Pulsar Functions 是轻量级计算流程，具有以下特点：

• 从一个或多个 Pulsar topic 中消费消息；
• 将用户提供的处理逻辑应用于每条消息；
• 将运行结果发布到另一个 topic。

Goals

Pulsar Functions背后的核心目标是使您能够轻松创建各种级别的复杂的的处理逻辑，而无需部署单独的类似系统(例如 Apache Storm, Apache Heron, Apache Flink, 等等) Pulsar Functions are computing infrastructure of Pulsar messaging system. The core goal is tied to a series of other goals:

• Developer productivity (language-native vs Pulsar Functions SDK functions)
• 简单的故障排查
• 操作简单（不需要外部处理系统）

Inspirations

Pulsar Functions are inspired by (and take cues from) several systems and paradigms:

• 流处理引擎，例如 Apache Storm、Apache Heron、Apache Flink
• “无服务器（Serverless）”和“Function as a Service”（FaaS）云平台，如：Amazon Web Services Lambda、Google Cloud Functions、Azure Cloud Functions 等

Pulsar Functions can be described as

• Lambda 样式的 functions
• specifically designed to use Pulsar as a message bus.

Programming model

Pulsar Functions provide a wide range of functionality, and the core programming model is simple. Functions receive messages from one or more input topics. Each time a message is received, the function will complete the following tasks.

• 将某些处理逻辑应用到输入并写入到输出：
  • An output topic in Pulsar
  • Apache BookKeeper
• Write logs to a log topic (potentially for debugging purposes)
• 增量 计数器

You can use Pulsar Functions to set up the following processing chain:

• A Python function listens for the raw-sentences topic and “sanitizes” incoming strings (removing extraneous whitespace and converting all characters to lowercase) and then publishes the results to a sanitized-sentences topic.
• A Java function listens for the sanitized-sentences topic, counts the number of times each word appears within a specified time window, and publishes the results to a results topic
• Finally, a Python function listens for the results topic and writes the results to a MySQL table.

词数统计示例

If you implement the classic word count example using Pulsar Functions, it looks something like this:

To write the function in Java with Pulsar Functions SDK for Java, you can write the function as follows.

package org.example.functions;
import org.apache.pulsar.functions.api.Context;
import org.apache.pulsar.functions.api.Function;
import java.util.Arrays;
public class WordCountFunction implements Function<String, Void> {
    // This function is invoked every time a message is published to the input topic
    @Override
    public Void process(String input, Context context) throws Exception {
        Arrays.asList(input.split(" ")).forEach(word -> {
            String counterKey = word.toLowerCase();
            context.incrCounter(counterKey, 1);
        });
        return null;
    }
}

Bundle and build the JAR file to be deployed, and then deploy it in your Pulsar cluster using the command line as follows.

$ bin/pulsar-admin functions create \
  --jar target/my-jar-with-dependencies.jar \
  --classname org.example.functions.WordCountFunction \
  --tenant public \
  --namespace default \
  --name word-count \
  --inputs persistent://public/default/sentences \
  --output persistent://public/default/count

Content-based routing example

Pulsar Functions are used in many cases. The following is a sophisticated example that involves content-based routing.

For example, a function takes items (strings) as input and publishes them to either a fruits or vegetables topic, depending on the item. Or, if an item is neither fruit nor vegetable, a warning is logged to a log topic. The following is a visual representation.

If you implement this routing functionality in Python, it looks something like this:

from pulsar import Function
class RoutingFunction(Function):
    def __init__(self):
        self.fruits_topic = "persistent://public/default/fruits"
        self.vegetables_topic = "persistent://public/default/vegetables"
    def is_fruit(item):
        return item in [b"apple", b"orange", b"pear", b"other fruits..."]
    def is_vegetable(item):
        return item in [b"carrot", b"lettuce", b"radish", b"other vegetables..."]
    def process(self, item, context):
        if self.is_fruit(item):
            context.publish(self.fruits_topic, item)
        elif self.is_vegetable(item):
            context.publish(self.vegetables_topic, item)
        else:
            warning = "The item {0} is neither a fruit nor a vegetable".format(item)
            context.get_logger().warn(warning)

If this code is stored in ~/router.py, then you can deploy it in your Pulsar cluster using the command line as follows.

$ bin/pulsar-admin functions create \
  --py ~/router.py \
  --classname router.RoutingFunction \
  --tenant public \
  --namespace default \
  --name route-fruit-veg \
  --inputs persistent://public/default/basket-items

Functions, messages and message types

Pulsar Functions take byte arrays as inputs and spit out byte arrays as output. However in languages that support typed interfaces(Java), you can write typed Functions, and bind messages to types in the following ways.

• Schema 管理
• SerDe

Fully Qualified Function Name (FQFN)

Each Pulsar Function has a Fully Qualified Function Name (FQFN) that consists of three elements: the function tenant, namespace, and function name. FQFN looks like this:

tenant/namespace/name

FQFNs enable you to create multiple functions with the same name provided that they are in different namespaces.

Supported languages

Currently, you can write Pulsar Functions in Java, Python, and Go. For details, refer to Develop Pulsar Functions.

Processing guarantees

Pulsar Functions provide three different messaging semantics that you can apply to any function.

Apply processing guarantees to a function

创建 Function 时，您可以为 Pulsar Function 设置 processing guarantees 。 The following pulsar-function create command creates a function with effectively-once guarantees applied.

$ bin/pulsar-admin functions create \
  --name my-effectively-once-function \
  --processing-guarantees EFFECTIVELY_ONCE \
  # Other function configs

The available options for --processing-guarantees are:

• ATMOST_ONCE
• ATLEAST_ONCE
• EFFECTIVELY_ONCE

By default, Pulsar Functions provide at-least-once delivery guarantees. So if you create a function without supplying a value for the --processingGuarantees flag, the function provides at-least-once guarantees.

Update the processing guarantees of a function

You can change the processing guarantees applied to a function using the update command. The following is an example.

$ bin/pulsar-admin functions update \
  --processing-guarantees ATMOST_ONCE \
  # Other function configs