Pastebin

To give you a taste of what a real Rocket application looks like, this section of the guide is a tutorial on how to create a Pastebin application in Rocket. A pastebin is a simple web application that allows users to upload a text document and later retrieve it via a special URL. They’re often used to share code snippets, configuration files, and error logs. In this tutorial, we’ll build a simple pastebin service that allows users to upload a file from their terminal. The service will respond back with a URL to the uploaded file.

Finished Product

A souped-up, completed version of the application you’re about to build is deployed live at paste.rs. Feel free to play with the application to get a feel for how it works. For example, to upload a text document named test.txt, you can do:

  1. curl --data-binary @test.txt https://paste.rs/

The finished product is composed of the following routes:

  • index: GET / - returns a simple HTML page with instructions about how to use the service
  • upload: POST / - accepts raw data in the body of the request and responds with a URL of a page containing the body’s content
  • retrieve: GET /id> - retrieves the content for the paste with id <id>

Getting Started

Let’s get started! First, create a fresh Cargo binary project named rocket-pastebin:

  1. cargo new --bin rocket-pastebin
  2. cd rocket-pastebin

Then add the usual Rocket dependencies to the Cargo.toml file:

  1. [dependencies]
  2. rocket = "0.3.17"
  3. rocket_codegen = "0.3.17"

And finally, create a skeleton Rocket application to work off of in src/main.rs:

  1. #![feature(plugin)]
  2. #![plugin(rocket_codegen)]
  3. extern crate rocket;
  4. fn main() {
  5. rocket::ignite().launch();
  6. }

Ensure everything works by running the application:

  1. cargo run

At this point, we haven’t declared any routes or handlers, so visiting any page will result in Rocket returning a 404 error. Throughout the rest of the tutorial, we’ll create the three routes and accompanying handlers.

Index

The first route we’ll create is the index route. This is the page users will see when they first visit the service. As such, the route should field requests of the form GET /. We declare the route and its handler by adding the index function below to src/main.rs:

  1. #[get("/")]
  2. fn index() -> &'static str {
  3. "
  4. USAGE
  5. POST /
  6. accepts raw data in the body of the request and responds with a URL of
  7. a page containing the body's content
  8. GET /<id>
  9. retrieves the content for the paste with id `<id>`
  10. "
  11. }

This declares the index route for requests to GET / as returning a static string with the specified contents. Rocket will take the string and return it as the body of a fully formed HTTP response with Content-Type: text/plain. You can read more about how Rocket formulates responses at the API documentation for the Responder trait.

Remember that routes first need to be mounted before Rocket dispatches requests to them. To mount the index route, modify the main function so that it reads:

  1. fn main() {
  2. rocket::ignite().mount("/", routes![index]).launch();
  3. }

You should now be able to cargo run the application and visit the root path (/) to see the text being displayed.

Uploading

The most complicated aspect of the pastebin, as you might imagine, is handling upload requests. When a user attempts to upload a pastebin, our service needs to generate a unique ID for the upload, read the data, write it out to a file or database, and then return a URL with the ID. We’ll take each of these one step at a time, beginning with generating IDs.

Unique IDs

Generating a unique and useful ID is an interesting topic, but it is outside the scope of this tutorial. Instead, we simply provide the code for a PasteID structure that represents a probably unique ID. Read through the code, then copy/paste it into a new file named paste_id.rs in the src/ directory:

