua_file and
content_by_lua_file) and Lua modules.

When turning off, every request served by ngx_lua will run in a separate Lua VM instance, starting from the 0.9.3 release. So the Lua files referenced in set_by_lua_file,
content_by_lua_file, access_by_lua_file,
and etc will not be cached
and all Lua modules used will be loaded from scratch. With this in place, developers can adopt an edit-and-refresh approach.

Please note however, that Lua code written inlined within nginx.conf
such as those specified by set_by_lua, content_by_lua,
access_by_lua, and rewrite_by_lua will not be updated when you edit the inlined Lua code in your nginx.conf file because only the Nginx config file parser can correctly parse the nginx.conf
file and the only way is to reload the config file
by sending a HUP signal or just to restart Nginx.

Even when the code cache is enabled, Lua files which are loaded by dofile or loadfile
in *_by_lua_file cannot be cached (unless you cache the results yourself). Usually you can either use the init_by_lua
or init_by_lua_file directives to load all such files or just make these Lua files true Lua modules
and load them via require.

The ngx_lua module does not support the stat mode available with the
Apache mod_lua module (yet).

Disabling the Lua code cache is strongly
discouraged for production use and should only be used during
development as it has a significant negative impact on overall performance. For example, the performance of a “hello world” Lua example can drop by an order of magnitude after disabling the Lua code cache.

Back to TOC

lua_regex_cache_max_entries

syntax: lua_regex_cache_max_entries <num>

default: lua_regex_cache_max_entries 1024

context: http

Specifies the maximum number of entries allowed in the worker process level compiled regex cache.

The regular expressions used in ngx.re.match, ngx.re.gmatch, ngx.re.sub, and ngx.re.gsub will be cached within this cache if the regex option o (i.e., compile-once flag) is specified.

The default number of entries allowed is 1024 and when this limit is reached, new regular expressions will not be cached (as if the o option was not specified) and there will be one, and only one, warning in the error.log file:

  1. 2011/08/27 23:18:26 [warn] 31997#0: *1 lua exceeding regex cache max entries (1024), ...

If you are using the ngx.re.* implementation of lua-resty-core by loading the resty.core.regex module (or just the resty.core module), then an LRU cache is used for the regex cache being used here.

Do not activate the o option for regular expressions (and/or replace string arguments for ngx.re.sub and ngx.re.gsub) that are generated on the fly and give rise to infinite variations to avoid hitting the specified limit.

Back to TOC

lua_regex_match_limit

syntax: lua_regex_match_limit <num>

default: lua_regex_match_limit 0

context: http

Specifies the “match limit” used by the PCRE library when executing the ngx.re API. To quote the PCRE manpage, “the limit … has the effect of limiting the amount of backtracking that can take place.”

When the limit is hit, the error string “pcre_exec() failed: -8” will be returned by the ngx.re API functions on the Lua land.

When setting the limit to 0, the default “match limit” when compiling the PCRE library is used. And this is the default value of this directive.

This directive was first introduced in the v0.8.5 release.

Back to TOC

lua_package_path

syntax: lua_package_path <lua-style-path-str>

default: The content of LUA_PATH environment variable or Lua’s compiled-in defaults.

context: http

Sets the Lua module search path used by scripts specified by set_by_lua,
content_by_lua and others. The path string is in standard Lua path form, and ;;
can be used to stand for the original search paths.

As from the v0.5.0rc29 release, the special notation $prefix or ${prefix} can be used in the search path string to indicate the path of the server prefix usually determined by the -p PATH command-line option while starting the Nginx server.

Back to TOC

lua_package_cpath

syntax: lua_package_cpath <lua-style-cpath-str>

default: The content of LUA_CPATH environment variable or Lua’s compiled-in defaults.

context: http

Sets the Lua C-module search path used by scripts specified by set_by_lua,
content_by_lua and others. The cpath string is in standard Lua cpath form, and ;;
can be used to stand for the original cpath.

As from the v0.5.0rc29 release, the special notation $prefix or ${prefix} can be used in the search path string to indicate the path of the server prefix usually determined by the -p PATH command-line option while starting the Nginx server.

Back to TOC

init_by_lua

syntax: init_by_lua <lua-script-str>

context: http

phase: loading-config

NOTE Use of this directive is discouraged following the v0.9.17 release. Use the init_by_lua_block directive instead.

Runs the Lua code specified by the argument <lua-script-str> on the global Lua VM level when the Nginx master process (if any) is loading the Nginx config file.

When Nginx receives the HUP signal and starts reloading the config file, the Lua VM will also be re-created and init_by_lua will run again on the new Lua VM. In case that the lua_code_cache directive is turned off (default on), the init_by_lua handler will run upon every request because in this special mode a standalone Lua VM is always created for each request.

Usually you can pre-load Lua modules at server start-up by means of this hook and take advantage of modern operating systems’ copy-on-write (COW) optimization. Here is an example for pre-loading Lua modules:

  1. # this runs before forking out nginx worker processes:
  2. init_by_lua_block { require "cjson" }
  3. server {
  4. location = /api {
  5. content_by_lua_block {
  6. -- the following require() will just return
  7. -- the alrady loaded module from package.loaded:
  8. ngx.say(require "cjson".encode{dog = 5, cat = 6})
  9. }
  10. }
  11. }

You can also initialize the lua_shared_dict shm storage at this phase. Here is an example for this:

  1. lua_shared_dict dogs 1m;
  2. init_by_lua_block {
  3. local dogs = ngx.shared.dogs;
  4. dogs:set("Tom", 56)
  5. }
  6. server {
  7. location = /api {
  8. content_by_lua_block {
  9. local dogs = ngx.shared.dogs;
  10. ngx.say(dogs:get("Tom"))
  11. }
  12. }
  13. }

But note that, the lua_shared_dict‘s shm storage will not be cleared through a config reload (via the HUP signal, for example). So if you do not want to re-initialize the shm storage in your init_by_lua code in this case, then you just need to set a custom flag in the shm storage and always check the flag in your init_by_lua code.

Because the Lua code in this context runs before Nginx forks its worker processes (if any), data or code loaded here will enjoy the Copy-on-write (COW) feature provided by many operating systems among all the worker processes, thus saving a lot of memory.

Do not initialize your own Lua global variables in this context because use of Lua global variables have performance penalties and can lead to global namespace pollution (see the Lua Variable Scope section for more details). The recommended way is to use proper Lua module files (but do not use the standard Lua function module() to define Lua modules because it pollutes the global namespace as well) and call require() to load your own module files in init_by_lua or other contexts (require() does cache the loaded Lua modules in the global package.loaded table in the Lua registry so your modules will only loaded once for the whole Lua VM instance).

Only a small set of the Nginx API for Lua is supported in this context:

More Nginx APIs for Lua may be supported in this context upon future user requests.

Basically you can safely use Lua libraries that do blocking I/O in this very context because blocking the master process during server start-up is completely okay. Even the Nginx core does blocking I/O (at least on resolving upstream’s host names) at the configure-loading phase.

You should be very careful about potential security vulnerabilities in your Lua code registered in this context because the Nginx master process is often run under the root account.

This directive was first introduced in the v0.5.5 release.

Back to TOC

init_by_lua_block

syntax: init_by_lua_block { lua-script }

context: http

phase: loading-config

Similar to the init_by_lua directive except that this directive inlines
the Lua source directly
inside a pair of curly braces ({}) instead of in an NGINX string literal (which requires
special character escaping).

For instance,

  1. init_by_lua_block {
  2. print("I need no extra escaping here, for example: \r\nblah")
  3. }

This directive was first introduced in the v0.9.17 release.

Back to TOC

init_by_lua_file

syntax: init_by_lua_file <path-to-lua-script-file>

context: http

phase: loading-config

Equivalent to init_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code or Lua/LuaJIT bytecode to be executed.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

This directive was first introduced in the v0.5.5 release.

Back to TOC

init_worker_by_lua

syntax: init_worker_by_lua <lua-script-str>

context: http

phase: starting-worker

NOTE Use of this directive is discouraged following the v0.9.17 release. Use the init_worker_by_lua_block directive instead.

Runs the specified Lua code upon every Nginx worker process’s startup when the master process is enabled. When the master process is disabled, this hook will just run after init_by_lua*.

This hook is often used to create per-worker reoccurring timers (via the ngx.timer.at Lua API), either for backend health-check or other timed routine work. Below is an example,

  1. init_worker_by_lua '
  2. local delay = 3 -- in seconds
  3. local new_timer = ngx.timer.at
  4. local log = ngx.log
  5. local ERR = ngx.ERR
  6. local check
  7. check = function(premature)
  8. if not premature then
  9. -- do the health check or other routine work
  10. local ok, err = new_timer(delay, check)
  11. if not ok then
  12. log(ERR, "failed to create timer: ", err)
  13. return
  14. end
  15. end
  16. end
  17. local hdl, err = new_timer(delay, check)
  18. if not hdl then
  19. log(ERR, "failed to create timer: ", err)
  20. return
  21. end
  22. ';

This directive was first introduced in the v0.9.5 release.

This hook no longer runs in the cache manager and cache loader processes since the v0.10.12 release.

Back to TOC

init_worker_by_lua_block

syntax: init_worker_by_lua_block { lua-script }

context: http

phase: starting-worker

Similar to the init_worker_by_lua directive except that this directive inlines
the Lua source directly
inside a pair of curly braces ({}) instead of in an NGINX string literal (which requires
special character escaping).

For instance,

  1. init_worker_by_lua_block {
  2. print("I need no extra escaping here, for example: \r\nblah")
  3. }

This directive was first introduced in the v0.9.17 release.

This hook no longer runs in the cache manager and cache loader processes since the v0.10.12 release.

Back to TOC

init_worker_by_lua_file

syntax: init_worker_by_lua_file <lua-file-path>

context: http

phase: starting-worker

Similar to init_worker_by_lua, but accepts the file path to a Lua source file or Lua bytecode file.

This directive was first introduced in the v0.9.5 release.

This hook no longer runs in the cache manager and cache loader processes since the v0.10.12 release.

Back to TOC

set_by_lua

syntax: set_by_lua $res <lua-script-str> [$arg1 $arg2 …]

context: server, server if, location, location if

phase: rewrite

NOTE Use of this directive is discouraged following the v0.9.17 release. Use the set_by_lua_block directive instead.

Executes code specified in <lua-script-str> with optional input arguments $arg1 $arg2 ..., and returns string output to $res.
The code in <lua-script-str> can make API calls and can retrieve input arguments from the ngx.arg table (index starts from 1 and increases sequentially).

This directive is designed to execute short, fast running code blocks as the Nginx event loop is blocked during code execution. Time consuming code sequences should therefore be avoided.

This directive is implemented by injecting custom commands into the standard ngx_http_rewrite_module‘s command list. Because ngx_http_rewrite_module does not support nonblocking I/O in its commands, Lua APIs requiring yielding the current Lua “light thread” cannot work in this directive.

At least the following API functions are currently disabled within the context of set_by_lua:

In addition, note that this directive can only write out a value to a single Nginx variable at
a time. However, a workaround is possible using the ngx.var.VARIABLE interface.

  1. location /foo {
  2. set $diff ''; # we have to predefine the $diff variable here
  3. set_by_lua $sum '
  4. local a = 32
  5. local b = 56
  6. ngx.var.diff = a - b; -- write to $diff directly
  7. return a + b; -- return the $sum value normally
  8. ';
  9. echo "sum = $sum, diff = $diff";
  10. }

This directive can be freely mixed with all directives of the ngx_http_rewrite_module, set-misc-nginx-module, and array-var-nginx-module modules. All of these directives will run in the same order as they appear in the config file.

  1. set $foo 32;
  2. set_by_lua $bar 'return tonumber(ngx.var.foo) + 1';
  3. set $baz "bar: $bar"; # $baz == "bar: 33"

As from the v0.5.0rc29 release, Nginx variable interpolation is disabled in the <lua-script-str> argument of this directive and therefore, the dollar sign character ($) can be used directly.

This directive requires the ngx_devel_kit module.

Back to TOC

set_by_lua_block

syntax: set_by_lua_block $res { lua-script }

context: server, server if, location, location if

phase: rewrite

Similar to the set_by_lua directive except that

  1. this directive inlines the Lua source directly
    inside a pair of curly braces ({}) instead of in an NGINX string literal (which requires
    special character escaping), and
  2. this directive does not support extra arguments after the Lua script as in set_by_lua.

For example,

  1. set_by_lua_block $res { return 32 + math.cos(32) }
  2. # $res now has the value "32.834223360507" or alike.

No special escaping is required in the Lua code block.

This directive was first introduced in the v0.9.17 release.

Back to TOC

set_by_lua_file

syntax: set_by_lua_file $res <path-to-lua-script-file> [$arg1 $arg2 …]

context: server, server if, location, location if

phase: rewrite

Equivalent to set_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

Nginx variable interpolation is supported in the <path-to-lua-script-file> argument string of this directive. But special care must be taken for injection attacks.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

When the Lua code cache is turned on (by default), the user code is loaded once at the first request and cached
and the Nginx config must be reloaded each time the Lua source file is modified.
The Lua code cache can be temporarily disabled during development by
switching lua_code_cache off in nginx.conf to avoid reloading Nginx.

This directive requires the ngx_devel_kit module.

Back to TOC

content_by_lua

syntax: content_by_lua <lua-script-str>

context: location, location if

phase: content

NOTE Use of this directive is discouraged following the v0.9.17 release. Use the content_by_lua_block directive instead.

Acts as a “content handler” and executes Lua code string specified in <lua-script-str> for every request.
The Lua code may make API calls and is executed as a new spawned coroutine in an independent global environment (i.e. a sandbox).

Do not use this directive and other content handler directives in the same location. For example, this directive and the proxy_pass directive should not be used in the same location.

Back to TOC

content_by_lua_block

syntax: content_by_lua_block { lua-script }

context: location, location if

phase: content

Similar to the content_by_lua directive except that this directive inlines
the Lua source directly
inside a pair of curly braces ({}) instead of in an NGINX string literal (which requires
special character escaping).

For instance,

  1. content_by_lua_block {
  2. ngx.say("I need no extra escaping here, for example: \r\nblah")
  3. }

This directive was first introduced in the v0.9.17 release.

Back to TOC

content_by_lua_file

syntax: content_by_lua_file <path-to-lua-script-file>

context: location, location if

phase: content

Equivalent to content_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

Nginx variables can be used in the <path-to-lua-script-file> string to provide flexibility. This however carries some risks and is not ordinarily recommended.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

When the Lua code cache is turned on (by default), the user code is loaded once at the first request and cached
and the Nginx config must be reloaded each time the Lua source file is modified.
The Lua code cache can be temporarily disabled during development by
switching lua_code_cache off in nginx.conf to avoid reloading Nginx.

Nginx variables are supported in the file path for dynamic dispatch, for example:

  1. # CAUTION: contents in nginx var must be carefully filtered,
  2. # otherwise there'll be great security risk!
  3. location ~ ^/app/([-_a-zA-Z0-9/]+) {
  4. set $path $1;
  5. content_by_lua_file /path/to/lua/app/root/$path.lua;
  6. }

But be very careful about malicious user inputs and always carefully validate or filter out the user-supplied path components.

Back to TOC

rewrite_by_lua

syntax: rewrite_by_lua <lua-script-str>

context: http, server, location, location if

phase: rewrite tail

NOTE Use of this directive is discouraged following the v0.9.17 release. Use the rewrite_by_lua_block directive instead.

Acts as a rewrite phase handler and executes Lua code string specified in <lua-script-str> for every request.
The Lua code may make API calls and is executed as a new spawned coroutine in an independent global environment (i.e. a sandbox).

Note that this handler always runs after the standard ngx_http_rewrite_module. So the following will work as expected:

  1. location /foo {
  2. set $a 12; # create and initialize $a
  3. set $b ""; # create and initialize $b
  4. rewrite_by_lua 'ngx.var.b = tonumber(ngx.var.a) + 1';
  5. echo "res = $b";
  6. }

because set $a 12 and set $b "" run before rewrite_by_lua.

On the other hand, the following will not work as expected:

  1. ? location /foo {
  2. ? set $a 12; # create and initialize $a
  3. ? set $b ''; # create and initialize $b
  4. ? rewrite_by_lua 'ngx.var.b = tonumber(ngx.var.a) + 1';
  5. ? if ($b = '13') {
  6. ? rewrite ^ /bar redirect;
  7. ? break;
  8. ? }
  9. ?
  10. ? echo "res = $b";
  11. ? }

because if runs before rewrite_by_lua even if it is placed after rewrite_by_lua in the config.

The right way of doing this is as follows:

  1. location /foo {
  2. set $a 12; # create and initialize $a
  3. set $b ''; # create and initialize $b
  4. rewrite_by_lua '
  5. ngx.var.b = tonumber(ngx.var.a) + 1
  6. if tonumber(ngx.var.b) == 13 then
  7. return ngx.redirect("/bar");
  8. end
  9. ';
  10. echo "res = $b";
  11. }

Note that the ngx_eval module can be approximated by using rewrite_by_lua. For example,

  1. location / {
  2. eval $res {
  3. proxy_pass http://foo.com/check-spam;
  4. }
  5. if ($res = 'spam') {
  6. rewrite ^ /terms-of-use.html redirect;
  7. }
  8. fastcgi_pass ...;
  9. }

can be implemented in ngx_lua as:

  1. location = /check-spam {
  2. internal;
  3. proxy_pass http://foo.com/check-spam;
  4. }
  5. location / {
  6. rewrite_by_lua '
  7. local res = ngx.location.capture("/check-spam")
  8. if res.body == "spam" then
  9. return ngx.redirect("/terms-of-use.html")
  10. end
  11. ';
  12. fastcgi_pass ...;
  13. }

Just as any other rewrite phase handlers, rewrite_by_lua also runs in subrequests.

Note that when calling ngx.exit(ngx.OK) within a rewrite_by_lua handler, the nginx request processing control flow will still continue to the content handler. To terminate the current request from within a rewrite_by_lua handler, calling ngx.exit with status >= 200 (ngx.HTTP_OK) and status < 300 (ngx.HTTP_SPECIAL_RESPONSE) for successful quits and ngx.exit(ngx.HTTP_INTERNAL_SERVER_ERROR) (or its friends) for failures.

If the ngx_http_rewrite_module‘s rewrite directive is used to change the URI and initiate location re-lookups (internal redirections), then any rewrite_by_lua or rewrite_by_lua_file code sequences within the current location will not be executed. For example,

  1. location /foo {
  2. rewrite ^ /bar;
  3. rewrite_by_lua 'ngx.exit(503)';
  4. }
  5. location /bar {
  6. ...
  7. }

Here the Lua code ngx.exit(503) will never run. This will be the case if rewrite ^ /bar last is used as this will similarly initiate an internal redirection. If the break modifier is used instead, there will be no internal redirection and the rewrite_by_lua code will be executed.

The rewrite_by_lua code will always run at the end of the rewrite request-processing phase unless rewrite_by_lua_no_postpone is turned on.

Back to TOC

rewrite_by_lua_block

syntax: rewrite_by_lua_block { lua-script }

context: http, server, location, location if

phase: rewrite tail

Similar to the rewrite_by_lua directive except that this directive inlines
the Lua source directly
inside a pair of curly braces ({}) instead of in an NGINX string literal (which requires
special character escaping).

For instance,

  1. rewrite_by_lua_block {
  2. do_something("hello, world!\nhiya\n")
  3. }

This directive was first introduced in the v0.9.17 release.

Back to TOC

rewrite_by_lua_file

syntax: rewrite_by_lua_file <path-to-lua-script-file>

context: http, server, location, location if

phase: rewrite tail

Equivalent to rewrite_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

Nginx variables can be used in the <path-to-lua-script-file> string to provide flexibility. This however carries some risks and is not ordinarily recommended.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

When the Lua code cache is turned on (by default), the user code is loaded once at the first request and cached and the Nginx config must be reloaded each time the Lua source file is modified. The Lua code cache can be temporarily disabled during development by switching lua_code_cache off in nginx.conf to avoid reloading Nginx.

The rewrite_by_lua_file code will always run at the end of the rewrite request-processing phase unless rewrite_by_lua_no_postpone is turned on.

Nginx variables are supported in the file path for dynamic dispatch just as in content_by_lua_file.

Back to TOC

access_by_lua

syntax: access_by_lua <lua-script-str>

context: http, server, location, location if

phase: access tail

NOTE Use of this directive is discouraged following the v0.9.17 release. Use the access_by_lua_block directive instead.

Acts as an access phase handler and executes Lua code string specified in <lua-script-str> for every request.
The Lua code may make API calls and is executed as a new spawned coroutine in an independent global environment (i.e. a sandbox).

Note that this handler always runs after the standard ngx_http_access_module. So the following will work as expected:

  1. location / {
  2. deny 192.168.1.1;
  3. allow 192.168.1.0/24;
  4. allow 10.1.1.0/16;
  5. deny all;
  6. access_by_lua '
  7. local res = ngx.location.capture("/mysql", { ... })
  8. ...
  9. ';
  10. # proxy_pass/fastcgi_pass/...
  11. }

That is, if a client IP address is in the blacklist, it will be denied before the MySQL query for more complex authentication is executed by access_by_lua.

Note that the ngx_auth_request module can be approximated by using access_by_lua:

  1. location / {
  2. auth_request /auth;
  3. # proxy_pass/fastcgi_pass/postgres_pass/...
  4. }

can be implemented in ngx_lua as:

  1. location / {
  2. access_by_lua '
  3. local res = ngx.location.capture("/auth")
  4. if res.status == ngx.HTTP_OK then
  5. return
  6. end
  7. if res.status == ngx.HTTP_FORBIDDEN then
  8. ngx.exit(res.status)
  9. end
  10. ngx.exit(ngx.HTTP_INTERNAL_SERVER_ERROR)
  11. ';
  12. # proxy_pass/fastcgi_pass/postgres_pass/...
  13. }

As with other access phase handlers, access_by_lua will not run in subrequests.

Note that when calling ngx.exit(ngx.OK) within a access_by_lua handler, the nginx request processing control flow will still continue to the content handler. To terminate the current request from within a access_by_lua handler, calling ngx.exit with status >= 200 (ngx.HTTP_OK) and status < 300 (ngx.HTTP_SPECIAL_RESPONSE) for successful quits and ngx.exit(ngx.HTTP_INTERNAL_SERVER_ERROR) (or its friends) for failures.

Starting from the v0.9.20 release, you can use the access_by_lua_no_postpone
directive to control when to run this handler inside the “access” request-processing phase
of NGINX.

Back to TOC

access_by_lua_block

syntax: access_by_lua_block { lua-script }

context: http, server, location, location if

phase: access tail

Similar to the access_by_lua directive except that this directive inlines
the Lua source directly
inside a pair of curly braces ({}) instead of in an NGINX string literal (which requires
special character escaping).

For instance,

  1. access_by_lua_block {
  2. do_something("hello, world!\nhiya\n")
  3. }

This directive was first introduced in the v0.9.17 release.

Back to TOC

access_by_lua_file

syntax: access_by_lua_file <path-to-lua-script-file>

context: http, server, location, location if

phase: access tail

Equivalent to access_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

Nginx variables can be used in the <path-to-lua-script-file> string to provide flexibility. This however carries some risks and is not ordinarily recommended.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

When the Lua code cache is turned on (by default), the user code is loaded once at the first request and cached
and the Nginx config must be reloaded each time the Lua source file is modified.
The Lua code cache can be temporarily disabled during development by switching lua_code_cache off in nginx.conf to avoid repeatedly reloading Nginx.

Nginx variables are supported in the file path for dynamic dispatch just as in content_by_lua_file.

Back to TOC

header_filter_by_lua

syntax: header_filter_by_lua <lua-script-str>

context: http, server, location, location if

phase: output-header-filter

NOTE Use of this directive is discouraged following the v0.9.17 release. Use the header_filter_by_lua_block directive instead.

Uses Lua code specified in <lua-script-str> to define an output header filter.

Note that the following API functions are currently disabled within this context:

Here is an example of overriding a response header (or adding one if absent) in our Lua header filter:

  1. location / {
  2. proxy_pass http://mybackend;
  3. header_filter_by_lua 'ngx.header.Foo = "blah"';
  4. }

This directive was first introduced in the v0.2.1rc20 release.

