AsyncAPI Codegen

asyncapi-codegen interprets an AsyncAPI spec and generates code and documentation accordingly. Only MQTT is supported.

There are two parts to this:

• Interpret the AsyncAPI spec into Python classes in ways that makes it easier to use.
• Use Jinja2 templates to create the code.

Terminology

With pub/sub architecture, the broker is a server and everything is a client to it. As such, is is useful to define the general nature of clients to differentiate roles between clients.

A provider is the MQTT service that provides functionality. The AsyncAPI spec describes the behavior of the provider. A utilizer is the MQTT client that consumes the functionality. When the AsyncAPI spec uses "publish" and "subscribe" terms, it is the utilizer that does the described action.

Supported Languages

Lanaguage	MQTT Library	JSON Library	JSON Schema Library
C++	Mosquitto Client	rapidjson	Built-in via json-schema-codegen
Python	Paho MQTT	Python built-in	None

Features:

Feature	Python	C++
Enforces JSON Schema	No	Yes

Documentation Formats

Formats:

• Markdown (with significant HTML), suitable for display on GitLab.

Output files

The generator creates the following types of output files:

• client interface. This is the main interface for publishing or subscribing to messages.
• servers. Represents a connection to a server. Currently, only unauthenticated MQTT connections are supported. A server object is provided to a client interface to establish the connection to the broker/server.
• parameters. These are just schema objects.
• messages. Currently, these are just schema objects. However, in the future these objects will also take protocol-specific message bindings.
• schemas. These are structures that always enforce schema compliance and provide (de-)serialization methods to rapidjson Value objects (which can thus be used to create JSON strings).

Generated code files should be annotated for doxygen document generation, or at least that is the goal.

Installation

pip3 install asyncapi-codegen

Python requirements for running code generator

See also requirements.txt

• python 3.10
• jinja2
• stringcase
• json-schema-codegen
• pyyaml

Python Code Generation

Requirements for the generated python code

• python 3.7
• parse
• paho-mqtt

C++ Code Generation

Requirements for the generated C++ code

• boost (boost::optional and boost::variant among others)
• rapidjson 1.1
• C++11

Using Docker

Utilizer code

To generate utilizer/client C++ code for an AsyncAPI spec, you could run this docker command, carefully adjusting the volume mounting to the directory containing the specs and the directory for the output.

When the YAML uses the words "publish" and "subscribe", the utilizer will perform those MQTT actions.

docker run --rm -t \
    -v $(pwd)/examples:/specs \
    -v $(pwd)/output:/output \
    --user $UID:$GID \
    docker.io/pearmaster/asyncapi-codegen:latest \
        --yaml /specs/streetlights-mqtt.yml \
        --name Streetlights \
        --cpp

Provider code

If you'd like instead to generate C++ code for the provider, you can append --progtype provider to the docker command to look like:

docker run --rm -t \
    -v $(pwd)/examples:/specs \
    -v $(pwd)/output:/output \
    --user $UID:$GID \
    docker.io/pearmaster/asyncapi-codegen:latest \
        --yaml /specs/streetlights-mqtt.yml \
        --name Streetlights \
        --cpp \
        --progtype provider

Markdown documentation

This command will generate documentation for utilizers. The entire spec is output in one big markdown file.

docker run --rm -t \
    -v $(pwd)/examples:/specs \
    -v $(pwd)/output:/output \
    --user $UID:$GID \
    docker.io/pearmaster/asyncapi-codegen:latest \
        --yaml /specs/streetlights-mqtt.yml \
        --name Streetlights
        --markdown

License

GPLv2