  1. use std::fmt;
  2. use std::borrow::Cow;
  3. use rand::{self, Rng};
  4. /// Table to retrieve base62 values from.
  5. const BASE62: &'static [u8] = b"0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz";
  6. /// A _probably_ unique paste ID.
  7. pub struct PasteID<'a>(Cow<'a, str>);
  8. impl<'a> PasteID<'a> {
  9. /// Generate a _probably_ unique ID with `size` characters. For readability,
  10. /// the characters used are from the sets [0-9], [A-Z], [a-z]. The
  11. /// probability of a collision depends on the value of `size` and the number
  12. /// of IDs generated thus far.
  13. pub fn new(size: usize) -> PasteID<'static> {
  14. let mut id = String::with_capacity(size);
  15. let mut rng = rand::thread_rng();
  16. for _ in 0..size {
  17. id.push(BASE62[rng.gen::<usize>() % 62] as char);
  18. }
  19. PasteID(Cow::Owned(id))
  20. }
  21. }
  22. impl<'a> fmt::Display for PasteID<'a> {
  23. fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
  24. write!(f, "{}", self.0)
  25. }
  26. }

Then, in src/main.rs, add the following after extern crate rocket:

  1. extern crate rand;
  2. mod paste_id;
  3. use paste_id::PasteID;

Finally, add a dependency for the rand crate to the Cargo.toml file:

  1. [dependencies]
  2. rand = "0.3"

Then, ensure that your application builds with the new code:

  1. cargo build

You’ll likely see many “unused” warnings for the new code we’ve added: that’s okay and expected. We’ll be using the new code soon.

Processing

Believe it or not, the hard part is done! (whew!).

To process the upload, we’ll need a place to store the uploaded files. To simplify things, we’ll store the uploads in a directory named upload/. Create an upload directory next to the src directory:

  1. mkdir upload

For the upload route, we’ll need to use a few items:

  1. use std::io;
  2. use std::path::Path;
  3. use rocket::Data;
  4. use rocket::http::RawStr;

The Data structure is key here: it represents an unopened stream to the incoming request body data. We’ll use it to efficiently stream the incoming request to a file.

Upload Route

We’re finally ready to write the upload route. Before we show you the code, you should attempt to write the route yourself. Here’s a hint: a possible route and handler signature look like this:

  1. #[post("/", data = "<paste>")]
  2. fn upload(paste: Data) -> io::Result<String>

Your code should:

  1. Create a new PasteID of a length of your choosing.
  2. Construct a filename inside upload/ given the PasteID.
  3. Stream the Data to the file with the constructed filename.
  4. Construct a URL given the PasteID.
  5. Return the URL to the client.

Here’s our version (in src/main.rs):

  1. #[post("/", data = "<paste>")]
  2. fn upload(paste: Data) -> io::Result<String> {
  3. let id = PasteID::new(3);
  4. let filename = format!("upload/{id}", id = id);
  5. let url = format!("{host}/{id}\n", host = "http://localhost:8000", id = id);
  6. // Write the paste out to the file and return the URL.
  7. paste.stream_to_file(Path::new(&filename))?;
  8. Ok(url)
  9. }

Make sure that the route is mounted at the root path:

  1. fn main() {
  2. rocket::ignite().mount("/", routes![index, upload]).launch();
  3. }

Test that your route works via cargo run. From a separate terminal, upload a file using curl. Then verify that the file was saved to the upload directory with the correct ID:

  1. cargo run
  2. echo "Hello, world." | curl --data-binary @- http://localhost:8000
  3. <ctrl-c> # kill running process
  4. ls upload # ensure the upload is there
  5. cat upload/* # ensure that contents are correct

Note that since we haven’t created a GET /<id> route, visiting the returned URL will result in a 404. We’ll fix that now.

Retrieving Pastes

The final step is to create the retrieve route which, given an <id>, will return the corresponding paste if it exists.

Here’s a first take at implementing the retrieve route. The route below takes in an <id> as a dynamic path element. The handler uses the id to construct a path to the paste inside upload/, and then attempts to open the file at that path, optionally returning the File if it exists. Rocket treats a None Responder as a 404 error, which is exactly what we want to return when the requested paste doesn’t exist.

  1. use std::fs::File;
  2. use rocket::http::RawStr;
  3. #[get("/<id>")]
  4. fn retrieve(id: &RawStr) -> Option<File> {
  5. let filename = format!("upload/{id}", id = id);
  6. File::open(&filename).ok()
  7. }

Unfortunately, there’s a problem with this code. Can you spot the issue? The RawStr type should tip you off!