Back to TOC

header_filter_by_lua_block

syntax: header_filter_by_lua_block { lua-script }

context: http, server, location, location if

phase: output-header-filter

Similar to the header_filter_by_lua directive except that this directive inlines
the Lua source directly
inside a pair of curly braces ({}) instead of in an NGINX string literal (which requires
special character escaping).

For instance,

  1. header_filter_by_lua_block {
  2. ngx.header["content-length"] = nil
  3. }

This directive was first introduced in the v0.9.17 release.

Back to TOC

header_filter_by_lua_file

syntax: header_filter_by_lua_file <path-to-lua-script-file>

context: http, server, location, location if

phase: output-header-filter

Equivalent to header_filter_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

This directive was first introduced in the v0.2.1rc20 release.

Back to TOC

body_filter_by_lua

syntax: body_filter_by_lua <lua-script-str>

context: http, server, location, location if

phase: output-body-filter

NOTE Use of this directive is discouraged following the v0.9.17 release. Use the body_filter_by_lua_block directive instead.

Uses Lua code specified in <lua-script-str> to define an output body filter.

The input data chunk is passed via ngx.arg[1] (as a Lua string value) and the “eof” flag indicating the end of the response body data stream is passed via ngx.arg[2] (as a Lua boolean value).

Behind the scene, the “eof” flag is just the last_buf (for main requests) or last_in_chain (for subrequests) flag of the Nginx chain link buffers. (Before the v0.7.14 release, the “eof” flag does not work at all in subrequests.)

The output data stream can be aborted immediately by running the following Lua statement:

  1. return ngx.ERROR

This will truncate the response body and usually result in incomplete and also invalid responses.

The Lua code can pass its own modified version of the input data chunk to the downstream Nginx output body filters by overriding ngx.arg[1] with a Lua string or a Lua table of strings. For example, to transform all the lowercase letters in the response body, we can just write:

  1. location / {
  2. proxy_pass http://mybackend;
  3. body_filter_by_lua 'ngx.arg[1] = string.upper(ngx.arg[1])';
  4. }

When setting nil or an empty Lua string value to ngx.arg[1], no data chunk will be passed to the downstream Nginx output filters at all.

Likewise, new “eof” flag can also be specified by setting a boolean value to ngx.arg[2]. For example,

  1. location /t {
  2. echo hello world;
  3. echo hiya globe;
  4. body_filter_by_lua '
  5. local chunk = ngx.arg[1]
  6. if string.match(chunk, "hello") then
  7. ngx.arg[2] = true -- new eof
  8. return
  9. end
  10. -- just throw away any remaining chunk data
  11. ngx.arg[1] = nil
  12. ';
  13. }

Then GET /t will just return the output

  1. hello world

That is, when the body filter sees a chunk containing the word “hello”, then it will set the “eof” flag to true immediately, resulting in truncated but still valid responses.

When the Lua code may change the length of the response body, then it is required to always clear out the Content-Length response header (if any) in a header filter to enforce streaming output, as in

  1. location /foo {
  2. # fastcgi_pass/proxy_pass/...
  3. header_filter_by_lua_block { ngx.header.content_length = nil }
  4. body_filter_by_lua 'ngx.arg[1] = string.len(ngx.arg[1]) .. "\\n"';
  5. }

Note that the following API functions are currently disabled within this context due to the limitations in NGINX output filter’s current implementation:

Nginx output filters may be called multiple times for a single request because response body may be delivered in chunks. Thus, the Lua code specified by in this directive may also run multiple times in the lifetime of a single HTTP request.

This directive was first introduced in the v0.5.0rc32 release.

Back to TOC

body_filter_by_lua_block

syntax: body_filter_by_lua_block { lua-script-str }

context: http, server, location, location if

phase: output-body-filter

Similar to the body_filter_by_lua directive except that this directive inlines
the Lua source directly
inside a pair of curly braces ({}) instead of in an NGINX string literal (which requires
special character escaping).

For instance,

  1. body_filter_by_lua_block {
  2. local data, eof = ngx.arg[1], ngx.arg[2]
  3. }

This directive was first introduced in the v0.9.17 release.

Back to TOC

body_filter_by_lua_file

syntax: body_filter_by_lua_file <path-to-lua-script-file>

context: http, server, location, location if

phase: output-body-filter

Equivalent to body_filter_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

This directive was first introduced in the v0.5.0rc32 release.

Back to TOC

log_by_lua

syntax: log_by_lua <lua-script-str>

context: http, server, location, location if

phase: log

NOTE Use of this directive is discouraged following the v0.9.17 release. Use the log_by_lua_block directive instead.

Runs the Lua source code inlined as the <lua-script-str> at the log request processing phase. This does not replace the current access logs, but runs before.

Note that the following API functions are currently disabled within this context:

Here is an example of gathering average data for $upstream_response_time:

  1. lua_shared_dict log_dict 5M;
  2. server {
  3. location / {
  4. proxy_pass http://mybackend;
  5. log_by_lua '
  6. local log_dict = ngx.shared.log_dict
  7. local upstream_time = tonumber(ngx.var.upstream_response_time)
  8. local sum = log_dict:get("upstream_time-sum") or 0
  9. sum = sum + upstream_time
  10. log_dict:set("upstream_time-sum", sum)
  11. local newval, err = log_dict:incr("upstream_time-nb", 1)
  12. if not newval and err == "not found" then
  13. log_dict:add("upstream_time-nb", 0)
  14. log_dict:incr("upstream_time-nb", 1)
  15. end
  16. ';
  17. }
  18. location = /status {
  19. content_by_lua_block {
  20. local log_dict = ngx.shared.log_dict
  21. local sum = log_dict:get("upstream_time-sum")
  22. local nb = log_dict:get("upstream_time-nb")
  23. if nb and sum then
  24. ngx.say("average upstream response time: ", sum / nb,
  25. " (", nb, " reqs)")
  26. else
  27. ngx.say("no data yet")
  28. end
  29. }
  30. }
  31. }

This directive was first introduced in the v0.5.0rc31 release.

Back to TOC

log_by_lua_block

syntax: log_by_lua_block { lua-script }

context: http, server, location, location if

phase: log

Similar to the log_by_lua directive except that this directive inlines
the Lua source directly
inside a pair of curly braces ({}) instead of in an NGINX string literal (which requires
special character escaping).

For instance,

  1. log_by_lua_block {
  2. print("I need no extra escaping here, for example: \r\nblah")
  3. }

This directive was first introduced in the v0.9.17 release.

Back to TOC

log_by_lua_file

syntax: log_by_lua_file <path-to-lua-script-file>

context: http, server, location, location if

phase: log

Equivalent to log_by_lua, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

This directive was first introduced in the v0.5.0rc31 release.

Back to TOC

balancer_by_lua_block

syntax: balancer_by_lua_block { lua-script }

context: upstream

phase: content

This directive runs Lua code as an upstream balancer for any upstream entities defined
by the upstream {} configuration block.

For instance,

  1. upstream foo {
  2. server 127.0.0.1;
  3. balancer_by_lua_block {
  4. -- use Lua to do something interesting here
  5. -- as a dynamic balancer
  6. }
  7. }
  8. server {
  9. location / {
  10. proxy_pass http://foo;
  11. }
  12. }

The resulting Lua load balancer can work with any existing nginx upstream modules
like ngx_proxy and
ngx_fastcgi.

Also, the Lua load balancer can work with the standard upstream connection pool mechanism,
i.e., the standard keepalive directive.
Just ensure that the keepalive directive
is used after this balancer_by_lua_block directive in a single upstream {} configuration block.

The Lua load balancer can totally ignore the list of servers defined in the upstream {} block
and select peer from a completely dynamic server list (even changing per request) via the
ngx.balancer module
from the lua-resty-core library.

The Lua code handler registered by this directive might get called more than once in a single
downstream request when the nginx upstream mechanism retries the request on conditions
specified by directives like the proxy_next_upstream
directive.

This Lua code execution context does not support yielding, so Lua APIs that may yield
(like cosockets and “light threads”) are disabled in this context. One can usually work
around this limitation by doing such operations in an earlier phase handler (like
access_by_lua*) and passing along the result into this context
via the ngx.ctx table.

This directive was first introduced in the v0.10.0 release.

Back to TOC

balancer_by_lua_file

syntax: balancer_by_lua_file <path-to-lua-script-file>

context: upstream

phase: content

Equivalent to balancer_by_lua_block, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

This directive was first introduced in the v0.10.0 release.

Back to TOC

lua_need_request_body

syntax: lua_need_request_body <on|off>

default: off

context: http, server, location, location if

phase: depends on usage

Determines whether to force the request body data to be read before running rewrite/access/access_by_lua* or not. The Nginx core does not read the client request body by default and if request body data is required, then this directive should be turned on or the ngx.req.read_body function should be called within the Lua code.

To read the request body data within the $request_body variable,
client_body_buffer_size must have the same value as client_max_body_size. Because when the content length exceeds client_body_buffer_size but less than client_max_body_size, Nginx will buffer the data into a temporary file on the disk, which will lead to empty value in the $request_body variable.

If the current location includes rewrite_by_lua* directives,
then the request body will be read just before the rewrite_by_lua* code is run (and also at the
rewrite phase). Similarly, if only content_by_lua is specified,
the request body will not be read until the content handler’s Lua code is
about to run (i.e., the request body will be read during the content phase).

It is recommended however, to use the ngx.req.read_body and ngx.req.discard_body functions for finer control over the request body reading process instead.

This also applies to access_by_lua*.

Back to TOC

ssl_certificate_by_lua_block

syntax: ssl_certificate_by_lua_block { lua-script }

context: server

phase: right-before-SSL-handshake

This directive runs user Lua code when NGINX is about to start the SSL handshake for the downstream
SSL (https) connections.

It is particularly useful for setting the SSL certificate chain and the corresponding private key on a per-request
basis. It is also useful to load such handshake configurations nonblockingly from the remote (for example,
with the cosocket API). And one can also do per-request OCSP stapling handling in pure
Lua here as well.

Another typical use case is to do SSL handshake traffic control nonblockingly in this context,
with the help of the lua-resty-limit-traffic#readme
library, for example.

One can also do interesting things with the SSL handshake requests from the client side, like
rejecting old SSL clients using the SSLv3 protocol or even below selectively.

The ngx.ssl
and ngx.ocsp Lua modules
provided by the lua-resty-core
library are particularly useful in this context. You can use the Lua API offered by these two Lua modules
to manipulate the SSL certificate chain and private key for the current SSL connection
being initiated.

This Lua handler does not run at all, however, when NGINX/OpenSSL successfully resumes
the SSL session via SSL session IDs or TLS session tickets for the current SSL connection. In
other words, this Lua handler only runs when NGINX has to initiate a full SSL handshake.

Below is a trivial example using the
ngx.ssl module
at the same time:

  1. server {
  2. listen 443 ssl;
  3. server_name test.com;
  4. ssl_certificate_by_lua_block {
  5. print("About to initiate a new SSL handshake!")
  6. }
  7. location / {
  8. root html;
  9. }
  10. }

See more complicated examples in the ngx.ssl
and ngx.ocsp
Lua modules’ official documentation.

Uncaught Lua exceptions in the user Lua code immediately abort the current SSL session, so does the
ngx.exit call with an error code like ngx.ERROR.

This Lua code execution context does support yielding, so Lua APIs that may yield
(like cosockets, sleeping, and “light threads”)
are enabled in this context.

Note, however, you still need to configure the ssl_certificate and
ssl_certificate_key
directives even though you will not use this static certificate and private key at all. This is
because the NGINX core requires their appearance otherwise you are seeing the following error
while starting NGINX:

  1. nginx: [emerg] no ssl configured for the server

This directive currently requires the following NGINX core patch to work correctly:

http://mailman.nginx.org/pipermail/nginx-devel/2016-January/007748.html

The bundled version of the NGINX core in OpenResty 1.9.7.2 (or above) already has this
patch applied.

Furthermore, one needs at least OpenSSL 1.0.2e for this directive to work.

This directive was first introduced in the v0.10.0 release.

Back to TOC

ssl_certificate_by_lua_file

syntax: ssl_certificate_by_lua_file <path-to-lua-script-file>

context: server

phase: right-before-SSL-handshake

Equivalent to ssl_certificate_by_lua_block, except that the file specified by <path-to-lua-script-file> contains the Lua code, or, as from the v0.5.0rc32 release, the Lua/LuaJIT bytecode to be executed.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

This directive was first introduced in the v0.10.0 release.

Back to TOC

ssl_session_fetch_by_lua_block

syntax: ssl_session_fetch_by_lua_block { lua-script }

context: http

phase: right-before-SSL-handshake

This directive runs Lua code to look up and load the SSL session (if any) according to the session ID
provided by the current SSL handshake request for the downstream.

The Lua API for obtaining the current session ID and loading a cached SSL session data
is provided in the ngx.ssl.session
Lua module shipped with the lua-resty-core
library.

Lua APIs that may yield, like ngx.sleep and cosockets,
are enabled in this context.

This hook, together with the ssl_session_store_by_lua* hook,
can be used to implement distributed caching mechanisms in pure Lua (based
on the cosocket API, for example). If a cached SSL session is found
and loaded into the current SSL connection context,
SSL session resumption can then get immediately initiated and bypass the full SSL handshake process which is very expensive in terms of CPU time.

Please note that TLS session tickets are very different and it is the clients’ responsibility
to cache the SSL session state when session tickets are used. SSL session resumptions based on
TLS session tickets would happen automatically without going through this hook (nor the
ssl_session_store_by_lua* hook). This hook is mainly
for older or less capable SSL clients that can only do SSL sessions by session IDs.

When ssl_certificate_by_lua* is specified at the same time,
this hook usually runs before ssl_certificate_by_lua*.
When the SSL session is found and successfully loaded for the current SSL connection,
SSL session resumption will happen and thus bypass the ssl_certificate_by_lua*
hook completely. In this case, NGINX also bypasses the ssl_session_store_by_lua*
hook, for obvious reasons.

To easily test this hook locally with a modern web browser, you can temporarily put the following line
in your https server block to disable the TLS session ticket support:

  1. ssl_session_tickets off;

But do not forget to comment this line out before publishing your site to the world.

If you are using the official pre-built packages for OpenResty
1.11.2.1 or later, then everything should work out of the box.

If you are using OpenSSL libraries not provided by OpenResty,
then you need to apply the following patch for OpenSSL 1.0.2h or later:

https://github.com/openresty/openresty/blob/master/patches/openssl-1.0.2h-sess_set_get_cb_yield.patch

If you are not using the NGINX core shipped with OpenResty 1.11.2.1 or later, then you need to
apply the following patch to the standard NGINX core 1.11.2 or later:

http://openresty.org/download/nginx-1.11.2-nonblocking_ssl_handshake_hooks.patch

This directive was first introduced in the v0.10.6 release.

Note that: this directive is only allowed to used in http context from the v0.10.7 release
(because SSL session resumption happens before server name dispatch).

Back to TOC

ssl_session_fetch_by_lua_file

syntax: ssl_session_fetch_by_lua_file <path-to-lua-script-file>

context: http

phase: right-before-SSL-handshake

Equivalent to ssl_session_fetch_by_lua_block, except that the file specified by <path-to-lua-script-file> contains the Lua code, or rather, the Lua/LuaJIT bytecode to be executed.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

This directive was first introduced in the v0.10.6 release.

Note that: this directive is only allowed to used in http context from the v0.10.7 release
(because SSL session resumption happens before server name dispatch).

Back to TOC

ssl_session_store_by_lua_block

syntax: ssl_session_store_by_lua_block { lua-script }

context: http

phase: right-after-SSL-handshake

This directive runs Lua code to fetch and save the SSL session (if any) according to the session ID
provided by the current SSL handshake request for the downstream. The saved or cached SSL
session data can be used for future SSL connections to resume SSL sessions without going
through the full SSL handshake process (which is very expensive in terms of CPU time).

Lua APIs that may yield, like ngx.sleep and cosockets,
are disabled in this context. You can still, however, use the ngx.timer.at API
to create 0-delay timers to save the SSL session data asynchronously to external services (like redis or memcached).

The Lua API for obtaining the current session ID and the associated session state data
is provided in the ngx.ssl.session
Lua module shipped with the lua-resty-core
library.

To easily test this hook locally with a modern web browser, you can temporarily put the following line
in your https server block to disable the TLS session ticket support:

  1. ssl_session_tickets off;

But do not forget to comment this line out before publishing your site to the world.

This directive was first introduced in the v0.10.6 release.

Note that: this directive is only allowed to used in http context from the v0.10.7 release
(because SSL session resumption happens before server name dispatch).

Back to TOC

ssl_session_store_by_lua_file

syntax: ssl_session_store_by_lua_file <path-to-lua-script-file>

context: http

phase: right-after-SSL-handshake

Equivalent to ssl_session_store_by_lua_block, except that the file specified by <path-to-lua-script-file> contains the Lua code, or rather, the Lua/LuaJIT bytecode to be executed.

When a relative path like foo/bar.lua is given, they will be turned into the absolute path relative to the server prefix path determined by the -p PATH command-line option while starting the Nginx server.

This directive was first introduced in the v0.10.6 release.

Note that: this directive is only allowed to used in http context from the v0.10.7 release
(because SSL session resumption happens before server name dispatch).

Back to TOC

lua_shared_dict

syntax: lua_shared_dict <name> <size>

default: no

context: http

phase: depends on usage

Declares a shared memory zone, <name>, to serve as storage for the shm based Lua dictionary ngx.shared.<name>.

Shared memory zones are always shared by all the nginx worker processes in the current nginx server instance.

The <size> argument accepts size units such as k and m:

  1. http {
  2. lua_shared_dict dogs 10m;
  3. ...
  4. }

The hard-coded minimum size is 8KB while the practical minimum size depends
on actual user data set (some people start with 12KB).

See ngx.shared.DICT for details.

This directive was first introduced in the v0.3.1rc22 release.

Back to TOC

lua_socket_connect_timeout

syntax: lua_socket_connect_timeout <time>

default: lua_socket_connect_timeout 60s

context: http, server, location

This directive controls the default timeout value used in TCP/unix-domain socket object’s connect method and can be overridden by the settimeout or settimeouts methods.

The <time> argument can be an integer, with an optional time unit, like s (second), ms (millisecond), m (minute). The default time unit is s, i.e., “second”. The default setting is 60s.

This directive was first introduced in the v0.5.0rc1 release.

Back to TOC

lua_socket_send_timeout

syntax: lua_socket_send_timeout <time>

default: lua_socket_send_timeout 60s

context: http, server, location

Controls the default timeout value used in TCP/unix-domain socket object’s send method and can be overridden by the settimeout or settimeouts methods.

The <time> argument can be an integer, with an optional time unit, like s (second), ms (millisecond), m (minute). The default time unit is s, i.e., “second”. The default setting is 60s.

This directive was first introduced in the v0.5.0rc1 release.

Back to TOC

lua_socket_send_lowat

syntax: lua_socket_send_lowat <size>

default: lua_socket_send_lowat 0

context: http, server, location

Controls the lowat (low water) value for the cosocket send buffer.

Back to TOC

lua_socket_read_timeout

syntax: lua_socket_read_timeout <time>

default: lua_socket_read_timeout 60s

context: http, server, location

phase: depends on usage

This directive controls the default timeout value used in TCP/unix-domain socket object’s receive method and iterator functions returned by the receiveuntil method. This setting can be overridden by the settimeout or settimeouts methods.

The <time> argument can be an integer, with an optional time unit, like s (second), ms (millisecond), m (minute). The default time unit is s, i.e., “second”. The default setting is 60s.

This directive was first introduced in the v0.5.0rc1 release.

Back to TOC

lua_socket_buffer_size

syntax: lua_socket_buffer_size <size>

default: lua_socket_buffer_size 4k/8k

context: http, server, location

Specifies the buffer size used by cosocket reading operations.

This buffer does not have to be that big to hold everything at the same time because cosocket supports 100% non-buffered reading and parsing. So even 1 byte buffer size should still work everywhere but the performance could be terrible.

This directive was first introduced in the v0.5.0rc1 release.

Back to TOC

lua_socket_pool_size

syntax: lua_socket_pool_size <size>

default: lua_socket_pool_size 30

context: http, server, location

Specifies the size limit (in terms of connection count) for every cosocket connection pool associated with every remote server (i.e., identified by either the host-port pair or the unix domain socket file path).

Default to 30 connections for every pool.

When the connection pool exceeds the available size limit, the least recently used (idle) connection already in the pool will be closed to make room for the current connection.

Note that the cosocket connection pool is per nginx worker process rather than per nginx server instance, so size limit specified here also applies to every single nginx worker process.

This directive was first introduced in the v0.5.0rc1 release.

Back to TOC

lua_socket_keepalive_timeout

syntax: lua_socket_keepalive_timeout <time>

default: lua_socket_keepalive_timeout 60s

context: http, server, location

This directive controls the default maximal idle time of the connections in the cosocket built-in connection pool. When this timeout reaches, idle connections will be closed and removed from the pool. This setting can be overridden by cosocket objects’ setkeepalive method.

The <time> argument can be an integer, with an optional time unit, like s (second), ms (millisecond), m (minute). The default time unit is s, i.e., “second”. The default setting is 60s.

This directive was first introduced in the v0.5.0rc1 release.

Back to TOC

lua_socket_log_errors

syntax: lua_socket_log_errors on|off

default: lua_socket_log_errors on

context: http, server, location

This directive can be used to toggle error logging when a failure occurs for the TCP or UDP cosockets. If you are already doing proper error handling and logging in your Lua code, then it is recommended to turn this directive off to prevent data flushing in your nginx error log files (which is usually rather expensive).

This directive was first introduced in the v0.5.13 release.

Back to TOC

lua_ssl_ciphers

syntax: lua_ssl_ciphers <ciphers>

default: lua_ssl_ciphers DEFAULT

context: http, server, location

Specifies the enabled ciphers for requests to a SSL/TLS server in the tcpsock:sslhandshake method. The ciphers are specified in the format understood by the OpenSSL library.

The full list can be viewed using the “openssl ciphers” command.

This directive was first introduced in the v0.9.11 release.

Back to TOC

lua_ssl_crl

syntax: lua_ssl_crl <file>

default: no

context: http, server, location

Specifies a file with revoked certificates (CRL) in the PEM format used to verify the certificate of the SSL/TLS server in the tcpsock:sslhandshake method.

This directive was first introduced in the v0.9.11 release.

Back to TOC

lua_ssl_protocols

syntax: lua_ssl_protocols [SSLv2] [SSLv3] [TLSv1] [TLSv1.1] [TLSv1.2] [TLSv1.3]

default: lua_ssl_protocols SSLv3 TLSv1 TLSv1.1 TLSv1.2

context: http, server, location

Enables the specified protocols for requests to a SSL/TLS server in the tcpsock:sslhandshake method.

The support for the TLSv1.3 parameter requires version v0.10.12 and OpenSSL 1.1.1.

This directive was first introduced in the v0.9.11 release.

Back to TOC

lua_ssl_trusted_certificate

syntax: lua_ssl_trusted_certificate <file>

default: no

context: http, server, location

Specifies a file path with trusted CA certificates in the PEM format used to verify the certificate of the SSL/TLS server in the tcpsock:sslhandshake method.

This directive was first introduced in the v0.9.11 release.

See also lua_ssl_verify_depth.

Back to TOC

lua_ssl_verify_depth

syntax: lua_ssl_verify_depth <number>

default: lua_ssl_verify_depth 1

context: http, server, location

Sets the verification depth in the server certificates chain.

This directive was first introduced in the v0.9.11 release.

See also lua_ssl_trusted_certificate.

Back to TOC

lua_http10_buffering

syntax: lua_http10_buffering on|off

default: lua_http10_buffering on

context: http, server, location, location-if

Enables or disables automatic response buffering for HTTP 1.0 (or older) requests. This buffering mechanism is mainly used for HTTP 1.0 keep-alive which relies on a proper Content-Length response header.

If the Lua code explicitly sets a Content-Length response header before sending the headers (either explicitly via ngx.send_headers or implicitly via the first ngx.say or ngx.print call), then the HTTP 1.0 response buffering will be disabled even when this directive is turned on.

To output very large response data in a streaming fashion (via the ngx.flush call, for example), this directive MUST be turned off to minimize memory usage.

This directive is turned on by default.

This directive was first introduced in the v0.5.0rc19 release.

Back to TOC

rewrite_by_lua_no_postpone

syntax: rewrite_by_lua_no_postpone on|off

default: rewrite_by_lua_no_postpone off

context: http

Controls whether or not to disable postponing rewrite_by_lua* directives to run at the end of the rewrite request-processing phase. By default, this directive is turned off and the Lua code is postponed to run at the end of the rewrite phase.

This directive was first introduced in the v0.5.0rc29 release.

Back to TOC

access_by_lua_no_postpone

syntax: access_by_lua_no_postpone on|off

default: access_by_lua_no_postpone off

context: http

Controls whether or not to disable postponing access_by_lua* directives to run at the end of the access request-processing phase. By default, this directive is turned off and the Lua code is postponed to run at the end of the access phase.

This directive was first introduced in the v0.9.20 release.

Back to TOC

lua_transform_underscores_in_response_headers

syntax: lua_transform_underscores_in_response_headers on|off

default: lua_transform_underscores_in_response_headers on

context: http, server, location, location-if

Controls whether to transform underscores (_) in the response header names specified in the ngx.header.HEADER API to hypens (-).

This directive was first introduced in the v0.5.0rc32 release.

Back to TOC

lua_check_client_abort

syntax: lua_check_client_abort on|off

default: lua_check_client_abort off

context: http, server, location, location-if

This directive controls whether to check for premature client connection abortion.

When this directive is on, the ngx_lua module will monitor the premature connection close event on the downstream connections and when there is such an event, it will call the user Lua function callback (registered by ngx.on_abort) or just stop and clean up all the Lua “light threads” running in the current request’s request handler when there is no user callback function registered.

According to the current implementation, however, if the client closes the connection before the Lua code finishes reading the request body data via ngx.req.socket, then ngx_lua will neither stop all the running “light threads” nor call the user callback (if ngx.on_abort has been called). Instead, the reading operation on ngx.req.socket will just return the error message “client aborted” as the second return value (the first return value is surely nil).

When TCP keepalive is disabled, it is relying on the client side to close the socket gracefully (by sending a FIN packet or something like that). For (soft) real-time web applications, it is highly recommended to configure the TCP keepalive support in your system’s TCP stack implementation in order to detect “half-open” TCP connections in time.

For example, on Linux, you can configure the standard listen directive in your nginx.conf file like this:

  1. listen 80 so_keepalive=2s:2s:8;

On FreeBSD, you can only tune the system-wide configuration for TCP keepalive, for example:

  1. # sysctl net.inet.tcp.keepintvl=2000
  2. # sysctl net.inet.tcp.keepidle=2000

This directive was first introduced in the v0.7.4 release.

See also ngx.on_abort.

Back to TOC

lua_max_pending_timers

syntax: lua_max_pending_timers <count>

default: lua_max_pending_timers 1024

context: http

Controls the maximum number of pending timers allowed.

Pending timers are those timers that have not expired yet.

When exceeding this limit, the ngx.timer.at call will immediately return nil and the error string “too many pending timers”.

This directive was first introduced in the v0.8.0 release.

Back to TOC

lua_max_running_timers

syntax: lua_max_running_timers <count>

default: lua_max_running_timers 256

context: http

Controls the maximum number of “running timers” allowed.

Running timers are those timers whose user callback functions are still running.

When exceeding this limit, Nginx will stop running the callbacks of newly expired timers and log an error message “N lua_max_running_timers are not enough” where “N” is the current value of this directive.

This directive was first introduced in the v0.8.0 release.

Back to TOC

lua_sa_restart

syntax: lua_sa_restart on|off

default: lua_sa_restart on

context: http

When enabled, this module will set the SA_RESTART flag on nginx workers signal dispositions.

This allows Lua I/O primitives to not be interrupted by nginx’s handling of various signals.

This directive was first introduced in the v0.10.14 release.

Back to TOC

Nginx API for Lua

Back to TOC

Introduction

The various *_by_lua, *_by_lua_block and *_by_lua_file configuration directives serve as gateways to the Lua API within the nginx.conf file. The Nginx Lua API described below can only be called within the user Lua code run in the context of these configuration directives.

The API is exposed to Lua in the form of two standard packages ngx and ndk. These packages are in the default global scope within ngx_lua and are always available within ngx_lua directives.

The packages can be introduced into external Lua modules like this:

  1. local say = ngx.say
  2. local _M = {}
  3. function _M.foo(a)
  4. say(a)
  5. end
  6. return _M

Use of the package.seeall flag is strongly discouraged due to its various bad side-effects.

It is also possible to directly require the packages in external Lua modules:

  1. local ngx = require "ngx"
  2. local ndk = require "ndk"

The ability to require these packages was introduced in the v0.2.1rc19 release.

Network I/O operations in user code should only be done through the Nginx Lua API calls as the Nginx event loop may be blocked and performance drop off dramatically otherwise. Disk operations with relatively small amount of data can be done using the standard Lua io library but huge file reading and writing should be avoided wherever possible as they may block the Nginx process significantly. Delegating all network and disk I/O operations to Nginx’s subrequests (via the ngx.location.capture method and similar) is strongly recommended for maximum performance.

Back to TOC

ngx.arg

syntax: val = ngx.arg[index]

context: set_by_lua*, body_filter_by_lua*

When this is used in the context of the set_by_lua* directives, this table is read-only and holds the input arguments to the config directives:

  1. value = ngx.arg[n]

Here is an example

  1. location /foo {
  2. set $a 32;
  3. set $b 56;
  4. set_by_lua $sum
  5. 'return tonumber(ngx.arg[1]) + tonumber(ngx.arg[2])'
  6. $a $b;
  7. echo $sum;
  8. }

that writes out 88, the sum of 32 and 56.

When this table is used in the context of body_filter_by_lua*, the first element holds the input data chunk to the output filter code and the second element holds the boolean flag for the “eof” flag indicating the end of the whole output data stream.

The data chunk and “eof” flag passed to the downstream Nginx output filters can also be overridden by assigning values directly to the corresponding table elements. When setting nil or an empty Lua string value to ngx.arg[1], no data chunk will be passed to the downstream Nginx output filters at all.

Back to TOC

ngx.var.VARIABLE

syntax: ngx.var.VAR_NAME

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Read and write Nginx variable values.

  1. value = ngx.var.some_nginx_variable_name
  2. ngx.var.some_nginx_variable_name = value

Note that only already defined nginx variables can be written to.
For example:

  1. location /foo {
  2. set $my_var ''; # this line is required to create $my_var at config time
  3. content_by_lua_block {
  4. ngx.var.my_var = 123;
  5. ...
  6. }
  7. }

That is, nginx variables cannot be created on-the-fly.

Some special nginx variables like $args and $limit_rate can be assigned a value,
many others are not, like $query_string, $arg_PARAMETER, and $http_NAME.

Nginx regex group capturing variables $1, $2, $3, and etc, can be read by this
interface as well, by writing ngx.var[1], ngx.var[2], ngx.var[3], and etc.

Setting ngx.var.Foo to a nil value will unset the $Foo Nginx variable.

  1. ngx.var.args = nil

CAUTION When reading from an Nginx variable, Nginx will allocate memory in the per-request memory pool which is freed only at request termination. So when you need to read from an Nginx variable repeatedly in your Lua code, cache the Nginx variable value to your own Lua variable, for example,

  1. local val = ngx.var.some_var
  2. --- use the val repeatedly later

to prevent (temporary) memory leaking within the current request’s lifetime. Another way of caching the result is to use the ngx.ctx table.

Undefined NGINX variables are evaluated to nil while uninitialized (but defined) NGINX variables are evaluated to an empty Lua string.

This API requires a relatively expensive metamethod call and it is recommended to avoid using it on hot code paths.

Back to TOC

Core constants

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, *log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

  1. ngx.OK (0)
  2. ngx.ERROR (-1)
  3. ngx.AGAIN (-2)
  4. ngx.DONE (-4)
  5. ngx.DECLINED (-5)

Note that only three of these constants are utilized by the Nginx API for Lua (i.e., ngx.exit accepts ngx.OK, ngx.ERROR, and ngx.DECLINED as input).

  1. ngx.null

The ngx.null constant is a NULL light userdata usually used to represent nil values in Lua tables etc and is similar to the lua-cjson library’s cjson.null constant. This constant was first introduced in the v0.5.0rc5 release.

The ngx.DECLINED constant was first introduced in the v0.5.0rc19 release.

Back to TOC

HTTP method constants

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

  1. ngx.HTTP_GET
  2. ngx.HTTP_HEAD
  3. ngx.HTTP_PUT
  4. ngx.HTTP_POST
  5. ngx.HTTP_DELETE
  6. ngx.HTTP_OPTIONS (added in the v0.5.0rc24 release)
  7. ngx.HTTP_MKCOL (added in the v0.8.2 release)
  8. ngx.HTTP_COPY (added in the v0.8.2 release)
  9. ngx.HTTP_MOVE (added in the v0.8.2 release)
  10. ngx.HTTP_PROPFIND (added in the v0.8.2 release)
  11. ngx.HTTP_PROPPATCH (added in the v0.8.2 release)
  12. ngx.HTTP_LOCK (added in the v0.8.2 release)
  13. ngx.HTTP_UNLOCK (added in the v0.8.2 release)
  14. ngx.HTTP_PATCH (added in the v0.8.2 release)
  15. ngx.HTTP_TRACE (added in the v0.8.2 release)

These constants are usually used in ngx.location.capture and ngx.location.capture_multi method calls.

Back to TOC

HTTP status constants

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

  1. value = ngx.HTTP_CONTINUE (100) (first added in the v0.9.20 release)
  2. value = ngx.HTTP_SWITCHING_PROTOCOLS (101) (first added in the v0.9.20 release)
  3. value = ngx.HTTP_OK (200)
  4. value = ngx.HTTP_CREATED (201)
  5. value = ngx.HTTP_ACCEPTED (202) (first added in the v0.9.20 release)
  6. value = ngx.HTTP_NO_CONTENT (204) (first added in the v0.9.20 release)
  7. value = ngx.HTTP_PARTIAL_CONTENT (206) (first added in the v0.9.20 release)
  8. value = ngx.HTTP_SPECIAL_RESPONSE (300)
  9. value = ngx.HTTP_MOVED_PERMANENTLY (301)
  10. value = ngx.HTTP_MOVED_TEMPORARILY (302)
  11. value = ngx.HTTP_SEE_OTHER (303)
  12. value = ngx.HTTP_NOT_MODIFIED (304)
  13. value = ngx.HTTP_TEMPORARY_REDIRECT (307) (first added in the v0.9.20 release)
  14. value = ngx.HTTP_PERMANENT_REDIRECT (308)
  15. value = ngx.HTTP_BAD_REQUEST (400)
  16. value = ngx.HTTP_UNAUTHORIZED (401)
  17. value = ngx.HTTP_PAYMENT_REQUIRED (402) (first added in the v0.9.20 release)
  18. value = ngx.HTTP_FORBIDDEN (403)
  19. value = ngx.HTTP_NOT_FOUND (404)
  20. value = ngx.HTTP_NOT_ALLOWED (405)
  21. value = ngx.HTTP_NOT_ACCEPTABLE (406) (first added in the v0.9.20 release)
  22. value = ngx.HTTP_REQUEST_TIMEOUT (408) (first added in the v0.9.20 release)
  23. value = ngx.HTTP_CONFLICT (409) (first added in the v0.9.20 release)
  24. value = ngx.HTTP_GONE (410)
  25. value = ngx.HTTP_UPGRADE_REQUIRED (426) (first added in the v0.9.20 release)
  26. value = ngx.HTTP_TOO_MANY_REQUESTS (429) (first added in the v0.9.20 release)
  27. value = ngx.HTTP_CLOSE (444) (first added in the v0.9.20 release)
  28. value = ngx.HTTP_ILLEGAL (451) (first added in the v0.9.20 release)
  29. value = ngx.HTTP_INTERNAL_SERVER_ERROR (500)
  30. value = ngx.HTTP_METHOD_NOT_IMPLEMENTED (501)
  31. value = ngx.HTTP_BAD_GATEWAY (502) (first added in the v0.9.20 release)
  32. value = ngx.HTTP_SERVICE_UNAVAILABLE (503)
  33. value = ngx.HTTP_GATEWAY_TIMEOUT (504) (first added in the v0.3.1rc38 release)
  34. value = ngx.HTTP_VERSION_NOT_SUPPORTED (505) (first added in the v0.9.20 release)
  35. value = ngx.HTTP_INSUFFICIENT_STORAGE (507) (first added in the v0.9.20 release)

Back to TOC

Nginx log level constants

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

  1. ngx.STDERR
  2. ngx.EMERG
  3. ngx.ALERT
  4. ngx.CRIT
  5. ngx.ERR
  6. ngx.WARN
  7. ngx.NOTICE
  8. ngx.INFO
  9. ngx.DEBUG

These constants are usually used by the ngx.log method.

Back to TOC

print

syntax: print(…)

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Writes argument values into the nginx error.log file with the ngx.NOTICE log level.

It is equivalent to

  1. ngx.log(ngx.NOTICE, ...)

Lua nil arguments are accepted and result in literal "nil" strings while Lua booleans result in literal "true" or "false" strings. And the ngx.null constant will yield the "null" string output.

There is a hard coded 2048 byte limitation on error message lengths in the Nginx core. This limit includes trailing newlines and leading time stamps. If the message size exceeds this limit, Nginx will truncate the message text accordingly. This limit can be manually modified by editing the NGX_MAX_ERROR_STR macro definition in the src/core/ngx_log.h file in the Nginx source tree.

Back to TOC

ngx.ctx

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*

This table can be used to store per-request Lua context data and has a life time identical to the current request (as with the Nginx variables).

Consider the following example,

  1. location /test {
  2. rewrite_by_lua_block {
  3. ngx.ctx.foo = 76
  4. }
  5. access_by_lua_block {
  6. ngx.ctx.foo = ngx.ctx.foo + 3
  7. }
  8. content_by_lua_block {
  9. ngx.say(ngx.ctx.foo)
  10. }
  11. }

Then GET /test will yield the output

  1. 79

That is, the ngx.ctx.foo entry persists across the rewrite, access, and content phases of a request.

Every request, including subrequests, has its own copy of the table. For example:

  1. location /sub {
  2. content_by_lua_block {
  3. ngx.say("sub pre: ", ngx.ctx.blah)
  4. ngx.ctx.blah = 32
  5. ngx.say("sub post: ", ngx.ctx.blah)
  6. }
  7. }
  8. location /main {
  9. content_by_lua_block {
  10. ngx.ctx.blah = 73
  11. ngx.say("main pre: ", ngx.ctx.blah)
  12. local res = ngx.location.capture("/sub")
  13. ngx.print(res.body)
  14. ngx.say("main post: ", ngx.ctx.blah)
  15. }
  16. }

Then GET /main will give the output

  1. main pre: 73
  2. sub pre: nil
  3. sub post: 32
  4. main post: 73

Here, modification of the ngx.ctx.blah entry in the subrequest does not affect the one in the parent request. This is because they have two separate versions of ngx.ctx.blah.

Internal redirection will destroy the original request ngx.ctx data (if any) and the new request will have an empty ngx.ctx table. For instance,

  1. location /new {
  2. content_by_lua_block {
  3. ngx.say(ngx.ctx.foo)
  4. }
  5. }
  6. location /orig {
  7. content_by_lua_block {
  8. ngx.ctx.foo = "hello"
  9. ngx.exec("/new")
  10. }
  11. }

Then GET /orig will give

  1. nil

rather than the original "hello" value.

Arbitrary data values, including Lua closures and nested tables, can be inserted into this “magic” table. It also allows the registration of custom meta methods.

Overriding ngx.ctx with a new Lua table is also supported, for example,

  1. ngx.ctx = { foo = 32, bar = 54 }

When being used in the context of init_worker_by_lua*, this table just has the same lifetime of the current Lua handler.

The ngx.ctx lookup requires relatively expensive metamethod calls and it is much slower than explicitly passing per-request data along by your own function arguments. So do not abuse this API for saving your own function arguments because it usually has quite some performance impact.

Because of the metamethod magic, never “local” the ngx.ctx table outside your Lua function scope on the Lua module level due to worker-level data sharing. For example, the following is bad:

  1. -- mymodule.lua
  2. local _M = {}
  3. -- the following line is bad since ngx.ctx is a per-request
  4. -- data while this <code>ctx</code> variable is on the Lua module level
  5. -- and thus is per-nginx-worker.
  6. local ctx = ngx.ctx
  7. function _M.main()
  8. ctx.foo = "bar"
  9. end
  10. return _M

Use the following instead:

  1. -- mymodule.lua
  2. local _M = {}
  3. function _M.main(ctx)
  4. ctx.foo = "bar"
  5. end
  6. return _M

That is, let the caller pass the ctx table explicitly via a function argument.

Back to TOC

ngx.location.capture

syntax: res = ngx.location.capture(uri, options?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Issues a synchronous but still non-blocking Nginx Subrequest using uri.

Nginx’s subrequests provide a powerful way to make non-blocking internal requests to other locations configured with disk file directory or any other nginx C modules like ngx_proxy, ngx_fastcgi, ngx_memc,
ngx_postgres, ngx_drizzle, and even ngx_lua itself and etc etc etc.

Also note that subrequests just mimic the HTTP interface but there is no extra HTTP/TCP traffic nor IPC involved. Everything works internally, efficiently, on the C level.

Subrequests are completely different from HTTP 301/302 redirection (via ngx.redirect) and internal redirection (via ngx.exec).

You should always read the request body (by either calling ngx.req.read_body or configuring lua_need_request_body on) before initiating a subrequest.

This API function (as well as ngx.location.capture_multi) always buffers the whole response body of the subrequest in memory. Thus, you should use cosockets
and streaming processing instead if you have to handle large subrequest responses.

Here is a basic example:

  1. res = ngx.location.capture(uri)

Returns a Lua table with 4 slots: res.status, res.header, res.body, and res.truncated.

res.status holds the response status code for the subrequest response.

res.header holds all the response headers of the
subrequest and it is a normal Lua table. For multi-value response headers,
the value is a Lua (array) table that holds all the values in the order that
they appear. For instance, if the subrequest response headers contain the following
lines:

  1. Set-Cookie: a=3
  2. Set-Cookie: foo=bar
  3. Set-Cookie: baz=blah

Then res.header["Set-Cookie"] will be evaluated to the table value
{"a=3", "foo=bar", "baz=blah"}.

res.body holds the subrequest’s response body data, which might be truncated. You always need to check the res.truncated boolean flag to see if res.body contains truncated data. The data truncation here can only be caused by those unrecoverable errors in your subrequests like the cases that the remote end aborts the connection prematurely in the middle of the response body data stream or a read timeout happens when your subrequest is receiving the response body data from the remote.

URI query strings can be concatenated to URI itself, for instance,

  1. res = ngx.location.capture('/foo/bar?a=3&b=4')

Named locations like @foo are not allowed due to a limitation in
the nginx core. Use normal locations combined with the internal directive to
prepare internal-only locations.

An optional option table can be fed as the second
argument, which supports the options:

  • method
    specify the subrequest’s request method, which only accepts constants like ngx.HTTP_POST.
  • body
    specify the subrequest’s request body (string value only).
  • args
    specify the subrequest’s URI query arguments (both string value and Lua tables are accepted)
  • ctx
    specify a Lua table to be the ngx.ctx table for the subrequest. It can be the current request’s ngx.ctx table, which effectively makes the parent and its subrequest to share exactly the same context table. This option was first introduced in the v0.3.1rc25 release.
  • vars
    take a Lua table which holds the values to set the specified Nginx variables in the subrequest as this option’s value. This option was first introduced in the v0.3.1rc31 release.
  • copy_all_vars
    specify whether to copy over all the Nginx variable values of the current request to the subrequest in question. modifications of the nginx variables in the subrequest will not affect the current (parent) request. This option was first introduced in the v0.3.1rc31 release.
  • share_all_vars
    specify whether to share all the Nginx variables of the subrequest with the current (parent) request. modifications of the Nginx variables in the subrequest will affect the current (parent) request. Enabling this option may lead to hard-to-debug issues due to bad side-effects and is considered bad and harmful. Only enable this option when you completely know what you are doing.
  • always_forward_body
    when set to true, the current (parent) request’s request body will always be forwarded to the subrequest being created if the body option is not specified. The request body read by either ngx.req.read_body() or lua_need_request_body on will be directly forwarded to the subrequest without copying the whole request body data when creating the subrequest (no matter the request body data is buffered in memory buffers or temporary files). By default, this option is false and when the body option is not specified, the request body of the current (parent) request is only forwarded when the subrequest takes the PUT or POST request method.

Issuing a POST subrequest, for example, can be done as follows

  1. res = ngx.location.capture(
  2. '/foo/bar',
  3. { method = ngx.HTTP_POST, body = 'hello, world' }
  4. )

See HTTP method constants methods other than POST.
The method option is ngx.HTTP_GET by default.

The args option can specify extra URI arguments, for instance,

  1. ngx.location.capture('/foo?a=1',
  2. { args = { b = 3, c = ':' } }
  3. )

is equivalent to

  1. ngx.location.capture('/foo?a=1&b=3&c=%3a')

that is, this method will escape argument keys and values according to URI rules and
concatenate them together into a complete query string. The format for the Lua table passed as the args argument is identical to the format used in the ngx.encode_args method.

The args option can also take plain query strings:

  1. ngx.location.capture('/foo?a=1',
  2. { args = 'b=3&c=%3a' } }
  3. )

This is functionally identical to the previous examples.

The share_all_vars option controls whether to share nginx variables among the current request and its subrequests.
If this option is set to true, then the current request and associated subrequests will share the same Nginx variable scope. Hence, changes to Nginx variables made by a subrequest will affect the current request.

Care should be taken in using this option as variable scope sharing can have unexpected side effects. The args, vars, or copy_all_vars options are generally preferable instead.

This option is set to false by default

  1. location /other {
  2. set $dog "$dog world";
  3. echo "$uri dog: $dog";
  4. }
  5. location /lua {
  6. set $dog 'hello';
  7. content_by_lua_block {
  8. res = ngx.location.capture("/other",
  9. { share_all_vars = true });
  10. ngx.print(res.body)
  11. ngx.say(ngx.var.uri, ": ", ngx.var.dog)
  12. }
  13. }

Accessing location /lua gives

  1. /other dog: hello world
  2. /lua: hello world

The copy_all_vars option provides a copy of the parent request’s Nginx variables to subrequests when such subrequests are issued. Changes made to these variables by such subrequests will not affect the parent request or any other subrequests sharing the parent request’s variables.

  1. location /other {
  2. set $dog "$dog world";
  3. echo "$uri dog: $dog";
  4. }
  5. location /lua {
  6. set $dog 'hello';
  7. content_by_lua_block {
  8. res = ngx.location.capture("/other",
  9. { copy_all_vars = true });
  10. ngx.print(res.body)
  11. ngx.say(ngx.var.uri, ": ", ngx.var.dog)
  12. }
  13. }

Request GET /lua will give the output

  1. /other dog: hello world
  2. /lua: hello

Note that if both share_all_vars and copy_all_vars are set to true, then share_all_vars takes precedence.

In addition to the two settings above, it is possible to specify
values for variables in the subrequest using the vars option. These
variables are set after the sharing or copying of variables has been
evaluated, and provides a more efficient method of passing specific
values to a subrequest over encoding them as URL arguments and
unescaping them in the Nginx config file.

  1. location /other {
  2. content_by_lua_block {
  3. ngx.say("dog = ", ngx.var.dog)
  4. ngx.say("cat = ", ngx.var.cat)
  5. }
  6. }
  7. location /lua {
  8. set $dog '';
  9. set $cat '';
  10. content_by_lua_block {
  11. res = ngx.location.capture("/other",
  12. { vars = { dog = "hello", cat = 32 }});
  13. ngx.print(res.body)
  14. }
  15. }

Accessing /lua will yield the output

  1. dog = hello
  2. cat = 32

The ctx option can be used to specify a custom Lua table to serve as the ngx.ctx table for the subrequest.