The issue is that the user controls the value of id, and as a result, can coerce the service into opening files inside upload/ that aren’t meant to be opened. For instance, imagine that you later decide that a special file upload/_credentials.txt will store some important, private information. If the user issues a GET request to /_credentials.txt, the server will read and return the upload/_credentials.txt file, leaking the sensitive information. This is a big problem; it’s known as the full path disclosure attack, and Rocket provides the tools to prevent this and other kinds of attacks from happening.

To prevent the attack, we need to validate id before we use it. Since the id is a dynamic parameter, we can use Rocket’s FromParam trait to implement the validation and ensure that the id is a valid PasteID before using it. We do this by implementing FromParam for PasteID in src/paste_id.rs, as below:

  1. use rocket::request::FromParam;
  2. /// Returns `true` if `id` is a valid paste ID and `false` otherwise.
  3. fn valid_id(id: &str) -> bool {
  4. id.chars().all(|c| {
  5. (c >= 'a' && c <= 'z')
  6. || (c >= 'A' && c <= 'Z')
  7. || (c >= '0' && c <= '9')
  8. })
  9. }
  10. /// Returns an instance of `PasteID` if the path segment is a valid ID.
  11. /// Otherwise returns the invalid ID as the `Err` value.
  12. impl<'a> FromParam<'a> for PasteID<'a> {
  13. type Error = &'a RawStr;
  14. fn from_param(param: &'a RawStr) -> Result<PasteID<'a>, &'a RawStr> {
  15. match valid_id(param) {
  16. true => Ok(PasteID(Cow::Borrowed(param))),
  17. false => Err(param)
  18. }
  19. }
  20. }

Then, we simply need to change the type of id in the handler to PasteID. Rocket will then ensure that <id> represents a valid PasteID before calling the retrieve route, preventing attacks on the retrieve route:

  1. #[get("/<id>")]
  2. fn retrieve(id: PasteID) -> Option<File> {
  3. let filename = format!("upload/{id}", id = id);
  4. File::open(&filename).ok()
  5. }

Note that our valid_id function is simplistic and could be improved by, for example, checking that the length of the id is within some known bound or potentially blacklisting sensitive files as needed.

The wonderful thing about using FromParam and other Rocket traits is that they centralize policies. For instance, here, we’ve centralized the policy for valid PasteIDs in dynamic parameters. At any point in the future, if other routes are added that require a PasteID, no further work has to be done: simply use the type in the signature and Rocket takes care of the rest.

Conclusion

That’s it! Ensure that all of your routes are mounted and test your application. You’ve now written a simple (~75 line!) pastebin in Rocket! There are many potential improvements to this small application, and we encourage you to work through some of them to get a better feel for Rocket. Here are some ideas:

  • Add a web form to the index where users can manually input new pastes. Accept the form at POST /. Use format and/or rank to specify which of the two POST / routes should be called.
  • Support deletion of pastes by adding a new DELETE /<id> route. Use PasteID to validate <id>.
  • Limit the upload to a maximum size. If the upload exceeds that size, return a 206 partial status code. Otherwise, return a 201 created status code.
  • Set the Content-Type of the return value in upload and retrieve to text/plain.
  • Return a unique “key” after each upload and require that the key is present and matches when doing deletion. Use one of Rocket’s core traits to do the key validation.
  • Add a PUT /<id> route that allows a user with the key for <id> to replace the existing paste, if any.
  • Add a new route, GET /<id>/<lang> that syntax highlights the paste with ID <id> for language <lang>. If <lang> is not a known language, do no highlighting. Possibly validate <lang> with FromParam.
  • Use the local module to write unit tests for your pastebin.
  • Dispatch a thread before launching Rocket in main that periodically cleans up idling old pastes in upload/.

You can find the full source code for the completed pastebin tutorial on GitHub.