  1. location /sub {
  2. content_by_lua_block {
  3. ngx.ctx.foo = "bar";
  4. }
  5. }
  6. location /lua {
  7. content_by_lua_block {
  8. local ctx = {}
  9. res = ngx.location.capture("/sub", { ctx = ctx })
  10. ngx.say(ctx.foo);
  11. ngx.say(ngx.ctx.foo);
  12. }
  13. }

Then request GET /lua gives

  1. bar
  2. nil

It is also possible to use this ctx option to share the same ngx.ctx table between the current (parent) request and the subrequest:

  1. location /sub {
  2. content_by_lua_block {
  3. ngx.ctx.foo = "bar";
  4. }
  5. }
  6. location /lua {
  7. content_by_lua_block {
  8. res = ngx.location.capture("/sub", { ctx = ngx.ctx })
  9. ngx.say(ngx.ctx.foo);
  10. }
  11. }

Request GET /lua yields the output

  1. bar

Note that subrequests issued by ngx.location.capture inherit all the
request headers of the current request by default and that this may have unexpected side effects on the
subrequest responses. For example, when using the standard ngx_proxy module to serve
subrequests, an “Accept-Encoding: gzip” header in the main request may result
in gzipped responses that cannot be handled properly in Lua code. Original request headers should be ignored by setting
proxy_pass_request_headers to off in subrequest locations.

When the body option is not specified and the always_forward_body option is false (the default value), the POST and PUT subrequests will inherit the request bodies of the parent request (if any).

There is a hard-coded upper limit on the number of concurrent subrequests possible for every main request. In older versions of Nginx, the limit was 50 concurrent subrequests and in more recent versions, Nginx 1.1.x onwards, this was increased to 200 concurrent subrequests. When this limit is exceeded, the following error message is added to the error.log file:

  1. [error] 13983#0: *1 subrequests cycle while processing "/uri"

The limit can be manually modified if required by editing the definition of the NGX_HTTP_MAX_SUBREQUESTS macro in the nginx/src/http/ngx_http_request.h file in the Nginx source tree.

Please also refer to restrictions on capturing locations configured by subrequest directives of other modules.

Back to TOC

ngx.location.capture_multi

syntax: res1, res2, … = ngx.location.capture_multi({ {uri, options?}, {uri, options?}, … })

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Just like ngx.location.capture, but supports multiple subrequests running in parallel.

This function issues several parallel subrequests specified by the input table and returns their results in the same order. For example,

  1. res1, res2, res3 = ngx.location.capture_multi{
  2. { "/foo", { args = "a=3&b=4" } },
  3. { "/bar" },
  4. { "/baz", { method = ngx.HTTP_POST, body = "hello" } },
  5. }
  6. if res1.status == ngx.HTTP_OK then
  7. ...
  8. end
  9. if res2.body == "BLAH" then
  10. ...
  11. end

This function will not return until all the subrequests terminate.
The total latency is the longest latency of the individual subrequests rather than the sum.

Lua tables can be used for both requests and responses when the number of subrequests to be issued is not known in advance:

  1. -- construct the requests table
  2. local reqs = {}
  3. table.insert(reqs, { "/mysql" })
  4. table.insert(reqs, { "/postgres" })
  5. table.insert(reqs, { "/redis" })
  6. table.insert(reqs, { "/memcached" })
  7. -- issue all the requests at once and wait until they all return
  8. local resps = { ngx.location.capture_multi(reqs) }
  9. -- loop over the responses table
  10. for i, resp in ipairs(resps) do
  11. -- process the response table "resp"
  12. end

The ngx.location.capture function is just a special form
of this function. Logically speaking, the ngx.location.capture can be implemented like this

  1. ngx.location.capture =
  2. function (uri, args)
  3. return ngx.location.capture_multi({ {uri, args} })
  4. end

Please also refer to restrictions on capturing locations configured by subrequest directives of other modules.

Back to TOC

ngx.status

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Read and write the current request’s response status. This should be called
before sending out the response headers.

  1. ngx.status = ngx.HTTP_CREATED
  2. status = ngx.status

Setting ngx.status after the response header is sent out has no effect but leaving an error message in your nginx’s error log file:

  1. attempt to set ngx.status after sending out response headers

Back to TOC

ngx.header.HEADER

syntax: ngx.header.HEADER = VALUE

syntax: value = ngx.header.HEADER

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Set, add to, or clear the current request’s HEADER response header that is to be sent.

Underscores (_) in the header names will be replaced by hyphens (-) by default. This transformation can be turned off via the lua_transform_underscores_in_response_headers directive.

The header names are matched case-insensitively.

  1. -- equivalent to ngx.header["Content-Type"] = 'text/plain'
  2. ngx.header.content_type = 'text/plain';
  3. ngx.header["X-My-Header"] = 'blah blah';

Multi-value headers can be set this way:

  1. ngx.header['Set-Cookie'] = {'a=32; path=/', 'b=4; path=/'}

will yield

  1. Set-Cookie: a=32; path=/
  2. Set-Cookie: b=4; path=/

in the response headers.

Only Lua tables are accepted (Only the last element in the table will take effect for standard headers such as Content-Type that only accept a single value).

  1. ngx.header.content_type = {'a', 'b'}

is equivalent to

  1. ngx.header.content_type = 'b'

Setting a slot to nil effectively removes it from the response headers:

  1. ngx.header["X-My-Header"] = nil;

The same applies to assigning an empty table:

  1. ngx.header["X-My-Header"] = {};

Setting ngx.header.HEADER after sending out response headers (either explicitly with ngx.send_headers or implicitly with ngx.print and similar) will log an error message.

Reading ngx.header.HEADER will return the value of the response header named HEADER.

Underscores (_) in the header names will also be replaced by dashes (-) and the header names will be matched case-insensitively. If the response header is not present at all, nil will be returned.

This is particularly useful in the context of header_filter_by_lua*, for example,

  1. location /test {
  2. set $footer '';
  3. proxy_pass http://some-backend;
  4. header_filter_by_lua_block {
  5. if ngx.header["X-My-Header"] == "blah" then
  6. ngx.var.footer = "some value"
  7. end
  8. }
  9. echo_after_body $footer;
  10. }

For multi-value headers, all of the values of header will be collected in order and returned as a Lua table. For example, response headers

  1. Foo: bar
  2. Foo: baz

will result in

  1. {"bar", "baz"}

to be returned when reading ngx.header.Foo.

Note that ngx.header is not a normal Lua table and as such, it is not possible to iterate through it using the Lua ipairs function.

For reading request headers, use the ngx.req.get_headers function instead.

Back to TOC

ngx.resp.get_headers

syntax: headers, err = ngx.resp.get_headers(max_headers?, raw?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, balancer_by_lua*

Returns a Lua table holding all the current response headers for the current request.

  1. local h, err = ngx.resp.get_headers()
  2. if err == "truncated" then
  3. -- one can choose to ignore or reject the current response here
  4. end
  5. for k, v in pairs(h) do
  6. ...
  7. end

This function has the same signature as ngx.req.get_headers except getting response headers instead of request headers.

Note that a maximum of 100 response headers are parsed by default (including those with the same name) and that additional response headers are silently discarded to guard against potential denial of service attacks. Since v0.10.13, when the limit is exceeded, it will return a second value which is the string "truncated".

This API was first introduced in the v0.9.5 release.

Back to TOC

ngx.req.is_internal

syntax: is_internal = ngx.req.is_internal()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Returns a boolean indicating whether the current request is an “internal request”, i.e.,
a request initiated from inside the current nginx server instead of from the client side.

Subrequests are all internal requests and so are requests after internal redirects.

This API was first introduced in the v0.9.20 release.

Back to TOC

ngx.req.start_time

syntax: secs = ngx.req.start_time()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Returns a floating-point number representing the timestamp (including milliseconds as the decimal part) when the current request was created.

The following example emulates the $request_time variable value (provided by ngx_http_log_module) in pure Lua:

  1. local request_time = ngx.now() - ngx.req.start_time()

This function was first introduced in the v0.7.7 release.

See also ngx.now and ngx.update_time.

Back to TOC

ngx.req.http_version

syntax: num = ngx.req.http_version()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*

Returns the HTTP version number for the current request as a Lua number.

Current possible values are 2.0, 1.0, 1.1, and 0.9. Returns nil for unrecognized values.

This method was first introduced in the v0.7.17 release.

Back to TOC

ngx.req.raw_header

syntax: str = ngx.req.raw_header(no_request_line?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*

Returns the original raw HTTP protocol header received by the Nginx server.

By default, the request line and trailing CR LF terminator will also be included. For example,

  1. ngx.print(ngx.req.raw_header())

gives something like this:

  1. GET /t HTTP/1.1
  2. Host: localhost
  3. Connection: close
  4. Foo: bar

You can specify the optional
no_request_line argument as a true value to exclude the request line from the result. For example,

  1. ngx.print(ngx.req.raw_header(true))

outputs something like this:

  1. Host: localhost
  2. Connection: close
  3. Foo: bar

This method was first introduced in the v0.7.17 release.

This method does not work in HTTP/2 requests yet.

Back to TOC

ngx.req.get_method

syntax: method_name = ngx.req.get_method()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, balancer_by_lua*

Retrieves the current request’s request method name. Strings like "GET" and "POST" are returned instead of numerical method constants.

If the current request is an Nginx subrequest, then the subrequest’s method name will be returned.

This method was first introduced in the v0.5.6 release.

See also ngx.req.set_method.

Back to TOC

ngx.req.set_method

syntax: ngx.req.set_method(method_id)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*

Overrides the current request’s request method with the method_id argument. Currently only numerical method constants are supported, like ngx.HTTP_POST and ngx.HTTP_GET.

If the current request is an Nginx subrequest, then the subrequest’s method will be overridden.

This method was first introduced in the v0.5.6 release.

See also ngx.req.get_method.

Back to TOC

ngx.req.set_uri

syntax: ngx.req.set_uri(uri, jump?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*

Rewrite the current request’s (parsed) URI by the uri argument. The uri argument must be a Lua string and cannot be of zero length, or a Lua exception will be thrown.

The optional boolean jump argument can trigger location rematch (or location jump) as ngx_http_rewrite_module‘s rewrite directive, that is, when jump is true (default to false), this function will never return and it will tell Nginx to try re-searching locations with the new URI value at the later post-rewrite phase and jumping to the new location.

Location jump will not be triggered otherwise, and only the current request’s URI will be modified, which is also the default behavior. This function will return but with no returned values when the jump argument is false or absent altogether.

For example, the following nginx config snippet

  1. rewrite ^ /foo last;

can be coded in Lua like this:

  1. ngx.req.set_uri("/foo", true)

Similarly, Nginx config

  1. rewrite ^ /foo break;

can be coded in Lua as

  1. ngx.req.set_uri("/foo", false)

or equivalently,

  1. ngx.req.set_uri("/foo")

The jump argument can only be set to true in rewrite_by_lua*. Use of jump in other contexts is prohibited and will throw out a Lua exception.

A more sophisticated example involving regex substitutions is as follows

  1. location /test {
  2. rewrite_by_lua_block {
  3. local uri = ngx.re.sub(ngx.var.uri, "^/test/(.*)", "/$1", "o")
  4. ngx.req.set_uri(uri)
  5. }
  6. proxy_pass http://my_backend;
  7. }

which is functionally equivalent to

  1. location /test {
  2. rewrite ^/test/(.*) /$1 break;
  3. proxy_pass http://my_backend;
  4. }

Note that it is not possible to use this interface to rewrite URI arguments and that ngx.req.set_uri_args should be used for this instead. For instance, Nginx config

  1. rewrite ^ /foo?a=3? last;

can be coded as

  1. ngx.req.set_uri_args("a=3")
  2. ngx.req.set_uri("/foo", true)

or

  1. ngx.req.set_uri_args({a = 3})
  2. ngx.req.set_uri("/foo", true)

This interface was first introduced in the v0.3.1rc14 release.

Back to TOC

ngx.req.set_uri_args

syntax: ngx.req.set_uri_args(args)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*

Rewrite the current request’s URI query arguments by the args argument. The args argument can be either a Lua string, as in

  1. ngx.req.set_uri_args("a=3&b=hello%20world")

or a Lua table holding the query arguments’ key-value pairs, as in

  1. ngx.req.set_uri_args({ a = 3, b = "hello world" })

where in the latter case, this method will escape argument keys and values according to the URI escaping rule.

Multi-value arguments are also supported:

  1. ngx.req.set_uri_args({ a = 3, b = {5, 6} })

which will result in a query string like a=3&b=5&b=6.

This interface was first introduced in the v0.3.1rc13 release.

See also ngx.req.set_uri.

Back to TOC

ngx.req.get_uri_args

syntax: args, err = ngx.req.get_uri_args(max_args?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, balancer_by_lua*

Returns a Lua table holding all the current request URL query arguments.

  1. location = /test {
  2. content_by_lua_block {
  3. local args, err = ngx.req.get_uri_args()
  4. if err == "truncated" then
  5. -- one can choose to ignore or reject the current request here
  6. end
  7. for key, val in pairs(args) do
  8. if type(val) == "table" then
  9. ngx.say(key, ": ", table.concat(val, ", "))
  10. else
  11. ngx.say(key, ": ", val)
  12. end
  13. end
  14. }
  15. }

Then GET /test?foo=bar&bar=baz&bar=blah will yield the response body

  1. foo: bar
  2. bar: baz, blah

Multiple occurrences of an argument key will result in a table value holding all the values for that key in order.

Keys and values are unescaped according to URI escaping rules. In the settings above, GET /test?a%20b=1%61+2 will yield:

  1. a b: 1a 2

Arguments without the =<value> parts are treated as boolean arguments. GET /test?foo&bar will yield:

  1. foo: true
  2. bar: true

That is, they will take Lua boolean values true. However, they are different from arguments taking empty string values. GET /test?foo=&bar= will give something like

  1. foo:
  2. bar:

Empty key arguments are discarded. GET /test?=hello&=world will yield an empty output for instance.

Updating query arguments via the nginx variable $args (or ngx.var.args in Lua) at runtime is also supported:

  1. ngx.var.args = "a=3&b=42"
  2. local args, err = ngx.req.get_uri_args()

Here the args table will always look like

  1. {a = 3, b = 42}

regardless of the actual request query string.

Note that a maximum of 100 request arguments are parsed by default (including those with the same name) and that additional request arguments are silently discarded to guard against potential denial of service attacks. Since v0.10.13, when the limit is exceeded, it will return a second value which is the string "truncated".

However, the optional max_args function argument can be used to override this limit:

  1. local args, err = ngx.req.get_uri_args(10)
  2. if err == "truncated" then
  3. -- one can choose to ignore or reject the current request here
  4. end

This argument can be set to zero to remove the limit and to process all request arguments received:

  1. local args, err = ngx.req.get_uri_args(0)

Removing the max_args cap is strongly discouraged.

Back to TOC

ngx.req.get_post_args

syntax: args, err = ngx.req.get_post_args(max_args?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Returns a Lua table holding all the current request POST query arguments (of the MIME type application/x-www-form-urlencoded). Call ngx.req.read_body to read the request body first or turn on the lua_need_request_body directive to avoid errors.

  1. location = /test {
  2. content_by_lua_block {
  3. ngx.req.read_body()
  4. local args, err = ngx.req.get_post_args()
  5. if err == "truncated" then
  6. -- one can choose to ignore or reject the current request here
  7. end
  8. if not args then
  9. ngx.say("failed to get post args: ", err)
  10. return
  11. end
  12. for key, val in pairs(args) do
  13. if type(val) == "table" then
  14. ngx.say(key, ": ", table.concat(val, ", "))
  15. else
  16. ngx.say(key, ": ", val)
  17. end
  18. end
  19. }
  20. }

Then

  1. # Post request with the body 'foo=bar&bar=baz&bar=blah'
  2. $ curl --data 'foo=bar&bar=baz&bar=blah' localhost/test

will yield the response body like

  1. foo: bar
  2. bar: baz, blah

Multiple occurrences of an argument key will result in a table value holding all of the values for that key in order.

Keys and values will be unescaped according to URI escaping rules.

With the settings above,

  1. # POST request with body 'a%20b=1%61+2'
  2. $ curl -d 'a%20b=1%61+2' localhost/test

will yield:

  1. a b: 1a 2

Arguments without the =<value> parts are treated as boolean arguments. POST /test with the request body foo&bar will yield:

  1. foo: true
  2. bar: true

That is, they will take Lua boolean values true. However, they are different from arguments taking empty string values. POST /test with request body foo=&bar= will return something like

  1. foo:
  2. bar:

Empty key arguments are discarded. POST /test with body =hello&=world will yield empty outputs for instance.

Note that a maximum of 100 request arguments are parsed by default (including those with the same name) and that additional request arguments are silently discarded to guard against potential denial of service attacks. Since v0.10.13, when the limit is exceeded, it will return a second value which is the string "truncated".

However, the optional max_args function argument can be used to override this limit:

  1. local args, err = ngx.req.get_post_args(10)
  2. if err == "truncated" then
  3. -- one can choose to ignore or reject the current request here
  4. end

This argument can be set to zero to remove the limit and to process all request arguments received:

  1. local args, err = ngx.req.get_post_args(0)

Removing the max_args cap is strongly discouraged.

Back to TOC

ngx.req.get_headers

syntax: headers, err = ngx.req.get_headers(max_headers?, raw?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Returns a Lua table holding all the current request headers.

  1. local h, err = ngx.req.get_headers()
  2. if err == "truncated" then
  3. -- one can choose to ignore or reject the current request here
  4. end
  5. for k, v in pairs(h) do
  6. ...
  7. end

To read an individual header:

  1. ngx.say("Host: ", ngx.req.get_headers()["Host"])

Note that the ngx.var.HEADER API call, which uses core $http_HEADER variables, may be more preferable for reading individual request headers.

For multiple instances of request headers such as:

  1. Foo: foo
  2. Foo: bar
  3. Foo: baz

the value of ngx.req.get_headers()["Foo"] will be a Lua (array) table such as:

  1. {"foo", "bar", "baz"}

Note that a maximum of 100 request headers are parsed by default (including those with the same name) and that additional request headers are silently discarded to guard against potential denial of service attacks. Since v0.10.13, when the limit is exceeded, it will return a second value which is the string "truncated".

However, the optional max_headers function argument can be used to override this limit:

  1. local headers, err = ngx.req.get_headers(10)
  2. if err == "truncated" then
  3. -- one can choose to ignore or reject the current request here
  4. end

This argument can be set to zero to remove the limit and to process all request headers received:

  1. local headers, err = ngx.req.get_headers(0)

Removing the max_headers cap is strongly discouraged.

Since the 0.6.9 release, all the header names in the Lua table returned are converted to the pure lower-case form by default, unless the raw argument is set to true (default to false).

Also, by default, an __index metamethod is added to the resulting Lua table and will normalize the keys to a pure lowercase form with all underscores converted to dashes in case of a lookup miss. For example, if a request header My-Foo-Header is present, then the following invocations will all pick up the value of this header correctly:

  1. ngx.say(headers.my_foo_header)
  2. ngx.say(headers["My-Foo-Header"])
  3. ngx.say(headers["my-foo-header"])

The __index metamethod will not be added when the raw argument is set to true.

Back to TOC

ngx.req.set_header

syntax: ngx.req.set_header(header_name, header_value)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*

Set the current request’s request header named header_name to value header_value, overriding any existing ones.

By default, all the subrequests subsequently initiated by ngx.location.capture and ngx.location.capture_multi will inherit the new header.

Here is an example of setting the Content-Type header:

  1. ngx.req.set_header("Content-Type", "text/css")

The header_value can take an array list of values,
for example,

  1. ngx.req.set_header("Foo", {"a", "abc"})

will produce two new request headers:

  1. Foo: a
  2. Foo: abc

and old Foo headers will be overridden if there is any.

When the header_value argument is nil, the request header will be removed. So

  1. ngx.req.set_header("X-Foo", nil)

is equivalent to

  1. ngx.req.clear_header("X-Foo")

Back to TOC

ngx.req.clear_header

syntax: ngx.req.clear_header(header_name)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*

Clears the current request’s request header named header_name. None of the current request’s existing subrequests will be affected but subsequently initiated subrequests will inherit the change by default.

Back to TOC

ngx.req.read_body

syntax: ngx.req.read_body()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Reads the client request body synchronously without blocking the Nginx event loop.

  1. ngx.req.read_body()
  2. local args = ngx.req.get_post_args()

If the request body is already read previously by turning on lua_need_request_body or by using other modules, then this function does not run and returns immediately.

If the request body has already been explicitly discarded, either by the ngx.req.discard_body function or other modules, this function does not run and returns immediately.

In case of errors, such as connection errors while reading the data, this method will throw out a Lua exception or terminate the current request with a 500 status code immediately.

The request body data read using this function can be retrieved later via ngx.req.get_body_data or, alternatively, the temporary file name for the body data cached to disk using ngx.req.get_body_file. This depends on

  1. whether the current request body is already larger than the client_body_buffer_size,
  2. and whether client_body_in_file_only has been switched on.

In cases where current request may have a request body and the request body data is not required, The ngx.req.discard_body function must be used to explicitly discard the request body to avoid breaking things under HTTP 1.1 keepalive or HTTP 1.1 pipelining.

This function was first introduced in the v0.3.1rc17 release.

Back to TOC

ngx.req.discard_body

syntax: ngx.req.discard_body()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Explicitly discard the request body, i.e., read the data on the connection and throw it away immediately (without using the request body by any means).

This function is an asynchronous call and returns immediately.

If the request body has already been read, this function does nothing and returns immediately.

This function was first introduced in the v0.3.1rc17 release.

See also ngx.req.read_body.

Back to TOC

ngx.req.get_body_data

syntax: data = ngx.req.get_body_data()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, log_by_lua*

Retrieves in-memory request body data. It returns a Lua string rather than a Lua table holding all the parsed query arguments. Use the ngx.req.get_post_args function instead if a Lua table is required.

This function returns nil if

  1. the request body has not been read,
  2. the request body has been read into disk temporary files,
  3. or the request body has zero size.

If the request body has not been read yet, call ngx.req.read_body first (or turn on lua_need_request_body to force this module to read the request body. This is not recommended however).

If the request body has been read into disk files, try calling the ngx.req.get_body_file function instead.

To force in-memory request bodies, try setting client_body_buffer_size to the same size value in client_max_body_size.

Note that calling this function instead of using ngx.var.request_body or ngx.var.echo_request_body is more efficient because it can save one dynamic memory allocation and one data copy.

This function was first introduced in the v0.3.1rc17 release.

See also ngx.req.get_body_file.

Back to TOC

ngx.req.get_body_file

syntax: file_name = ngx.req.get_body_file()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Retrieves the file name for the in-file request body data. Returns nil if the request body has not been read or has been read into memory.

The returned file is read only and is usually cleaned up by Nginx’s memory pool. It should not be manually modified, renamed, or removed in Lua code.

If the request body has not been read yet, call ngx.req.read_body first (or turn on lua_need_request_body to force this module to read the request body. This is not recommended however).

If the request body has been read into memory, try calling the ngx.req.get_body_data function instead.

To force in-file request bodies, try turning on client_body_in_file_only.

This function was first introduced in the v0.3.1rc17 release.

See also ngx.req.get_body_data.

Back to TOC

ngx.req.set_body_data

syntax: ngx.req.set_body_data(data)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Set the current request’s request body using the in-memory data specified by the data argument.

If the request body has not been read yet, call ngx.req.read_body first (or turn on lua_need_request_body to force this module to read the request body. This is not recommended however). Additionally, the request body must not have been previously discarded by ngx.req.discard_body.

Whether the previous request body has been read into memory or buffered into a disk file, it will be freed or the disk file will be cleaned up immediately, respectively.

This function was first introduced in the v0.3.1rc18 release.

See also ngx.req.set_body_file.

Back to TOC

ngx.req.set_body_file

syntax: ngx.req.set_body_file(file_name, auto_clean?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Set the current request’s request body using the in-file data specified by the file_name argument.

If the request body has not been read yet, call ngx.req.read_body first (or turn on lua_need_request_body to force this module to read the request body. This is not recommended however). Additionally, the request body must not have been previously discarded by ngx.req.discard_body.

If the optional auto_clean argument is given a true value, then this file will be removed at request completion or the next time this function or ngx.req.set_body_data are called in the same request. The auto_clean is default to false.

Please ensure that the file specified by the file_name argument exists and is readable by an Nginx worker process by setting its permission properly to avoid Lua exception errors.

Whether the previous request body has been read into memory or buffered into a disk file, it will be freed or the disk file will be cleaned up immediately, respectively.

This function was first introduced in the v0.3.1rc18 release.

See also ngx.req.set_body_data.

Back to TOC

ngx.req.init_body

syntax: ngx.req.init_body(buffer_size?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*

Creates a new blank request body for the current request and inializes the buffer for later request body data writing via the ngx.req.append_body and ngx.req.finish_body APIs.

If the buffer_size argument is specified, then its value will be used for the size of the memory buffer for body writing with ngx.req.append_body. If the argument is omitted, then the value specified by the standard client_body_buffer_size directive will be used instead.

When the data can no longer be hold in the memory buffer for the request body, then the data will be flushed onto a temporary file just like the standard request body reader in the Nginx core.

It is important to always call the ngx.req.finish_body after all the data has been appended onto the current request body. Also, when this function is used together with ngx.req.socket, it is required to call ngx.req.socket before this function, or you will get the “request body already exists” error message.

The usage of this function is often like this:

  1. ngx.req.init_body(128 * 1024) -- buffer is 128KB
  2. for chunk in next_data_chunk() do
  3. ngx.req.append_body(chunk) -- each chunk can be 4KB
  4. end
  5. ngx.req.finish_body()

This function can be used with ngx.req.append_body, ngx.req.finish_body, and ngx.req.socket to implement efficient input filters in pure Lua (in the context of rewrite_by_lua* or access_by_lua*), which can be used with other Nginx content handler or upstream modules like ngx_http_proxy_module and ngx_http_fastcgi_module.

This function was first introduced in the v0.5.11 release.

Back to TOC

ngx.req.append_body

syntax: ngx.req.append_body(data_chunk)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*

Append new data chunk specified by the data_chunk argument onto the existing request body created by the ngx.req.init_body call.

When the data can no longer be hold in the memory buffer for the request body, then the data will be flushed onto a temporary file just like the standard request body reader in the Nginx core.

It is important to always call the ngx.req.finish_body after all the data has been appended onto the current request body.

This function can be used with ngx.req.init_body, ngx.req.finish_body, and ngx.req.socket to implement efficient input filters in pure Lua (in the context of rewrite_by_lua* or access_by_lua*), which can be used with other Nginx content handler or upstream modules like ngx_http_proxy_module and ngx_http_fastcgi_module.

This function was first introduced in the v0.5.11 release.

See also ngx.req.init_body.

Back to TOC

ngx.req.finish_body

syntax: ngx.req.finish_body()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*

Completes the construction process of the new request body created by the ngx.req.init_body and ngx.req.append_body calls.

This function can be used with ngx.req.init_body, ngx.req.append_body, and ngx.req.socket to implement efficient input filters in pure Lua (in the context of rewrite_by_lua* or access_by_lua*), which can be used with other Nginx content handler or upstream modules like ngx_http_proxy_module and ngx_http_fastcgi_module.

This function was first introduced in the v0.5.11 release.

See also ngx.req.init_body.

Back to TOC

ngx.req.socket

syntax: tcpsock, err = ngx.req.socket()

syntax: tcpsock, err = ngx.req.socket(raw)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Returns a read-only cosocket object that wraps the downstream connection. Only receive and receiveuntil methods are supported on this object.

In case of error, nil will be returned as well as a string describing the error.

The socket object returned by this method is usually used to read the current request’s body in a streaming fashion. Do not turn on the lua_need_request_body directive, and do not mix this call with ngx.req.read_body and ngx.req.discard_body.

If any request body data has been pre-read into the Nginx core request header buffer, the resulting cosocket object will take care of this to avoid potential data loss resulting from such pre-reading.
Chunked request bodies are not yet supported in this API.

Since the v0.9.0 release, this function accepts an optional boolean raw argument. When this argument is true, this function returns a full-duplex cosocket object wrapping around the raw downstream connection socket, upon which you can call the receive, receiveuntil, and send methods.

When the raw argument is true, it is required that no pending data from any previous ngx.say, ngx.print, or ngx.send_headers calls exists. So if you have these downstream output calls previously, you should call ngx.flush(true) before calling ngx.req.socket(true) to ensure that there is no pending output data. If the request body has not been read yet, then this “raw socket” can also be used to read the request body.

You can use the “raw request socket” returned by ngx.req.socket(true) to implement fancy protocols like WebSocket, or just emit your own raw HTTP response header or body data. You can refer to the lua-resty-websocket library for a real world example.

This function was first introduced in the v0.5.0rc1 release.

Back to TOC

ngx.exec

syntax: ngx.exec(uri, args?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Does an internal redirect to uri with args and is similar to the echo_exec directive of the echo-nginx-module.

  1. ngx.exec('/some-location');
  2. ngx.exec('/some-location', 'a=3&b=5&c=6');
  3. ngx.exec('/some-location?a=3&b=5', 'c=6');

The optional second args can be used to specify extra URI query arguments, for example:

  1. ngx.exec("/foo", "a=3&b=hello%20world")

Alternatively, a Lua table can be passed for the args argument for ngx_lua to carry out URI escaping and string concatenation.

  1. ngx.exec("/foo", { a = 3, b = "hello world" })

The result is exactly the same as the previous example.

The format for the Lua table passed as the args argument is identical to the format used in the ngx.encode_args method.

Named locations are also supported but the second args argument will be ignored if present and the querystring for the new target is inherited from the referring location (if any).

GET /foo/file.php?a=hello will return “hello” and not “goodbye” in the example below

  1. location /foo {
  2. content_by_lua_block {
  3. ngx.exec("@bar", "a=goodbye");
  4. }
  5. }
  6. location @bar {
  7. content_by_lua_block {
  8. local args = ngx.req.get_uri_args()
  9. for key, val in pairs(args) do
  10. if key == "a" then
  11. ngx.say(val)
  12. end
  13. end
  14. }
  15. }

Note that the ngx.exec method is different from ngx.redirect in that
it is purely an internal redirect and that no new external HTTP traffic is involved.

Also note that this method call terminates the processing of the current request and that it must be called before ngx.send_headers or explicit response body
outputs by either ngx.print or ngx.say.

It is recommended that a coding style that combines this method call with the return statement, i.e., return ngx.exec(...) be adopted when this method call is used in contexts other than header_filter_by_lua* to reinforce the fact that the request processing is being terminated.

Back to TOC

ngx.redirect

syntax: ngx.redirect(uri, status?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Issue an HTTP 301 or 302 redirection to uri.

The optional status parameter specifies the HTTP status code to be used. The following status codes are supported right now:

  • 301
  • 302 (default)
  • 303
  • 307
  • 308

It is 302 (ngx.HTTP_MOVED_TEMPORARILY) by default.

Here is an example assuming the current server name is localhost and that it is listening on port 1984:

  1. return ngx.redirect("/foo")

which is equivalent to

  1. return ngx.redirect("/foo", ngx.HTTP_MOVED_TEMPORARILY)

Redirecting arbitrary external URLs is also supported, for example:

  1. return ngx.redirect("http://www.google.com")

We can also use the numerical code directly as the second status argument:

  1. return ngx.redirect("/foo", 301)

This method is similar to the rewrite directive with the redirect modifier in the standard
ngx_http_rewrite_module, for example, this nginx.conf snippet

  1. rewrite ^ /foo? redirect; # nginx config

is equivalent to the following Lua code

  1. return ngx.redirect('/foo'); -- Lua code

while

  1. rewrite ^ /foo? permanent; # nginx config

is equivalent to

  1. return ngx.redirect('/foo', ngx.HTTP_MOVED_PERMANENTLY) -- Lua code

URI arguments can be specified as well, for example:

  1. return ngx.redirect('/foo?a=3&b=4')

Note that this method call terminates the processing of the current request and that it must be called before ngx.send_headers or explicit response body
outputs by either ngx.print or ngx.say.

It is recommended that a coding style that combines this method call with the return statement, i.e., return ngx.redirect(...) be adopted when this method call is used in contexts other than header_filter_by_lua* to reinforce the fact that the request processing is being terminated.

Back to TOC

ngx.send_headers

syntax: ok, err = ngx.send_headers()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Explicitly send out the response headers.

Since v0.8.3 this function returns 1 on success, or returns nil and a string describing the error otherwise.

Note that there is normally no need to manually send out response headers as ngx_lua will automatically send headers out
before content is output with ngx.say or ngx.print or when content_by_lua* exits normally.

Back to TOC

ngx.headers_sent

syntax: value = ngx.headers_sent

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*

Returns true if the response headers have been sent (by ngx_lua), and false otherwise.

This API was first introduced in ngx_lua v0.3.1rc6.

Back to TOC

ngx.print

syntax: ok, err = ngx.print(…)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Emits arguments concatenated to the HTTP client (as response body). If response headers have not been sent, this function will send headers out first and then output body data.

Since v0.8.3 this function returns 1 on success, or returns nil and a string describing the error otherwise.

Lua nil values will output "nil" strings and Lua boolean values will output "true" and "false" literal strings respectively.

Nested arrays of strings are permitted and the elements in the arrays will be sent one by one:

  1. local table = {
  2. "hello, ",
  3. {"world: ", true, " or ", false,
  4. {": ", nil}}
  5. }
  6. ngx.print(table)

will yield the output

  1. hello, world: true or false: nil

Non-array table arguments will cause a Lua exception to be thrown.

The ngx.null constant will yield the "null" string output.

This is an asynchronous call and will return immediately without waiting for all the data to be written into the system send buffer. To run in synchronous mode, call ngx.flush(true) after calling ngx.print. This can be particularly useful for streaming output. See ngx.flush for more details.

Please note that both ngx.print and ngx.say will always invoke the whole Nginx output body filter chain, which is an expensive operation. So be careful when calling either of these two in a tight loop; buffer the data yourself in Lua and save the calls.

Back to TOC

ngx.say

syntax: ok, err = ngx.say(…)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Just as ngx.print but also emit a trailing newline.

Back to TOC

ngx.log

syntax: ngx.log(log_level, …)

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Log arguments concatenated to error.log with the given logging level.

Lua nil arguments are accepted and result in literal "nil" string while Lua booleans result in literal "true" or "false" string outputs. And the ngx.null constant will yield the "null" string output.

The log_level argument can take constants like ngx.ERR and ngx.WARN. Check out Nginx log level constants for details.

There is a hard coded 2048 byte limitation on error message lengths in the Nginx core. This limit includes trailing newlines and leading time stamps. If the message size exceeds this limit, Nginx will truncate the message text accordingly. This limit can be manually modified by editing the NGX_MAX_ERROR_STR macro definition in the src/core/ngx_log.h file in the Nginx source tree.

Back to TOC

ngx.flush

syntax: ok, err = ngx.flush(wait?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Flushes response output to the client.

ngx.flush accepts an optional boolean wait argument (Default: false) first introduced in the v0.3.1rc34 release. When called with the default argument, it issues an asynchronous call (Returns immediately without waiting for output data to be written into the system send buffer). Calling the function with the wait argument set to true switches to synchronous mode.

In synchronous mode, the function will not return until all output data has been written into the system send buffer or until the send_timeout setting has expired. Note that using the Lua coroutine mechanism means that this function does not block the Nginx event loop even in the synchronous mode.

When ngx.flush(true) is called immediately after ngx.print or ngx.say, it causes the latter functions to run in synchronous mode. This can be particularly useful for streaming output.

Note that ngx.flush is not functional when in the HTTP 1.0 output buffering mode. See HTTP 1.0 support.

Since v0.8.3 this function returns 1 on success, or returns nil and a string describing the error otherwise.

Back to TOC

ngx.exit

syntax: ngx.exit(status)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

When status >= 200 (i.e., ngx.HTTP_OK and above), it will interrupt the execution of the current request and return status code to nginx.

When status == 0 (i.e., ngx.OK), it will only quit the current phase handler (or the content handler if the content_by_lua* directive is used) and continue to run later phases (if any) for the current request.

The status argument can be ngx.OK, ngx.ERROR, ngx.HTTP_NOT_FOUND,
ngx.HTTP_MOVED_TEMPORARILY, or other HTTP status constants.

To return an error page with custom contents, use code snippets like this:

  1. ngx.status = ngx.HTTP_GONE
  2. ngx.say("This is our own content")
  3. -- to cause quit the whole request rather than the current phase handler
  4. ngx.exit(ngx.HTTP_OK)

The effect in action:

  1. $ curl -i http://localhost/test
  2. HTTP/1.1 410 Gone
  3. Server: nginx/1.0.6
  4. Date: Thu, 15 Sep 2011 00:51:48 GMT
  5. Content-Type: text/plain
  6. Transfer-Encoding: chunked
  7. Connection: keep-alive
  8. This is our own content

Number literals can be used directly as the argument, for instance,

  1. ngx.exit(501)

Note that while this method accepts all HTTP status constants as input, it only accepts ngx.OK and ngx.ERROR of the core constants.

Also note that this method call terminates the processing of the current request and that it is recommended that a coding style that combines this method call with the return statement, i.e., return ngx.exit(...) be used to reinforce the fact that the request processing is being terminated.

When being used in the contexts of header_filter_by_lua*, balancer_by_lua*, and
ssl_session_store_by_lua*, ngx.exit() is
an asynchronous operation and will return immediately. This behavior may change in future and it is recommended that users always use return in combination as suggested above.

Back to TOC

ngx.eof

syntax: ok, err = ngx.eof()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Explicitly specify the end of the response output stream. In the case of HTTP 1.1 chunked encoded output, it will just trigger the Nginx core to send out the “last chunk”.

When you disable the HTTP 1.1 keep-alive feature for your downstream connections, you can rely on well written HTTP clients to close the connection actively for you when you call this method. This trick can be used do back-ground jobs without letting the HTTP clients to wait on the connection, as in the following example:

  1. location = /async {
  2. keepalive_timeout 0;
  3. content_by_lua_block {
  4. ngx.say("got the task!")
  5. ngx.eof() -- well written HTTP clients will close the connection at this point
  6. -- access MySQL, PostgreSQL, Redis, Memcached, and etc here...
  7. }
  8. }

But if you create subrequests to access other locations configured by Nginx upstream modules, then you should configure those upstream modules to ignore client connection abortions if they are not by default. For example, by default the standard ngx_http_proxy_module will terminate both the subrequest and the main request as soon as the client closes the connection, so it is important to turn on the proxy_ignore_client_abort directive in your location block configured by ngx_http_proxy_module:

  1. proxy_ignore_client_abort on;

A better way to do background jobs is to use the ngx.timer.at API.

Since v0.8.3 this function returns 1 on success, or returns nil and a string describing the error otherwise.

Back to TOC

ngx.sleep

syntax: ngx.sleep(seconds)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Sleeps for the specified seconds without blocking. One can specify time resolution up to 0.001 seconds (i.e., one milliseconds).

Behind the scene, this method makes use of the Nginx timers.

Since the 0.7.20 release, The 0 time argument can also be specified.

This method was introduced in the 0.5.0rc30 release.

Back to TOC

ngx.escape_uri

syntax: newstr = ngx.escape_uri(str)

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Escape str as a URI component.

Back to TOC

ngx.unescape_uri

syntax: newstr = ngx.unescape_uri(str)

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*

Unescape str as an escaped URI component.

For example,

  1. ngx.say(ngx.unescape_uri("b%20r56+7"))

gives the output

  1. b r56 7

Back to TOC

ngx.encode_args

syntax: str = ngx.encode_args(table)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*

Encode the Lua table to a query args string according to the URI encoded rules.

For example,

  1. ngx.encode_args({foo = 3, ["b r"] = "hello world"})

yields

  1. foo=3&b%20r=hello%20world

The table keys must be Lua strings.

Multi-value query args are also supported. Just use a Lua table for the argument’s value, for example:

  1. ngx.encode_args({baz = {32, "hello"}})

gives

  1. baz=32&baz=hello

If the value table is empty and the effect is equivalent to the nil value.

Boolean argument values are also supported, for instance,

  1. ngx.encode_args({a = true, b = 1})

yields

  1. a&b=1

If the argument value is false, then the effect is equivalent to the nil value.

This method was first introduced in the v0.3.1rc27 release.

Back to TOC

ngx.decode_args

syntax: table, err = ngx.decode_args(str, max_args?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Decodes a URI encoded query-string into a Lua table. This is the inverse function of ngx.encode_args.

The optional max_args argument can be used to specify the maximum number of arguments parsed from the str argument. By default, a maximum of 100 request arguments are parsed (including those with the same name) and that additional URI arguments are silently discarded to guard against potential denial of service attacks. Since v0.10.13, when the limit is exceeded, it will return a second value which is the string "truncated".

This argument can be set to zero to remove the limit and to process all request arguments received:

  1. local args = ngx.decode_args(str, 0)

Removing the max_args cap is strongly discouraged.

This method was introduced in the v0.5.0rc29.

Back to TOC

ngx.encode_base64

syntax: newstr = ngx.encode_base64(str, no_padding?)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Encodes str to a base64 digest.

Since the 0.9.16 release, an optional boolean-typed no_padding argument can be specified to control whether the base64 padding should be appended to the resulting digest (default to false, i.e., with padding enabled).

Back to TOC

ngx.decode_base64

syntax: newstr = ngx.decode_base64(str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Decodes the str argument as a base64 digest to the raw form. Returns nil if str is not well formed.

Back to TOC

ngx.crc32_short

syntax: intval = ngx.crc32_short(str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Calculates the CRC-32 (Cyclic Redundancy Code) digest for the str argument.

This method performs better on relatively short str inputs (i.e., less than 30 ~ 60 bytes), as compared to ngx.crc32_long. The result is exactly the same as ngx.crc32_long.

Behind the scene, it is just a thin wrapper around the ngx_crc32_short function defined in the Nginx core.

This API was first introduced in the v0.3.1rc8 release.

Back to TOC

ngx.crc32_long

syntax: intval = ngx.crc32_long(str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Calculates the CRC-32 (Cyclic Redundancy Code) digest for the str argument.

This method performs better on relatively long str inputs (i.e., longer than 30 ~ 60 bytes), as compared to ngx.crc32_short. The result is exactly the same as ngx.crc32_short.

Behind the scene, it is just a thin wrapper around the ngx_crc32_long function defined in the Nginx core.

This API was first introduced in the v0.3.1rc8 release.

Back to TOC

ngx.hmac_sha1

syntax: digest = ngx.hmac_sha1(secret_key, str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Computes the HMAC-SHA1 digest of the argument str and turns the result using the secret key <secret_key>.

The raw binary form of the HMAC-SHA1 digest will be generated, use ngx.encode_base64, for example, to encode the result to a textual representation if desired.

For example,

  1. local key = "thisisverysecretstuff"
  2. local src = "some string we want to sign"
  3. local digest = ngx.hmac_sha1(key, src)
  4. ngx.say(ngx.encode_base64(digest))

yields the output

  1. R/pvxzHC4NLtj7S+kXFg/NePTmk=

This API requires the OpenSSL library enabled in the Nginx build (usually by passing the --with-http_ssl_module option to the ./configure script).

This function was first introduced in the v0.3.1rc29 release.

Back to TOC

ngx.md5

syntax: digest = ngx.md5(str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the hexadecimal representation of the MD5 digest of the str argument.

For example,

  1. location = /md5 {
  2. content_by_lua_block { ngx.say(ngx.md5("hello")) }
  3. }

yields the output

  1. 5d41402abc4b2a76b9719d911017c592

See ngx.md5_bin if the raw binary MD5 digest is required.

Back to TOC

ngx.md5_bin

syntax: digest = ngx.md5_bin(str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the binary form of the MD5 digest of the str argument.

See ngx.md5 if the hexadecimal form of the MD5 digest is required.

Back to TOC

ngx.sha1_bin

syntax: digest = ngx.sha1_bin(str)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the binary form of the SHA-1 digest of the str argument.

This function requires SHA-1 support in the Nginx build. (This usually just means OpenSSL should be installed while building Nginx).

This function was first introduced in the v0.5.0rc6.

Back to TOC

ngx.quote_sql_str

syntax: quoted_value = ngx.quote_sql_str(raw_value)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns a quoted SQL string literal according to the MySQL quoting rules.

Back to TOC

ngx.today

syntax: str = ngx.today()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns current date (in the format yyyy-mm-dd) from the nginx cached time (no syscall involved unlike Lua’s date library).

This is the local time.

Back to TOC

ngx.time

syntax: secs = ngx.time()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the elapsed seconds from the epoch for the current time stamp from the nginx cached time (no syscall involved unlike Lua’s date library).

Updates of the Nginx time cache can be forced by calling ngx.update_time first.

Back to TOC

ngx.now

syntax: secs = ngx.now()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns a floating-point number for the elapsed time in seconds (including milliseconds as the decimal part) from the epoch for the current time stamp from the nginx cached time (no syscall involved unlike Lua’s date library).

You can forcibly update the Nginx time cache by calling ngx.update_time first.

This API was first introduced in v0.3.1rc32.

Back to TOC

ngx.update_time

syntax: ngx.update_time()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Forcibly updates the Nginx current time cache. This call involves a syscall and thus has some overhead, so do not abuse it.

This API was first introduced in v0.3.1rc32.

Back to TOC

ngx.localtime

syntax: str = ngx.localtime()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the current time stamp (in the format yyyy-mm-dd hh:mm:ss) of the nginx cached time (no syscall involved unlike Lua’s os.date function).

This is the local time.

Back to TOC

ngx.utctime

syntax: str = ngx.utctime()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the current time stamp (in the format yyyy-mm-dd hh:mm:ss) of the nginx cached time (no syscall involved unlike Lua’s os.date function).

This is the UTC time.

Back to TOC

syntax: str = ngx.cookie_time(sec)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns a formatted string can be used as the cookie expiration time. The parameter sec is the time stamp in seconds (like those returned from ngx.time).

  1. ngx.say(ngx.cookie_time(1290079655))
  2. -- yields "Thu, 18-Nov-10 11:27:35 GMT"

Back to TOC

ngx.http_time

syntax: str = ngx.http_time(sec)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns a formated string can be used as the http header time (for example, being used in Last-Modified header). The parameter sec is the time stamp in seconds (like those returned from ngx.time).

  1. ngx.say(ngx.http_time(1290079655))
  2. -- yields "Thu, 18 Nov 2010 11:27:35 GMT"

Back to TOC

ngx.parse_http_time

syntax: sec = ngx.parse_http_time(str)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Parse the http time string (as returned by ngx.http_time) into seconds. Returns the seconds or nil if the input string is in bad forms.

  1. local time = ngx.parse_http_time("Thu, 18 Nov 2010 11:27:35 GMT")
  2. if time == nil then
  3. ...
  4. end

Back to TOC

ngx.is_subrequest

syntax: value = ngx.is_subrequest

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*

Returns true if the current request is an nginx subrequest, or false otherwise.

Back to TOC

ngx.re.match

syntax: captures, err = ngx.re.match(subject, regex, options?, ctx?, res_table?)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Matches the subject string using the Perl compatible regular expression regex with the optional options.

Only the first occurrence of the match is returned, or nil if no match is found. In case of errors, like seeing a bad regular expression or exceeding the PCRE stack limit, nil and a string describing the error will be returned.

When a match is found, a Lua table captures is returned, where captures[0] holds the whole substring being matched, and captures[1] holds the first parenthesized sub-pattern’s capturing, captures[2] the second, and so on.

  1. local m, err = ngx.re.match("hello, 1234", "[0-9]+")
  2. if m then
  3. -- m[0] == "1234"
  4. else
  5. if err then
  6. ngx.log(ngx.ERR, "error: ", err)
  7. return
  8. end
  9. ngx.say("match not found")
  10. end
  1. local m, err = ngx.re.match("hello, 1234", "([0-9])[0-9]+")
  2. -- m[0] == "1234"
  3. -- m[1] == "1"

Named captures are also supported since the v0.7.14 release
and are returned in the same Lua table as key-value pairs as the numbered captures.

  1. local m, err = ngx.re.match("hello, 1234", "([0-9])(?<remaining>[0-9]+)")
  2. -- m[0] == "1234"
  3. -- m[1] == "1"
  4. -- m[2] == "234"
  5. -- m["remaining"] == "234"

Unmatched subpatterns will have false values in their captures table fields.

  1. local m, err = ngx.re.match("hello, world", "(world)|(hello)|(?<named>howdy)")
  2. -- m[0] == "hello"
  3. -- m[1] == false
  4. -- m[2] == "hello"
  5. -- m[3] == false
  6. -- m["named"] == false

Specify options to control how the match operation will be performed. The following option characters are supported:

  1. a anchored mode (only match from the beginning)
  2. d enable the DFA mode (or the longest token match semantics).
  3. this requires PCRE 6.0+ or else a Lua exception will be thrown.
  4. first introduced in ngx_lua v0.3.1rc30.
  5. D enable duplicate named pattern support. This allows named
  6. subpattern names to be repeated, returning the captures in
  7. an array-like Lua table. for example,
  8. local m = ngx.re.match("hello, world",
  9. "(?<named>\w+), (?<named>\w+)",
  10. "D")
  11. -- m["named"] == {"hello", "world"}
  12. this option was first introduced in the v0.7.14 release.
  13. this option requires at least PCRE 8.12.
  14. i case insensitive mode (similar to Perl's /i modifier)
  15. j enable PCRE JIT compilation, this requires PCRE 8.21+ which
  16. must be built with the --enable-jit option. for optimum performance,
  17. this option should always be used together with the 'o' option.
  18. first introduced in ngx_lua v0.3.1rc30.
  19. J enable the PCRE Javascript compatible mode. this option was
  20. first introduced in the v0.7.14 release. this option requires
  21. at least PCRE 8.12.
  22. m multi-line mode (similar to Perl's /m modifier)
  23. o compile-once mode (similar to Perl's /o modifier),
  24. to enable the worker-process-level compiled-regex cache
  25. s single-line mode (similar to Perl's /s modifier)
  26. u UTF-8 mode. this requires PCRE to be built with
  27. the --enable-utf8 option or else a Lua exception will be thrown.
  28. U similar to "u" but disables PCRE's UTF-8 validity check on
  29. the subject string. first introduced in ngx_lua v0.8.1.
  30. x extended mode (similar to Perl's /x modifier)

These options can be combined:

  1. local m, err = ngx.re.match("hello, world", "HEL LO", "ix")
  2. -- m[0] == "hello"
  1. local m, err = ngx.re.match("hello, 美好生活", "HELLO, (.{2})", "iu")
  2. -- m[0] == "hello, 美好"
  3. -- m[1] == "美好"

The o option is useful for performance tuning, because the regex pattern in question will only be compiled once, cached in the worker-process level, and shared among all requests in the current Nginx worker process. The upper limit of the regex cache can be tuned via the lua_regex_cache_max_entries directive.

The optional fourth argument, ctx, can be a Lua table holding an optional pos field. When the pos field in the ctx table argument is specified, ngx.re.match will start matching from that offset (starting from 1). Regardless of the presence of the pos field in the ctx table, ngx.re.match will always set this pos field to the position after the substring matched by the whole pattern in case of a successful match. When match fails, the ctx table will be left intact.

  1. local ctx = {}
  2. local m, err = ngx.re.match("1234, hello", "[0-9]+", "", ctx)
  3. -- m[0] = "1234"
  4. -- ctx.pos == 5
  1. local ctx = { pos = 2 }
  2. local m, err = ngx.re.match("1234, hello", "[0-9]+", "", ctx)
  3. -- m[0] = "234"
  4. -- ctx.pos == 5

The ctx table argument combined with the a regex modifier can be used to construct a lexer atop ngx.re.match.

Note that, the options argument is not optional when the ctx argument is specified and that the empty Lua string ("") must be used as placeholder for options if no meaningful regex options are required.

This method requires the PCRE library enabled in Nginx. (Known Issue With Special Escaping Sequences).

To confirm that PCRE JIT is enabled, activate the Nginx debug log by adding the --with-debug option to Nginx or OpenResty’s ./configure script. Then, enable the “debug” error log level in error_log directive. The following message will be generated if PCRE JIT is enabled:

  1. pcre JIT compiling result: 1

Starting from the 0.9.4 release, this function also accepts a 5th argument, res_table, for letting the caller supply the Lua table used to hold all the capturing results. Starting from 0.9.6, it is the caller’s responsibility to ensure this table is empty. This is very useful for recycling Lua tables and saving GC and table allocation overhead.

This feature was introduced in the v0.2.1rc11 release.

Back to TOC

ngx.re.find

syntax: from, to, err = ngx.re.find(subject, regex, options?, ctx?, nth?)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to ngx.re.match but only returns the beginning index (from) and end index (to) of the matched substring. The returned indexes are 1-based and can be fed directly into the string.sub API function to obtain the matched substring.

In case of errors (like bad regexes or any PCRE runtime errors), this API function returns two nil values followed by a string describing the error.

If no match is found, this function just returns a nil value.

Below is an example:

  1. local s = "hello, 1234"
  2. local from, to, err = ngx.re.find(s, "([0-9]+)", "jo")
  3. if from then
  4. ngx.say("from: ", from)
  5. ngx.say("to: ", to)
  6. ngx.say("matched: ", string.sub(s, from, to))
  7. else
  8. if err then
  9. ngx.say("error: ", err)
  10. return
  11. end
  12. ngx.say("not matched!")
  13. end

This example produces the output

  1. from: 8
  2. to: 11
  3. matched: 1234

Because this API function does not create new Lua strings nor new Lua tables, it is much faster than ngx.re.match. It should be used wherever possible.

Since the 0.9.3 release, an optional 5th argument, nth, is supported to specify which (submatch) capture’s indexes to return. When nth is 0 (which is the default), the indexes for the whole matched substring is returned; when nth is 1, then the 1st submatch capture’s indexes are returned; when nth is 2, then the 2nd submatch capture is returned, and so on. When the specified submatch does not have a match, then two nil values will be returned. Below is an example for this:

  1. local str = "hello, 1234"
  2. local from, to = ngx.re.find(str, "([0-9])([0-9]+)", "jo", nil, 2)
  3. if from then
  4. ngx.say("matched 2nd submatch: ", string.sub(str, from, to)) -- yields "234"
  5. end

This API function was first introduced in the v0.9.2 release.

Back to TOC

ngx.re.gmatch

syntax: iterator, err = ngx.re.gmatch(subject, regex, options?)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to ngx.re.match, but returns a Lua iterator instead, so as to let the user programmer iterate all the matches over the <subject> string argument with the PCRE regex.

In case of errors, like seeing an ill-formed regular expression, nil and a string describing the error will be returned.

Here is a small example to demonstrate its basic usage:

  1. local iterator, err = ngx.re.gmatch("hello, world!", "([a-z]+)", "i")
  2. if not iterator then
  3. ngx.log(ngx.ERR, "error: ", err)
  4. return
  5. end
  6. local m
  7. m, err = iterator() -- m[0] == m[1] == "hello"
  8. if err then
  9. ngx.log(ngx.ERR, "error: ", err)
  10. return
  11. end
  12. m, err = iterator() -- m[0] == m[1] == "world"
  13. if err then
  14. ngx.log(ngx.ERR, "error: ", err)
  15. return
  16. end
  17. m, err = iterator() -- m == nil
  18. if err then
  19. ngx.log(ngx.ERR, "error: ", err)
  20. return
  21. end

More often we just put it into a Lua loop:

  1. local it, err = ngx.re.gmatch("hello, world!", "([a-z]+)", "i")
  2. if not it then
  3. ngx.log(ngx.ERR, "error: ", err)
  4. return
  5. end
  6. while true do
  7. local m, err = it()
  8. if err then
  9. ngx.log(ngx.ERR, "error: ", err)
  10. return
  11. end
  12. if not m then
  13. -- no match found (any more)
  14. break
  15. end
  16. -- found a match
  17. ngx.say(m[0])
  18. ngx.say(m[1])
  19. end

The optional options argument takes exactly the same semantics as the ngx.re.match method.

The current implementation requires that the iterator returned should only be used in a single request. That is, one should not assign it to a variable belonging to persistent namespace like a Lua package.

This method requires the PCRE library enabled in Nginx. (Known Issue With Special Escaping Sequences).

This feature was first introduced in the v0.2.1rc12 release.

Back to TOC

ngx.re.sub

syntax: newstr, n, err = ngx.re.sub(subject, regex, replace, options?)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Substitutes the first match of the Perl compatible regular expression regex on the subject argument string with the string or function argument replace. The optional options argument has exactly the same meaning as in ngx.re.match.

This method returns the resulting new string as well as the number of successful substitutions. In case of failures, like syntax errors in the regular expressions or the <replace> string argument, it will return nil and a string describing the error.

When the replace is a string, then it is treated as a special template for string replacement. For example,

  1. local newstr, n, err = ngx.re.sub("hello, 1234", "([0-9])[0-9]", "[$0][$1]")
  2. if newstr then
  3. -- newstr == "hello, [12][1]34"
  4. -- n == 1
  5. else
  6. ngx.log(ngx.ERR, "error: ", err)
  7. return
  8. end

where $0 referring to the whole substring matched by the pattern and $1 referring to the first parenthesized capturing substring.

Curly braces can also be used to disambiguate variable names from the background string literals:

  1. local newstr, n, err = ngx.re.sub("hello, 1234", "[0-9]", "${0}00")
  2. -- newstr == "hello, 100234"
  3. -- n == 1

Literal dollar sign characters ($) in the replace string argument can be escaped by another dollar sign, for instance,

  1. local newstr, n, err = ngx.re.sub("hello, 1234", "[0-9]", "$$")
  2. -- newstr == "hello, $234"
  3. -- n == 1

Do not use backlashes to escape dollar signs; it will not work as expected.

When the replace argument is of type “function”, then it will be invoked with the “match table” as the argument to generate the replace string literal for substitution. The “match table” fed into the replace function is exactly the same as the return value of ngx.re.match. Here is an example:

  1. local func = function (m)
  2. return "[" .. m[0] .. "][" .. m[1] .. "]"
  3. end
  4. local newstr, n, err = ngx.re.sub("hello, 1234", "( [0-9] ) [0-9]", func, "x")
  5. -- newstr == "hello, [12][1]34"
  6. -- n == 1

The dollar sign characters in the return value of the replace function argument are not special at all.

This method requires the PCRE library enabled in Nginx. (Known Issue With Special Escaping Sequences).

This feature was first introduced in the v0.2.1rc13 release.

Back to TOC

ngx.re.gsub

syntax: newstr, n, err = ngx.re.gsub(subject, regex, replace, options?)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Just like ngx.re.sub, but does global substitution.

Here is some examples:

  1. local newstr, n, err = ngx.re.gsub("hello, world", "([a-z])[a-z]+", "[$0,$1]", "i")
  2. if newstr then
  3. -- newstr == "[hello,h], [world,w]"
  4. -- n == 2
  5. else
  6. ngx.log(ngx.ERR, "error: ", err)
  7. return
  8. end
  1. local func = function (m)
  2. return "[" .. m[0] .. "," .. m[1] .. "]"
  3. end
  4. local newstr, n, err = ngx.re.gsub("hello, world", "([a-z])[a-z]+", func, "i")
  5. -- newstr == "[hello,h], [world,w]"
  6. -- n == 2

This method requires the PCRE library enabled in Nginx. (Known Issue With Special Escaping Sequences).

This feature was first introduced in the v0.2.1rc15 release.

Back to TOC

ngx.shared.DICT

syntax: dict = ngx.shared.DICT

syntax: dict = ngx.shared[name_var]

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Fetching the shm-based Lua dictionary object for the shared memory zone named DICT defined by the lua_shared_dict directive.

Shared memory zones are always shared by all the nginx worker processes in the current nginx server instance.

The resulting object dict has the following methods:

All these methods are atomic operations, that is, safe from concurrent accesses from multiple nginx worker processes for the same lua_shared_dict zone.

Here is an example:

  1. http {
  2. lua_shared_dict dogs 10m;
  3. server {
  4. location /set {
  5. content_by_lua_block {
  6. local dogs = ngx.shared.dogs
  7. dogs:set("Jim", 8)
  8. ngx.say("STORED")
  9. }
  10. }
  11. location /get {
  12. content_by_lua_block {
  13. local dogs = ngx.shared.dogs
  14. ngx.say(dogs:get("Jim"))
  15. }
  16. }
  17. }
  18. }

Let us test it:

  1. $ curl localhost/set
  2. STORED
  3. $ curl localhost/get
  4. 8
  5. $ curl localhost/get
  6. 8

The number 8 will be consistently output when accessing /get regardless of how many Nginx workers there are because the dogs dictionary resides in the shared memory and visible to all of the worker processes.

The shared dictionary will retain its contents through a server config reload (either by sending the HUP signal to the Nginx process or by using the -s reload command-line option).

The contents in the dictionary storage will be lost, however, when the Nginx server quits.

This feature was first introduced in the v0.3.1rc22 release.

Back to TOC

ngx.shared.DICT.get

syntax: value, flags = ngx.shared.DICT:get(key)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Retrieving the value in the dictionary ngx.shared.DICT for the key key. If the key does not exist or has expired, then nil will be returned.

In case of errors, nil and a string describing the error will be returned.

The value returned will have the original data type when they were inserted into the dictionary, for example, Lua booleans, numbers, or strings.

The first argument to this method must be the dictionary object itself, for example,

  1. local cats = ngx.shared.cats
  2. local value, flags = cats.get(cats, "Marry")

or use Lua’s syntactic sugar for method calls:

  1. local cats = ngx.shared.cats
  2. local value, flags = cats:get("Marry")

These two forms are fundamentally equivalent.

If the user flags is 0 (the default), then no flags value will be returned.

This feature was first introduced in the v0.3.1rc22 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.get_stale

syntax: value, flags, stale = ngx.shared.DICT:get_stale(key)

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to the get method but returns the value even if the key has already expired.

Returns a 3rd value, stale, indicating whether the key has expired or not.

Note that the value of an expired key is not guaranteed to be available so one should never rely on the availability of expired items.

This method was first introduced in the 0.8.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.set

syntax: success, err, forcible = ngx.shared.DICT:set(key, value, exptime?, flags?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Unconditionally sets a key-value pair into the shm-based dictionary ngx.shared.DICT. Returns three values:

  • success: boolean value to indicate whether the key-value pair is stored or not.
  • err: textual error message, can be "no memory".
  • forcible: a boolean value to indicate whether other valid items have been removed forcibly when out of storage in the shared memory zone.

The value argument inserted can be Lua booleans, numbers, strings, or nil. Their value type will also be stored into the dictionary and the same data type can be retrieved later via the get method.

The optional exptime argument specifies expiration time (in seconds) for the inserted key-value pair. The time resolution is 0.001 seconds. If the exptime takes the value 0 (which is the default), then the item will never expire.

The optional flags argument specifies a user flags value associated with the entry to be stored. It can also be retrieved later with the value. The user flags is stored as an unsigned 32-bit integer internally. Defaults to 0. The user flags argument was first introduced in the v0.5.0rc2 release.

When it fails to allocate memory for the current key-value item, then set will try removing existing items in the storage according to the Least-Recently Used (LRU) algorithm. Note that, LRU takes priority over expiration time here. If up to tens of existing items have been removed and the storage left is still insufficient (either due to the total capacity limit specified by lua_shared_dict or memory segmentation), then the err return value will be no memory and success will be false.

If this method succeeds in storing the current item by forcibly removing other not-yet-expired items in the dictionary via LRU, the forcible return value will be true. If it stores the item without forcibly removing other valid items, then the return value forcible will be false.

The first argument to this method must be the dictionary object itself, for example,

  1. local cats = ngx.shared.cats
  2. local succ, err, forcible = cats.set(cats, "Marry", "it is a nice cat!")

or use Lua’s syntactic sugar for method calls:

  1. local cats = ngx.shared.cats
  2. local succ, err, forcible = cats:set("Marry", "it is a nice cat!")

These two forms are fundamentally equivalent.

This feature was first introduced in the v0.3.1rc22 release.

Please note that while internally the key-value pair is set atomically, the atomicity does not go across the method call boundary.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.safe_set

syntax: ok, err = ngx.shared.DICT:safe_set(key, value, exptime?, flags?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to the set method, but never overrides the (least recently used) unexpired items in the store when running out of storage in the shared memory zone. In this case, it will immediately return nil and the string “no memory”.

This feature was first introduced in the v0.7.18 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.add

syntax: success, err, forcible = ngx.shared.DICT:add(key, value, exptime?, flags?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Just like the set method, but only stores the key-value pair into the dictionary ngx.shared.DICT if the key does not exist.

If the key argument already exists in the dictionary (and not expired for sure), the success return value will be false and the err return value will be "exists".

This feature was first introduced in the v0.3.1rc22 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.safe_add

syntax: ok, err = ngx.shared.DICT:safe_add(key, value, exptime?, flags?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to the add method, but never overrides the (least recently used) unexpired items in the store when running out of storage in the shared memory zone. In this case, it will immediately return nil and the string “no memory”.

This feature was first introduced in the v0.7.18 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.replace

syntax: success, err, forcible = ngx.shared.DICT:replace(key, value, exptime?, flags?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Just like the set method, but only stores the key-value pair into the dictionary ngx.shared.DICT if the key does exist.

If the key argument does not exist in the dictionary (or expired already), the success return value will be false and the err return value will be "not found".

This feature was first introduced in the v0.3.1rc22 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.delete

syntax: ngx.shared.DICT:delete(key)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Unconditionally removes the key-value pair from the shm-based dictionary ngx.shared.DICT.

It is equivalent to ngx.shared.DICT:set(key, nil).

This feature was first introduced in the v0.3.1rc22 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.incr

syntax: newval, err, forcible? = ngx.shared.DICT:incr(key, value, init?, init_ttl?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

optional requirement: resty.core.shdict or resty.core

Increments the (numerical) value for key in the shm-based dictionary ngx.shared.DICT by the step value value. Returns the new resulting number if the operation is successfully completed or nil and an error message otherwise.

When the key does not exist or has already expired in the shared dictionary,

  1. if the init argument is not specified or takes the value nil, this method will return nil and the error string "not found", or
  2. if the init argument takes a number value, this method will create a new key with the value init + value.

Like the add method, it also overrides the (least recently used) unexpired items in the store when running out of storage in the shared memory zone.

The optional init_ttl argument specifies expiration time (in seconds) of the value when it is initialized via the init argument. The time resolution is 0.001 seconds. If init_ttl takes the value 0 (which is the default), then the item will never expire. This argument cannot be provided without providing the init argument as well, and has no effect if the value already exists (e.g., if it was previously inserted via set or the likes).

Note: Usage of the init_ttl argument requires the resty.core.shdict or resty.core modules from the lua-resty-core library. Example:

  1. require "resty.core"
  2. local cats = ngx.shared.cats
  3. local newval, err = cats:incr("black_cats", 1, 0, 0.1)
  4. print(newval) -- 1
  5. ngx.sleep(0.2)
  6. local val, err = cats:get("black_cats")
  7. print(val) -- nil

The forcible return value will always be nil when the init argument is not specified.

If this method succeeds in storing the current item by forcibly removing other not-yet-expired items in the dictionary via LRU, the forcible return value will be true. If it stores the item without forcibly removing other valid items, then the return value forcible will be false.

If the original value is not a valid Lua number in the dictionary, it will return nil and "not a number".

The value argument and init argument can be any valid Lua numbers, like negative numbers or floating-point numbers.

This method was first introduced in the v0.3.1rc22 release.

The optional init parameter was first added in the v0.10.6 release.

The optional init_ttl parameter was introduced in the v0.10.12rc2 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.lpush

syntax: length, err = ngx.shared.DICT:lpush(key, value)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Inserts the specified (numerical or string) value at the head of the list named key in the shm-based dictionary ngx.shared.DICT. Returns the number of elements in the list after the push operation.

If key does not exist, it is created as an empty list before performing the push operation. When the key already takes a value that is not a list, it will return nil and "value not a list".

It never overrides the (least recently used) unexpired items in the store when running out of storage in the shared memory zone. In this case, it will immediately return nil and the string “no memory”.

This feature was first introduced in the v0.10.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.rpush

syntax: length, err = ngx.shared.DICT:rpush(key, value)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to the lpush method, but inserts the specified (numerical or string) value at the tail of the list named key.

This feature was first introduced in the v0.10.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.lpop

syntax: val, err = ngx.shared.DICT:lpop(key)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Removes and returns the first element of the list named key in the shm-based dictionary ngx.shared.DICT.

If key does not exist, it will return nil. When the key already takes a value that is not a list, it will return nil and "value not a list".

This feature was first introduced in the v0.10.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.rpop

syntax: val, err = ngx.shared.DICT:rpop(key)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Removes and returns the last element of the list named key in the shm-based dictionary ngx.shared.DICT.

If key does not exist, it will return nil. When the key already takes a value that is not a list, it will return nil and "value not a list".

This feature was first introduced in the v0.10.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.llen

syntax: len, err = ngx.shared.DICT:llen(key)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the number of elements in the list named key in the shm-based dictionary ngx.shared.DICT.

If key does not exist, it is interpreted as an empty list and 0 is returned. When the key already takes a value that is not a list, it will return nil and "value not a list".

This feature was first introduced in the v0.10.6 release.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.ttl

syntax: ttl, err = ngx.shared.DICT:ttl(key)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

requires: resty.core.shdict or resty.core

Retrieves the remaining TTL (time-to-live in seconds) of a key-value pair in the shm-based dictionary ngx.shared.DICT. Returns the TTL as a number if the operation is successfully completed or nil and an error message otherwise.

If the key does not exist (or has already expired), this method will return nil and the error string "not found".

The TTL is originally determined by the exptime argument of the set, add, replace (and the likes) methods. It has a time resolution of 0.001 seconds. A value of 0 means that the item will never expire.

Example:

  1. require "resty.core"
  2. local cats = ngx.shared.cats
  3. local succ, err = cats:set("Marry", "a nice cat", 0.5)
  4. ngx.sleep(0.2)
  5. local ttl, err = cats:ttl("Marry")
  6. ngx.say(ttl) -- 0.3

This feature was first introduced in the v0.10.11 release.

Note: This method requires the resty.core.shdict or resty.core modules from the lua-resty-core library.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.expire

syntax: success, err = ngx.shared.DICT:expire(key, exptime)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

requires: resty.core.shdict or resty.core

Updates the exptime (in second) of a key-value pair in the shm-based dictionary ngx.shared.DICT. Returns a boolean indicating success if the operation completes or nil and an error message otherwise.

If the key does not exist, this method will return nil and the error string "not found".

The exptime argument has a resolution of 0.001 seconds. If exptime is 0, then the item will never expire.

Example:

  1. require "resty.core"
  2. local cats = ngx.shared.cats
  3. local succ, err = cats:set("Marry", "a nice cat", 0.1)
  4. succ, err = cats:expire("Marry", 0.5)
  5. ngx.sleep(0.2)
  6. local val, err = cats:get("Marry")
  7. ngx.say(val) -- "a nice cat"

This feature was first introduced in the v0.10.11 release.

Note: This method requires the resty.core.shdict or resty.core modules from the lua-resty-core library.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.flush_all

syntax: ngx.shared.DICT:flush_all()

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Flushes out all the items in the dictionary. This method does not actuall free up all the memory blocks in the dictionary but just marks all the existing items as expired.

This feature was first introduced in the v0.5.0rc17 release.

See also ngx.shared.DICT.flush_expired and ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.flush_expired

syntax: flushed = ngx.shared.DICT:flush_expired(max_count?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Flushes out the expired items in the dictionary, up to the maximal number specified by the optional max_count argument. When the max_count argument is given 0 or not given at all, then it means unlimited. Returns the number of items that have actually been flushed.

Unlike the flush_all method, this method actually free up the memory used by the expired items.

This feature was first introduced in the v0.6.3 release.

See also ngx.shared.DICT.flush_all and ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.get_keys

syntax: keys = ngx.shared.DICT:get_keys(max_count?)

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Fetch a list of the keys from the dictionary, up to <max_count>.

By default, only the first 1024 keys (if any) are returned. When the <max_count> argument is given the value 0, then all the keys will be returned even there is more than 1024 keys in the dictionary.

CAUTION Avoid calling this method on dictionaries with a very large number of keys as it may lock the dictionary for significant amount of time and block Nginx worker processes trying to access the dictionary.

This feature was first introduced in the v0.7.3 release.

Back to TOC

ngx.shared.DICT.capacity

syntax: capacity_bytes = ngx.shared.DICT:capacity()

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

requires: resty.core.shdict or resty.core

Retrieves the capacity in bytes for the shm-based dictionary ngx.shared.DICT declared with
the lua_shared_dict directive.

Example:

  1. require "resty.core.shdict"
  2. local cats = ngx.shared.cats
  3. local capacity_bytes = cats:capacity()

This feature was first introduced in the v0.10.11 release.

Note: This method requires the resty.core.shdict or resty.core modules from the lua-resty-core library.

This feature requires at least nginx core version 0.7.3.

See also ngx.shared.DICT.

Back to TOC

ngx.shared.DICT.free_space

syntax: free_page_bytes = ngx.shared.DICT:free_space()

context: init_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

requires: resty.core.shdict or resty.core

Retrieves the free page size in bytes for the shm-based dictionary ngx.shared.DICT.

Note: The memory for ngx.shared.DICT is allocated via the nginx slab allocator which has each slot for
data size ranges like ~8, 9~16, 17~32, …, 1025~2048, 2048~ bytes. And pages are assigned to a slot if there
is no room in already assigned pages for the slot.

So even if the return value of the free_space method is zero, there may be room in already assigned pages, so
you may successfully set a new key value pair to the shared dict without getting true for forcible or
non nil err from the ngx.shared.DICT.set.

On the other hand, if already assigned pages for a slot are full and a new key value pair is added to the
slot and there is no free page, you may get true for forcible or non nil err from the
ngx.shared.DICT.set method.

Example:

  1. require "resty.core.shdict"
  2. local cats = ngx.shared.cats
  3. local free_page_bytes = cats:free_space()

This feature was first introduced in the v0.10.11 release.

Note: This method requires the resty.core.shdict or resty.core modules from the lua-resty-core library.

This feature requires at least nginx core version 1.11.7.

See also ngx.shared.DICT.

Back to TOC

ngx.socket.udp

syntax: udpsock = ngx.socket.udp()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Creates and returns a UDP or datagram-oriented unix domain socket object (also known as one type of the “cosocket” objects). The following methods are supported on this object:

It is intended to be compatible with the UDP API of the LuaSocket library but is 100% nonblocking out of the box.

This feature was first introduced in the v0.5.7 release.

See also ngx.socket.tcp.

Back to TOC

udpsock:setpeername

syntax: ok, err = udpsock:setpeername(host, port)

syntax: ok, err = udpsock:setpeername(“unix:/path/to/unix-domain.socket”)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Attempts to connect a UDP socket object to a remote server or to a datagram unix domain socket file. Because the datagram protocol is actually connection-less, this method does not really establish a “connection”, but only just set the name of the remote peer for subsequent read/write operations.

Both IP addresses and domain names can be specified as the host argument. In case of domain names, this method will use Nginx core’s dynamic resolver to parse the domain name without blocking and it is required to configure the resolver directive in the nginx.conf file like this:

  1. resolver 8.8.8.8; # use Google's public DNS nameserver

If the nameserver returns multiple IP addresses for the host name, this method will pick up one randomly.

In case of error, the method returns nil followed by a string describing the error. In case of success, the method returns 1.

Here is an example for connecting to a UDP (memcached) server:

  1. location /test {
  2. resolver 8.8.8.8;
  3. content_by_lua_block {
  4. local sock = ngx.socket.udp()
  5. local ok, err = sock:setpeername("my.memcached.server.domain", 11211)
  6. if not ok then
  7. ngx.say("failed to connect to memcached: ", err)
  8. return
  9. end
  10. ngx.say("successfully connected to memcached!")
  11. sock:close()
  12. }
  13. }

Since the v0.7.18 release, connecting to a datagram unix domain socket file is also possible on Linux:

  1. local sock = ngx.socket.udp()
  2. local ok, err = sock:setpeername("unix:/tmp/some-datagram-service.sock")
  3. if not ok then
  4. ngx.say("failed to connect to the datagram unix domain socket: ", err)
  5. return
  6. end

assuming the datagram service is listening on the unix domain socket file /tmp/some-datagram-service.sock and the client socket will use the “autobind” feature on Linux.

Calling this method on an already connected socket object will cause the original connection to be closed first.

This method was first introduced in the v0.5.7 release.

Back to TOC

udpsock:send

syntax: ok, err = udpsock:send(data)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Sends data on the current UDP or datagram unix domain socket object.

In case of success, it returns 1. Otherwise, it returns nil and a string describing the error.

The input argument data can either be a Lua string or a (nested) Lua table holding string fragments. In case of table arguments, this method will copy all the string elements piece by piece to the underlying Nginx socket send buffers, which is usually optimal than doing string concatenation operations on the Lua land.

This feature was first introduced in the v0.5.7 release.

Back to TOC

udpsock:receive

syntax: data, err = udpsock:receive(size?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Receives data from the UDP or datagram unix domain socket object with an optional receive buffer size argument, size.

This method is a synchronous operation and is 100% nonblocking.

In case of success, it returns the data received; in case of error, it returns nil with a string describing the error.

If the size argument is specified, then this method will use this size as the receive buffer size. But when this size is greater than 8192, then 8192 will be used instead.

If no argument is specified, then the maximal buffer size, 8192 is assumed.

Timeout for the reading operation is controlled by the lua_socket_read_timeout config directive and the settimeout method. And the latter takes priority. For example:

  1. sock:settimeout(1000) -- one second timeout
  2. local data, err = sock:receive()
  3. if not data then
  4. ngx.say("failed to read a packet: ", err)
  5. return
  6. end
  7. ngx.say("successfully read a packet: ", data)

It is important here to call the settimeout method before calling this method.

This feature was first introduced in the v0.5.7 release.

Back to TOC

udpsock:close

syntax: ok, err = udpsock:close()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Closes the current UDP or datagram unix domain socket. It returns the 1 in case of success and returns nil with a string describing the error otherwise.

Socket objects that have not invoked this method (and associated connections) will be closed when the socket object is released by the Lua GC (Garbage Collector) or the current client HTTP request finishes processing.

This feature was first introduced in the v0.5.7 release.

Back to TOC

udpsock:settimeout

syntax: udpsock:settimeout(time)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Set the timeout value in milliseconds for subsequent socket operations (like receive).

Settings done by this method takes priority over those config directives, like lua_socket_read_timeout.

This feature was first introduced in the v0.5.7 release.

Back to TOC

ngx.socket.stream

Just an alias to ngx.socket.tcp. If the stream-typed cosocket may also connect to a unix domain
socket, then this API name is preferred.

This API function was first added to the v0.10.1 release.

Back to TOC

ngx.socket.tcp

syntax: tcpsock = ngx.socket.tcp()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Creates and returns a TCP or stream-oriented unix domain socket object (also known as one type of the “cosocket” objects). The following methods are supported on this object:

It is intended to be compatible with the TCP API of the LuaSocket library but is 100% nonblocking out of the box. Also, we introduce some new APIs to provide more functionalities.

The cosocket object created by this API function has exactly the same lifetime as the Lua handler creating it. So never pass the cosocket object to any other Lua handler (including ngx.timer callback functions) and never share the cosocket object between different NGINX requests.

For every cosocket object’s underlying connection, if you do not
explicitly close it (via close) or put it back to the connection
pool (via setkeepalive), then it is automatically closed when one of
the following two events happens:

  • the current request handler completes, or
  • the Lua cosocket object value gets collected by the Lua GC.

Fatal errors in cosocket operations always automatically close the current
connection (note that, read timeout error is the only error that is
not fatal), and if you call close on a closed connection, you will get
the “closed” error.

Starting from the 0.9.9 release, the cosocket object here is full-duplex, that is, a reader “light thread” and a writer “light thread” can operate on a single cosocket object simultaneously (both “light threads” must belong to the same Lua handler though, see reasons above). But you cannot have two “light threads” both reading (or writing or connecting) the same cosocket, otherwise you might get an error like “socket busy reading” when calling the methods of the cosocket object.

This feature was first introduced in the v0.5.0rc1 release.

See also ngx.socket.udp.

Back to TOC

tcpsock:connect

syntax: ok, err = tcpsock:connect(host, port, options_table?)

syntax: ok, err = tcpsock:connect(“unix:/path/to/unix-domain.socket”, options_table?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Attempts to connect a TCP socket object to a remote server or to a stream unix domain socket file without blocking.

Before actually resolving the host name and connecting to the remote backend, this method will always look up the connection pool for matched idle connections created by previous calls of this method (or the ngx.socket.connect function).

Both IP addresses and domain names can be specified as the host argument. In case of domain names, this method will use Nginx core’s dynamic resolver to parse the domain name without blocking and it is required to configure the resolver directive in the nginx.conf file like this:

  1. resolver 8.8.8.8; # use Google's public DNS nameserver

If the nameserver returns multiple IP addresses for the host name, this method will pick up one randomly.

In case of error, the method returns nil followed by a string describing the error. In case of success, the method returns 1.

Here is an example for connecting to a TCP server:

  1. location /test {
  2. resolver 8.8.8.8;
  3. content_by_lua_block {
  4. local sock = ngx.socket.tcp()
  5. local ok, err = sock:connect("www.google.com", 80)
  6. if not ok then
  7. ngx.say("failed to connect to google: ", err)
  8. return
  9. end
  10. ngx.say("successfully connected to google!")
  11. sock:close()
  12. }
  13. }

Connecting to a Unix Domain Socket file is also possible:

  1. local sock = ngx.socket.tcp()
  2. local ok, err = sock:connect("unix:/tmp/memcached.sock")
  3. if not ok then
  4. ngx.say("failed to connect to the memcached unix domain socket: ", err)
  5. return
  6. end

assuming memcached (or something else) is listening on the unix domain socket file /tmp/memcached.sock.

Timeout for the connecting operation is controlled by the lua_socket_connect_timeout config directive and the settimeout method. And the latter takes priority. For example:

  1. local sock = ngx.socket.tcp()
  2. sock:settimeout(1000) -- one second timeout
  3. local ok, err = sock:connect(host, port)

It is important here to call the settimeout method before calling this method.

Calling this method on an already connected socket object will cause the original connection to be closed first.

An optional Lua table can be specified as the last argument to this method to specify various connect options:

  • pool
    specify a custom name for the connection pool being used. If omitted, then the connection pool name will be generated from the string template "<host>:<port>" or "<unix-socket-path>".

The support for the options table argument was first introduced in the v0.5.7 release.

This method was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:sslhandshake

syntax: session, err = tcpsock:sslhandshake(reused_session?, server_name?, ssl_verify?, send_status_req?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Does SSL/TLS handshake on the currently established connection.

The optional reused_session argument can take a former SSL
session userdata returned by a previous sslhandshake
call for exactly the same target. For short-lived connections, reusing SSL
sessions can usually speed up the handshake by one order by magnitude but it
is not so useful if the connection pool is enabled. This argument defaults to
nil. If this argument takes the boolean false value, no SSL session
userdata would return by this call and only a Lua boolean will be returned as
the first return value; otherwise the current SSL session will
always be returned as the first argument in case of successes.

The optional server_name argument is used to specify the server
name for the new TLS extension Server Name Indication (SNI). Use of SNI can
make different servers share the same IP address on the server side. Also,
when SSL verification is enabled, this server_name argument is
also used to validate the server name specified in the server certificate sent from
the remote.

The optional ssl_verify argument takes a Lua boolean value to
control whether to perform SSL verification. When set to true, the server
certificate will be verified according to the CA certificates specified by
the lua_ssl_trusted_certificate directive.
You may also need to adjust the lua_ssl_verify_depth
directive to control how deep we should follow along the certificate chain.
Also, when the ssl_verify argument is true and the
server_name argument is also specified, the latter will be used
to validate the server name in the server certificate.

The optional send_status_req argument takes a boolean that controls whether to send
the OCSP status request in the SSL handshake request (which is for requesting OCSP stapling).

For connections that have already done SSL/TLS handshake, this method returns
immediately.

This method was first introduced in the v0.9.11 release.

Back to TOC

tcpsock:send

syntax: bytes, err = tcpsock:send(data)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Sends data without blocking on the current TCP or Unix Domain Socket connection.

This method is a synchronous operation that will not return until all the data has been flushed into the system socket send buffer or an error occurs.

In case of success, it returns the total number of bytes that have been sent. Otherwise, it returns nil and a string describing the error.

The input argument data can either be a Lua string or a (nested) Lua table holding string fragments. In case of table arguments, this method will copy all the string elements piece by piece to the underlying Nginx socket send buffers, which is usually optimal than doing string concatenation operations on the Lua land.

Timeout for the sending operation is controlled by the lua_socket_send_timeout config directive and the settimeout method. And the latter takes priority. For example:

  1. sock:settimeout(1000) -- one second timeout
  2. local bytes, err = sock:send(request)

It is important here to call the settimeout method before calling this method.

In case of any connection errors, this method always automatically closes the current connection.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:receive

syntax: data, err, partial = tcpsock:receive(size)

syntax: data, err, partial = tcpsock:receive(pattern?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Receives data from the connected socket according to the reading pattern or size.

This method is a synchronous operation just like the send method and is 100% nonblocking.

In case of success, it returns the data received; in case of error, it returns nil with a string describing the error and the partial data received so far.

If a number-like argument is specified (including strings that look like numbers), then it is interpreted as a size. This method will not return until it reads exactly this size of data or an error occurs.

If a non-number-like string argument is specified, then it is interpreted as a “pattern”. The following patterns are supported:

  • '*a': reads from the socket until the connection is closed. No end-of-line translation is performed;
  • '*l': reads a line of text from the socket. The line is terminated by a Line Feed (LF) character (ASCII 10), optionally preceded by a Carriage Return (CR) character (ASCII 13). The CR and LF characters are not included in the returned line. In fact, all CR characters are ignored by the pattern.

If no argument is specified, then it is assumed to be the pattern '*l', that is, the line reading pattern.

Timeout for the reading operation is controlled by the lua_socket_read_timeout config directive and the settimeout method. And the latter takes priority. For example:

  1. sock:settimeout(1000) -- one second timeout
  2. local line, err, partial = sock:receive()
  3. if not line then
  4. ngx.say("failed to read a line: ", err)
  5. return
  6. end
  7. ngx.say("successfully read a line: ", line)

It is important here to call the settimeout method before calling this method.

Since the v0.8.8 release, this method no longer automatically closes the current connection when the read timeout error happens. For other connection errors, this method always automatically closes the connection.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:receiveany

syntax: data, err = tcpsock:receiveany(max)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Returns any data received by the connected socket, at most max bytes.

This method is a synchronous operation just like the send method and is 100% nonblocking.

In case of success, it returns the data received; in case of error, it returns nil with a string describing the error.

If the received data is more than this size, this method will return with exactly this size of data.
The remaining data in the underlying receive buffer could be returned in the next reading operation.

Timeout for the reading operation is controlled by the lua_socket_read_timeout config directive and the settimeouts method. And the latter takes priority. For example:

  1. sock:settimeouts(1000, 1000, 1000) -- one second timeout for connect/read/write
  2. local data, err = sock:receiveany(10 * 1024 * 1024) -- read any data, at most 10K
  3. if not data then
  4. ngx.say("failed to read any data: ", err)
  5. return
  6. end
  7. ngx.say("successfully read: ", data)

This method doesn’t automatically close the current connection when the read timeout error occurs. For other connection errors, this method always automatically closes the connection.

This feature was first introduced in the v0.10.14 release.

Back to TOC

tcpsock:receiveuntil

syntax: iterator = tcpsock:receiveuntil(pattern, options?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

This method returns an iterator Lua function that can be called to read the data stream until it sees the specified pattern or an error occurs.

Here is an example for using this method to read a data stream with the boundary sequence --abcedhb:

  1. local reader = sock:receiveuntil("\r\n--abcedhb")
  2. local data, err, partial = reader()
  3. if not data then
  4. ngx.say("failed to read the data stream: ", err)
  5. end
  6. ngx.say("read the data stream: ", data)

When called without any argument, the iterator function returns the received data right before the specified pattern string in the incoming data stream. So for the example above, if the incoming data stream is 'hello, world! -agentzh\r\n--abcedhb blah blah', then the string 'hello, world! -agentzh' will be returned.

In case of error, the iterator function will return nil along with a string describing the error and the partial data bytes that have been read so far.

The iterator function can be called multiple times and can be mixed safely with other cosocket method calls or other iterator function calls.

The iterator function behaves differently (i.e., like a real iterator) when it is called with a size argument. That is, it will read that size of data on each invocation and will return nil at the last invocation (either sees the boundary pattern or meets an error). For the last successful invocation of the iterator function, the err return value will be nil too. The iterator function will be reset after the last successful invocation that returns nil data and nil error. Consider the following example:

  1. local reader = sock:receiveuntil("\r\n--abcedhb")
  2. while true do
  3. local data, err, partial = reader(4)
  4. if not data then
  5. if err then
  6. ngx.say("failed to read the data stream: ", err)
  7. break
  8. end
  9. ngx.say("read done")
  10. break
  11. end
  12. ngx.say("read chunk: [", data, "]")
  13. end

Then for the incoming data stream 'hello, world! -agentzh\r\n--abcedhb blah blah', we shall get the following output from the sample code above:

  1. read chunk: [hell]
  2. read chunk: [o, w]
  3. read chunk: [orld]
  4. read chunk: [! -a]
  5. read chunk: [gent]
  6. read chunk: [zh]
  7. read done

Note that, the actual data returned might be a little longer than the size limit specified by the size argument when the boundary pattern has ambiguity for streaming parsing. Near the boundary of the data stream, the data string actually returned could also be shorter than the size limit.

Timeout for the iterator function’s reading operation is controlled by the lua_socket_read_timeout config directive and the settimeout method. And the latter takes priority. For example:

  1. local readline = sock:receiveuntil("\r\n")
  2. sock:settimeout(1000) -- one second timeout
  3. line, err, partial = readline()
  4. if not line then
  5. ngx.say("failed to read a line: ", err)
  6. return
  7. end
  8. ngx.say("successfully read a line: ", line)

It is important here to call the settimeout method before calling the iterator function (note that the receiveuntil call is irrelevant here).

As from the v0.5.1 release, this method also takes an optional options table argument to control the behavior. The following options are supported:

  • inclusive

The inclusive takes a boolean value to control whether to include the pattern string in the returned data string. Default to false. For example,

  1. local reader = tcpsock:receiveuntil("_END_", { inclusive = true })
  2. local data = reader()
  3. ngx.say(data)

Then for the input data stream "hello world _END_ blah blah blah", then the example above will output hello world _END_, including the pattern string _END_ itself.

Since the v0.8.8 release, this method no longer automatically closes the current connection when the read timeout error happens. For other connection errors, this method always automatically closes the connection.

This method was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:close

syntax: ok, err = tcpsock:close()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Closes the current TCP or stream unix domain socket. It returns the 1 in case of success and returns nil with a string describing the error otherwise.

Note that there is no need to call this method on socket objects that have invoked the setkeepalive method because the socket object is already closed (and the current connection is saved into the built-in connection pool).

Socket objects that have not invoked this method (and associated connections) will be closed when the socket object is released by the Lua GC (Garbage Collector) or the current client HTTP request finishes processing.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:settimeout

syntax: tcpsock:settimeout(time)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Set the timeout value in milliseconds for subsequent socket operations (connect, receive, and iterators returned from receiveuntil).

Settings done by this method takes priority over those config directives, i.e., lua_socket_connect_timeout, lua_socket_send_timeout, and lua_socket_read_timeout.

Note that this method does not affect the lua_socket_keepalive_timeout setting; the timeout argument to the setkeepalive method should be used for this purpose instead.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:settimeouts

syntax: tcpsock:settimeouts(connect_timeout, send_timeout, read_timeout)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Sets the connect timeout thresold, send timeout threshold, and read timeout threshold, respetively, in milliseconds, for subsequent socket
operations (connect, send, receive, and iterators returned from receiveuntil).

Settings done by this method takes priority over those config directives, i.e., lua_socket_connect_timeout, lua_socket_send_timeout, and lua_socket_read_timeout.

You are recommended to use settimeouts instead of settimeout.

Note that this method does not affect the lua_socket_keepalive_timeout setting; the timeout argument to the setkeepalive method should be used for this purpose instead.

This feature was first introduced in the v0.10.7 release.

Back to TOC

tcpsock:setoption

syntax: tcpsock:setoption(option, value?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

This function is added for LuaSocket API compatibility and does nothing for now. Its functionality will be implemented in future.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:setkeepalive

syntax: ok, err = tcpsock:setkeepalive(timeout?, size?)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Puts the current socket’s connection immediately into the cosocket built-in connection pool and keep it alive until other connect method calls request it or the associated maximal idle timeout is expired.

The first optional argument, timeout, can be used to specify the maximal idle timeout (in milliseconds) for the current connection. If omitted, the default setting in the lua_socket_keepalive_timeout config directive will be used. If the 0 value is given, then the timeout interval is unlimited.

The second optional argument, size, can be used to specify the maximal number of connections allowed in the connection pool for the current server (i.e., the current host-port pair or the unix domain socket file path). Note that the size of the connection pool cannot be changed once the pool is created. When this argument is omitted, the default setting in the lua_socket_pool_size config directive will be used.

When the connection pool exceeds the available size limit, the least recently used (idle) connection already in the pool will be closed to make room for the current connection.

Note that the cosocket connection pool is per Nginx worker process rather than per Nginx server instance, so the size limit specified here also applies to every single Nginx worker process.

Idle connections in the pool will be monitored for any exceptional events like connection abortion or unexpected incoming data on the line, in which cases the connection in question will be closed and removed from the pool.

In case of success, this method returns 1; otherwise, it returns nil and a string describing the error.

When the system receive buffer for the current connection has unread data, then this method will return the “connection in dubious state” error message (as the second return value) because the previous session has unread data left behind for the next session and the connection is not safe to be reused.

This method also makes the current cosocket object enter the “closed” state, so there is no need to manually call the close method on it afterwards.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

tcpsock:getreusedtimes

syntax: count, err = tcpsock:getreusedtimes()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

This method returns the (successfully) reused times for the current connection. In case of error, it returns nil and a string describing the error.

If the current connection does not come from the built-in connection pool, then this method always returns 0, that is, the connection has never been reused (yet). If the connection comes from the connection pool, then the return value is always non-zero. So this method can also be used to determine if the current connection comes from the pool.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

ngx.socket.connect

syntax: tcpsock, err = ngx.socket.connect(host, port)

syntax: tcpsock, err = ngx.socket.connect(“unix:/path/to/unix-domain.socket”)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*

This function is a shortcut for combining ngx.socket.tcp() and the connect() method call in a single operation. It is actually implemented like this:

  1. local sock = ngx.socket.tcp()
  2. local ok, err = sock:connect(...)
  3. if not ok then
  4. return nil, err
  5. end
  6. return sock

There is no way to use the settimeout method to specify connecting timeout for this method and the lua_socket_connect_timeout directive must be set at configure time instead.

This feature was first introduced in the v0.5.0rc1 release.

Back to TOC

ngx.get_phase

syntax: str = ngx.get_phase()

context: init_by_lua*, init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Retrieves the current running phase name. Possible return values are

This API was first introduced in the v0.5.10 release.

Back to TOC

ngx.thread.spawn

syntax: co = ngx.thread.spawn(func, arg1, arg2, …)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Spawns a new user “light thread” with the Lua function func as well as those optional arguments arg1, arg2, and etc. Returns a Lua thread (or Lua coroutine) object represents this “light thread”.

“Light threads” are just a special kind of Lua coroutines that are scheduled by the ngx_lua module.

Before ngx.thread.spawn returns, the func will be called with those optional arguments until it returns, aborts with an error, or gets yielded due to I/O operations via the Nginx API for Lua (like tcpsock:receive).

After ngx.thread.spawn returns, the newly-created “light thread” will keep running asynchronously usually at various I/O events.

All the Lua code chunks running by rewrite_by_lua, access_by_lua, and content_by_lua are in a boilerplate “light thread” created automatically by ngx_lua. Such boilerplate “light thread” are also called “entry threads”.

By default, the corresponding Nginx handler (e.g., rewrite_by_lua handler) will not terminate until

  1. both the “entry thread” and all the user “light threads” terminates,
  2. a “light thread” (either the “entry thread” or a user “light thread” aborts by calling ngx.exit, ngx.exec, ngx.redirect, or ngx.req.set_uri(uri, true), or
  3. the “entry thread” terminates with a Lua error.

When the user “light thread” terminates with a Lua error, however, it will not abort other running “light threads” like the “entry thread” does.

Due to the limitation in the Nginx subrequest model, it is not allowed to abort a running Nginx subrequest in general. So it is also prohibited to abort a running “light thread” that is pending on one ore more Nginx subrequests. You must call ngx.thread.wait to wait for those “light thread” to terminate before quitting the “world”. A notable exception here is that you can abort pending subrequests by calling ngx.exit with and only with the status code ngx.ERROR (-1), 408, 444, or 499.

The “light threads” are not scheduled in a pre-emptive way. In other words, no time-slicing is performed automatically. A “light thread” will keep running exclusively on the CPU until

  1. a (nonblocking) I/O operation cannot be completed in a single run,
  2. it calls coroutine.yield to actively give up execution, or
  3. it is aborted by a Lua error or an invocation of ngx.exit, ngx.exec, ngx.redirect, or ngx.req.set_uri(uri, true).

For the first two cases, the “light thread” will usually be resumed later by the ngx_lua scheduler unless a “stop-the-world” event happens.

User “light threads” can create “light threads” themselves. And normal user coroutines created by coroutine.create can also create “light threads”. The coroutine (be it a normal Lua coroutine or a “light thread”) that directly spawns the “light thread” is called the “parent coroutine” for the “light thread” newly spawned.

The “parent coroutine” can call ngx.thread.wait to wait on the termination of its child “light thread”.

You can call coroutine.status() and coroutine.yield() on the “light thread” coroutines.

The status of the “light thread” coroutine can be “zombie” if

  1. the current “light thread” already terminates (either successfully or with an error),
  2. its parent coroutine is still alive, and
  3. its parent coroutine is not waiting on it with ngx.thread.wait.

The following example demonstrates the use of coroutine.yield() in the “light thread” coroutines
to do manual time-slicing:

  1. local yield = coroutine.yield
  2. function f()
  3. local self = coroutine.running()
  4. ngx.say("f 1")
  5. yield(self)
  6. ngx.say("f 2")
  7. yield(self)
  8. ngx.say("f 3")
  9. end
  10. local self = coroutine.running()
  11. ngx.say("0")
  12. yield(self)
  13. ngx.say("1")
  14. ngx.thread.spawn(f)
  15. ngx.say("2")
  16. yield(self)
  17. ngx.say("3")
  18. yield(self)
  19. ngx.say("4")

Then it will generate the output

  1. 0
  2. 1
  3. f 1
  4. 2
  5. f 2
  6. 3
  7. f 3
  8. 4

“Light threads” are mostly useful for making concurrent upstream requests in a single Nginx request handler, much like a generalized version of ngx.location.capture_multi that can work with all the Nginx API for Lua. The following example demonstrates parallel requests to MySQL, Memcached, and upstream HTTP services in a single Lua handler, and outputting the results in the order that they actually return (similar to Facebook’s BigPipe model):

  1. -- query mysql, memcached, and a remote http service at the same time,
  2. -- output the results in the order that they
  3. -- actually return the results.
  4. local mysql = require "resty.mysql"
  5. local memcached = require "resty.memcached"
  6. local function query_mysql()
  7. local db = mysql:new()
  8. db:connect{
  9. host = "127.0.0.1",
  10. port = 3306,
  11. database = "test",
  12. user = "monty",
  13. password = "mypass"
  14. }
  15. local res, err, errno, sqlstate =
  16. db:query("select * from cats order by id asc")
  17. db:set_keepalive(0, 100)
  18. ngx.say("mysql done: ", cjson.encode(res))
  19. end
  20. local function query_memcached()
  21. local memc = memcached:new()
  22. memc:connect("127.0.0.1", 11211)
  23. local res, err = memc:get("some_key")
  24. ngx.say("memcached done: ", res)
  25. end
  26. local function query_http()
  27. local res = ngx.location.capture("/my-http-proxy")
  28. ngx.say("http done: ", res.body)
  29. end
  30. ngx.thread.spawn(query_mysql) -- create thread 1
  31. ngx.thread.spawn(query_memcached) -- create thread 2
  32. ngx.thread.spawn(query_http) -- create thread 3

This API was first enabled in the v0.7.0 release.

Back to TOC

ngx.thread.wait

syntax: ok, res1, res2, … = ngx.thread.wait(thread1, thread2, …)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*

Waits on one or more child “light threads” and returns the results of the first “light thread” that terminates (either successfully or with an error).

The arguments thread1, thread2, and etc are the Lua thread objects returned by earlier calls of ngx.thread.spawn.

The return values have exactly the same meaning as coroutine.resume, that is, the first value returned is a boolean value indicating whether the “light thread” terminates successfully or not, and subsequent values returned are the return values of the user Lua function that was used to spawn the “light thread” (in case of success) or the error object (in case of failure).

Only the direct “parent coroutine” can wait on its child “light thread”, otherwise a Lua exception will be raised.

The following example demonstrates the use of ngx.thread.wait and ngx.location.capture to emulate ngx.location.capture_multi:

  1. local capture = ngx.location.capture
  2. local spawn = ngx.thread.spawn
  3. local wait = ngx.thread.wait
  4. local say = ngx.say
  5. local function fetch(uri)
  6. return capture(uri)
  7. end
  8. local threads = {
  9. spawn(fetch, "/foo"),
  10. spawn(fetch, "/bar"),
  11. spawn(fetch, "/baz")
  12. }
  13. for i = 1, #threads do
  14. local ok, res = wait(threads[i])
  15. if not ok then
  16. say(i, ": failed to run: ", res)
  17. else
  18. say(i, ": status: ", res.status)
  19. say(i, ": body: ", res.body)
  20. end
  21. end

Here it essentially implements the “wait all” model.

And below is an example demonstrating the “wait any” model:

  1. function f()
  2. ngx.sleep(0.2)
  3. ngx.say("f: hello")
  4. return "f done"
  5. end
  6. function g()
  7. ngx.sleep(0.1)
  8. ngx.say("g: hello")
  9. return "g done"
  10. end
  11. local tf, err = ngx.thread.spawn(f)
  12. if not tf then
  13. ngx.say("failed to spawn thread f: ", err)
  14. return
  15. end
  16. ngx.say("f thread created: ", coroutine.status(tf))
  17. local tg, err = ngx.thread.spawn(g)
  18. if not tg then
  19. ngx.say("failed to spawn thread g: ", err)
  20. return
  21. end
  22. ngx.say("g thread created: ", coroutine.status(tg))
  23. ok, res = ngx.thread.wait(tf, tg)
  24. if not ok then
  25. ngx.say("failed to wait: ", res)
  26. return
  27. end
  28. ngx.say("res: ", res)
  29. -- stop the "world", aborting other running threads
  30. ngx.exit(ngx.OK)

And it will generate the following output:

  1. f thread created: running
  2. g thread created: running
  3. g: hello
  4. res: g done

This API was first enabled in the v0.7.0 release.

Back to TOC

ngx.thread.kill

syntax: ok, err = ngx.thread.kill(thread)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, ngx.timer.*

Kills a running “light thread” created by ngx.thread.spawn. Returns a true value when successful or nil and a string describing the error otherwise.

According to the current implementation, only the parent coroutine (or “light thread”) can kill a thread. Also, a running “light thread” with pending NGINX subrequests (initiated by ngx.location.capture for example) cannot be killed due to a limitation in the NGINX core.

This API was first enabled in the v0.9.9 release.

Back to TOC

ngx.on_abort

syntax: ok, err = ngx.on_abort(callback)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*

Registers a user Lua function as the callback which gets called automatically when the client closes the (downstream) connection prematurely.

Returns 1 if the callback is registered successfully or returns nil and a string describing the error otherwise.

All the Nginx API for Lua can be used in the callback function because the function is run in a special “light thread”, just as those “light threads” created by ngx.thread.spawn.

The callback function can decide what to do with the client abortion event all by itself. For example, it can simply ignore the event by doing nothing and the current Lua request handler will continue executing without interruptions. And the callback function can also decide to terminate everything by calling ngx.exit, for example,

  1. local function my_cleanup()
  2. -- custom cleanup work goes here, like cancelling a pending DB transaction
  3. -- now abort all the "light threads" running in the current request handler
  4. ngx.exit(499)
  5. end
  6. local ok, err = ngx.on_abort(my_cleanup)
  7. if not ok then
  8. ngx.log(ngx.ERR, "failed to register the on_abort callback: ", err)
  9. ngx.exit(500)
  10. end

When lua_check_client_abort is set to off (which is the default), then this function call will always return the error message “lua_check_client_abort is off”.

According to the current implementation, this function can only be called once in a single request handler; subsequent calls will return the error message “duplicate call”.

This API was first introduced in the v0.7.4 release.

See also lua_check_client_abort.

Back to TOC

ngx.timer.at

syntax: hdl, err = ngx.timer.at(delay, callback, user_arg1, user_arg2, …)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Creates an Nginx timer with a user callback function as well as optional user arguments.

The first argument, delay, specifies the delay for the timer,
in seconds. One can specify fractional seconds like 0.001 to mean 1
millisecond here. 0 delay can also be specified, in which case the
timer will immediately expire when the current handler yields
execution.

The second argument, callback, can
be any Lua function, which will be invoked later in a background
“light thread” after the delay specified. The user callback will be
called automatically by the Nginx core with the arguments premature,
user_arg1, user_arg2, and etc, where the premature
argument takes a boolean value indicating whether it is a premature timer
expiration or not, and user_arg1, user_arg2, and etc, are
those (extra) user arguments specified when calling ngx.timer.at
as the remaining arguments.

Premature timer expiration happens when the Nginx worker process is
trying to shut down, as in an Nginx configuration reload triggered by
the HUP signal or in an Nginx server shutdown. When the Nginx worker
is trying to shut down, one can no longer call ngx.timer.at to
create new timers with nonzero delays and in that case ngx.timer.at will return a “conditional false” value and
a string describing the error, that is, “process exiting”.

Starting from the v0.9.3 release, it is allowed to create zero-delay timers even when the Nginx worker process starts shutting down.

When a timer expires, the user Lua code in the timer callback is
running in a “light thread” detached completely from the original
request creating the timer. So objects with the same lifetime as the
request creating them, like cosockets, cannot be shared between the
original request and the timer user callback function.

Here is a simple example:

  1. location / {
  2. ...
  3. log_by_lua_block {
  4. local function push_data(premature, uri, args, status)
  5. -- push the data uri, args, and status to the remote
  6. -- via ngx.socket.tcp or ngx.socket.udp
  7. -- (one may want to buffer the data in Lua a bit to
  8. -- save I/O operations)
  9. end
  10. local ok, err = ngx.timer.at(0, push_data,
  11. ngx.var.uri, ngx.var.args, ngx.header.status)
  12. if not ok then
  13. ngx.log(ngx.ERR, "failed to create timer: ", err)
  14. return
  15. end
  16. }
  17. }

One can also create infinite re-occurring timers, for instance, a timer getting triggered every 5 seconds, by calling ngx.timer.at recursively in the timer callback function. Here is such an example,

  1. local delay = 5
  2. local handler
  3. handler = function (premature)
  4. -- do some routine job in Lua just like a cron job
  5. if premature then
  6. return
  7. end
  8. local ok, err = ngx.timer.at(delay, handler)
  9. if not ok then
  10. ngx.log(ngx.ERR, "failed to create the timer: ", err)
  11. return
  12. end
  13. end
  14. local ok, err = ngx.timer.at(delay, handler)
  15. if not ok then
  16. ngx.log(ngx.ERR, "failed to create the timer: ", err)
  17. return
  18. end

It is recommended, however, to use the ngx.timer.every API function
instead for creating recurring timers since it is more robust.

Because timer callbacks run in the background and their running time
will not add to any client request’s response time, they can easily
accumulate in the server and exhaust system resources due to either
Lua programming mistakes or just too much client traffic. To prevent
extreme consequences like crashing the Nginx server, there are
built-in limitations on both the number of “pending timers” and the
number of “running timers” in an Nginx worker process. The “pending
timers” here mean timers that have not yet been expired and “running
timers” are those whose user callbacks are currently running.

The maximal number of pending timers allowed in an Nginx
worker is controlled by the lua_max_pending_timers
directive. The maximal number of running timers is controlled by the
lua_max_running_timers directive.

According to the current implementation, each “running timer” will
take one (fake) connection record from the global connection record
list configured by the standard worker_connections directive in
nginx.conf. So ensure that the
worker_connections directive is set to
a large enough value that takes into account both the real connections
and fake connections required by timer callbacks (as limited by the
lua_max_running_timers directive).

A lot of the Lua APIs for Nginx are enabled in the context of the timer
callbacks, like stream/datagram cosockets (ngx.socket.tcp and ngx.socket.udp), shared
memory dictionaries (ngx.shared.DICT), user coroutines (coroutine.*),
user “light threads” (ngx.thread.*), ngx.exit, ngx.now/ngx.time,
ngx.md5/ngx.sha1_bin, are all allowed. But the subrequest API (like
ngx.location.capture), the ngx.req.* API, the downstream output API
(like ngx.say, ngx.print, and ngx.flush) are explicitly disabled in
this context.

You can pass most of the standard Lua values (nils, booleans, numbers, strings, tables, closures, file handles, and etc) into the timer callback, either explicitly as user arguments or implicitly as upvalues for the callback closure. There are several exceptions, however: you cannot pass any thread objects returned by coroutine.create and ngx.thread.spawn or any cosocket objects returned by ngx.socket.tcp, ngx.socket.udp, and ngx.req.socket because these objects’ lifetime is bound to the request context creating them while the timer callback is detached from the creating request’s context (by design) and runs in its own (fake) request context. If you try to share the thread or cosocket objects across the boundary of the creating request, then you will get the “no co ctx found” error (for threads) or “bad request” (for cosockets). It is fine, however, to create all these objects inside your timer callback.

This API was first introduced in the v0.8.0 release.

Back to TOC

ngx.timer.every

syntax: hdl, err = ngx.timer.every(delay, callback, user_arg1, user_arg2, …)

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to the ngx.timer.at API function, but

  1. delay cannot be zero,
  2. timer will be created every delay seconds until the current Nginx worker process starts exiting.

When success, returns a “conditional true” value (but not a true). Otherwise, returns a “conditional false” value and a string describing the error.

This API also respect the lua_max_pending_timers and lua_max_running_timers.

This API was first introduced in the v0.10.9 release.

Back to TOC

ngx.timer.running_count

syntax: count = ngx.timer.running_count()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the number of timers currently running.

This directive was first introduced in the v0.9.20 release.

Back to TOC

ngx.timer.pending_count

syntax: count = ngx.timer.pending_count()

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Returns the number of pending timers.

This directive was first introduced in the v0.9.20 release.

Back to TOC

ngx.config.subsystem

syntax: subsystem = ngx.config.subsystem

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

This string field indicates the current NGINX subsystem the current Lua environment is based on. For this module, this field always takes the string value "http". For
ngx_stream_lua_module, however, this field takes the value "stream".

This field was first introduced in the 0.10.1.

Back to TOC

ngx.config.debug

syntax: debug = ngx.config.debug

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

This boolean field indicates whether the current Nginx is a debug build, i.e., being built by the ./configure option --with-debug.

This field was first introduced in the 0.8.7.

Back to TOC

ngx.config.prefix

syntax: prefix = ngx.config.prefix()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

Returns the Nginx server “prefix” path, as determined by the -p command-line option when running the nginx executable, or the path specified by the --prefix command-line option when building Nginx with the ./configure script.

This function was first introduced in the 0.9.2.

Back to TOC

ngx.config.nginx_version

syntax: ver = ngx.config.nginx_version

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

This field take an integral value indicating the version number of the current Nginx core being used. For example, the version number 1.4.3 results in the Lua number 1004003.

This API was first introduced in the 0.9.3 release.

Back to TOC

ngx.config.nginx_configure

syntax: str = ngx.config.nginx_configure()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*

This function returns a string for the NGINX ./configure command’s arguments string.

This API was first introduced in the 0.9.5 release.

Back to TOC

ngx.config.ngx_lua_version

syntax: ver = ngx.config.ngx_lua_version

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*

This field take an integral value indicating the version number of the current ngx_lua module being used. For example, the version number 0.9.3 results in the Lua number 9003.

This API was first introduced in the 0.9.3 release.

Back to TOC

ngx.worker.exiting

syntax: exiting = ngx.worker.exiting()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

This function returns a boolean value indicating whether the current Nginx worker process already starts exiting. Nginx worker process exiting happens on Nginx server quit or configuration reload (aka HUP reload).

This API was first introduced in the 0.9.3 release.

Back to TOC

ngx.worker.pid

syntax: pid = ngx.worker.pid()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

This function returns a Lua number for the process ID (PID) of the current Nginx worker process. This API is more efficient than ngx.var.pid and can be used in contexts where the ngx.var.VARIABLE API cannot be used (like init_worker_by_lua).

This API was first introduced in the 0.9.5 release.

Back to TOC

ngx.worker.count

syntax: count = ngx.worker.count()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_by_lua*, init_worker_by_lua*

Returns the total number of the Nginx worker processes (i.e., the value configured
by the worker_processes
directive in nginx.conf).

This API was first introduced in the 0.9.20 release.

Back to TOC

ngx.worker.id

syntax: count = ngx.worker.id()

context: set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, init_worker_by_lua*

Returns the ordinal number of the current Nginx worker processes (starting from number 0).

So if the total number of workers is N, then this method may return a number between 0
and N - 1 (inclusive).

This function returns meaningful values only for NGINX 1.9.1+. With earlier versions of NGINX, it
always returns nil.

See also ngx.worker.count.

This API was first introduced in the 0.9.20 release.

Back to TOC

ngx.semaphore

syntax: local semaphore = require “ngx.semaphore”

This is a Lua module that implements a classic-style semaphore API for efficient synchronizations among
different “light threads”. Sharing the same semaphore among different “light threads” created in different (request)
contexts are also supported as long as the “light threads” reside in the same NGINX worker process
and the lua_code_cache directive is turned on (which is the default).

This Lua module does not ship with this ngx_lua module itself rather it is shipped with
the
lua-resty-core library.

Please refer to the documentation
for this ngx.semaphore Lua module in lua-resty-core
for more details.

This feature requires at least ngx_lua v0.10.0.

Back to TOC

ngx.balancer

syntax: local balancer = require “ngx.balancer”

This is a Lua module that provides a Lua API to allow defining completely dynamic load balancers
in pure Lua.

This Lua module does not ship with this ngx_lua module itself rather it is shipped with
the
lua-resty-core library.

Please refer to the documentation
for this ngx.balancer Lua module in lua-resty-core
for more details.

This feature requires at least ngx_lua v0.10.0.

Back to TOC

ngx.ssl

syntax: local ssl = require “ngx.ssl”

This Lua module provides API functions to control the SSL handshake process in contexts like
ssl_certificate_by_lua*.

This Lua module does not ship with this ngx_lua module itself rather it is shipped with
the
lua-resty-core library.

Please refer to the documentation
for this ngx.ssl Lua module for more details.

This feature requires at least ngx_lua v0.10.0.

Back to TOC

ngx.ocsp

syntax: local ocsp = require “ngx.ocsp”

This Lua module provides API to perform OCSP queries, OCSP response validations, and
OCSP stapling planting.

Usually, this module is used together with the ngx.ssl
module in the
context of ssl_certificate_by_lua*.

This Lua module does not ship with this ngx_lua module itself rather it is shipped with
the
lua-resty-core library.

Please refer to the documentation
for this ngx.ocsp Lua module for more details.

This feature requires at least ngx_lua v0.10.0.

Back to TOC

ndk.set_var.DIRECTIVE

syntax: res = ndk.set_var.DIRECTIVE_NAME

context: init_worker_by_lua*, set_by_lua*, rewrite_by_lua*, access_by_lua*, content_by_lua*, header_filter_by_lua*, body_filter_by_lua*, log_by_lua*, ngx.timer.*, balancer_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

This mechanism allows calling other nginx C modules’ directives that are implemented by Nginx Devel Kit (NDK)’s set_var submodule’s ndk_set_var_value.

For example, the following set-misc-nginx-module directives can be invoked this way:

For instance,

  1. local res = ndk.set_var.set_escape_uri('a/b');
  2. -- now res == 'a%2fb'

Similarly, the following directives provided by encrypted-session-nginx-module can be invoked from within Lua too:

This feature requires the ngx_devel_kit module.

Back to TOC

coroutine.create

syntax: co = coroutine.create(f)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, init_by_lua*, ngx.timer.*, header_filter_by_lua*, body_filter_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Creates a user Lua coroutines with a Lua function, and returns a coroutine object.

Similar to the standard Lua coroutine.create API, but works in the context of the Lua coroutines created by ngx_lua.

This API was first usable in the context of init_by_lua* since the 0.9.2.

This API was first introduced in the v0.6.0 release.

Back to TOC

coroutine.resume

syntax: ok, … = coroutine.resume(co, …)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, init_by_lua*, ngx.timer.*, header_filter_by_lua*, body_filter_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Resumes the executation of a user Lua coroutine object previously yielded or just created.

Similar to the standard Lua coroutine.resume API, but works in the context of the Lua coroutines created by ngx_lua.

This API was first usable in the context of init_by_lua* since the 0.9.2.

This API was first introduced in the v0.6.0 release.

Back to TOC

coroutine.yield

syntax: … = coroutine.yield(…)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, init_by_lua*, ngx.timer.*, header_filter_by_lua*, body_filter_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Yields the execution of the current user Lua coroutine.

Similar to the standard Lua coroutine.yield API, but works in the context of the Lua coroutines created by ngx_lua.

This API was first usable in the context of init_by_lua* since the 0.9.2.

This API was first introduced in the v0.6.0 release.

Back to TOC

coroutine.wrap

syntax: co = coroutine.wrap(f)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, init_by_lua*, ngx.timer.*, header_filter_by_lua*, body_filter_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Similar to the standard Lua coroutine.wrap API, but works in the context of the Lua coroutines created by ngx_lua.

This API was first usable in the context of init_by_lua* since the 0.9.2.

This API was first introduced in the v0.6.0 release.

Back to TOC

coroutine.running

syntax: co = coroutine.running()

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, init_by_lua*, ngx.timer.*, header_filter_by_lua*, body_filter_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Identical to the standard Lua coroutine.running API.

This API was first usable in the context of init_by_lua* since the 0.9.2.

This API was first enabled in the v0.6.0 release.

Back to TOC

coroutine.status

syntax: status = coroutine.status(co)

context: rewrite_by_lua*, access_by_lua*, content_by_lua*, init_by_lua*, ngx.timer.*, header_filter_by_lua*, body_filter_by_lua*, ssl_certificate_by_lua*, ssl_session_fetch_by_lua*, ssl_session_store_by_lua*

Identical to the standard Lua coroutine.status API.

This API was first usable in the context of init_by_lua* since the 0.9.2.

This API was first enabled in the v0.6.0 release.

Back to TOC

Obsolete Sections

This section is just holding obsolete documentation sections that have been either renamed or removed so that existing links over the web are still valid.

Back to TOC

Special PCRE Sequences

This section has been renamed to Special Escaping Sequences.

原文: https://github.com/chaoslawful/lua-nginx-module/blob/master/README.markdown