pw_docgen#

The docgen module provides tools to generate documentation for Pigweed-based projects, and for Pigweed itself.

Pigweed-based projects typically use a subset of Pigweed’s modules and add their own product-specific modules on top of that, which may have product-specific documentation. Docgen provides a convenient way to combine all of the relevant documentation for a project into one place, allowing downstream consumers of release bundles (e.g. factory teams, QA teams, beta testers, etc.) to have a unified source of documentation early on.

The documentation generation is integrated directly into the build system. Any build target can depend on documentation, which allows it to be included as part of a factory release build, for example. Additionally, documentation itself can depend on other build targets, such as report cards for binary size/profiling. Any time the code is changed, documentation will be regenerated with the updated reports.

Documentation Overview#

Each Pigweed module provides documentation describing its functionality, use cases, and programming API.

Included in a module’s documentation are report cards which show an overview of the module’s size cost and performance benchmarks. These allow prospective users to evaluate the impact of including the module in their projects.

Build Integration#

Pigweed documentation files are written in reStructuredText format and rendered to HTML using Sphinx through Pigweed’s GN build system.

There are additonal Sphinx plugins used for rendering diagrams within reStructuredText files including:

Documentation source and asset files are placed alongside code within a module and registered as a pw_doc_group target within a BUILD.gn file. These groups become available for import within a special documentation generation target, which accumulates all of them and renders the resulting HTML. This system can either be used directly within Pigweed, or integrated into a downstream project.

GN Templates#

pw_doc_group#

The main template for defining documentation files is pw_doc_group. It is used to logically group a collection of documentation source files and assets. Each Pigweed module is expected to provide at least one pw_doc_group target defining the module’s documentation. A pw_doc_group can depend on other groups, causing them to be built with it.

Arguments

  • sources: RST documentation source files.

  • inputs: Additional resources required for the docs (images, data files, etc.)

  • group_deps: Other pw_doc_group targets required by this one.

  • report_deps: Report card generating targets (e.g. pw_size_diff) on which the docs depend.

  • other_deps: Any other GN targets that should be run before this pw_doc_group runs that is not included in one of the above dep categories.

Example

pw_doc_group("my_doc_group") {
  sources = [ "docs.rst" ]
  inputs = [ "face-with-tears-of-joy-emoji.svg" ]
  group_deps = [ ":sub_doc_group" ]
  report_deps = [ ":my_size_report" ]
}

pw_doc_gen#

The pw_doc_gen template creates a target which renders complete HTML documentation for a project. It depends on registered pw_doc_group targets and creates an action which collects and renders them.

To generate the complete docs, the template also requires a conf.py file configuring Sphinx’s output, and a top level index.rst for the main page of the documentation. These are added at the root level of the built documentation to tie everything together.

Arguments

  • conf: Path to the conf.py to use for Sphinx.

  • index: Path to the top-level index.rst file.

  • output_directory: Directory in which to render HTML output.

  • deps: List of all pw_doc_group targets required for the documentation.

  • python_metadata_deps: Python-related dependencies that are only used as deps for generating Python package metadata list, not the overall documentation generation. This should rarely be used by non-Pigweed code.

Example

pw_doc_gen("my_docs") {
  conf = "//my_docs/conf.py"
  index = "//my_docs/index.rst"
  output_directory = target_gen_dir
  deps = [
    "//my_module:my_doc_group",
  ]
}

Generating Documentation#

All source files listed under a pw_doc_gen target and its pw_doc_group dependencies get copied out into a directory structure mirroring the original layout of the modules in which the sources appear. This is demonstrated below using a subset of Pigweed’s core documentation.

Consider the following target in $dir_pigweed/docs/BUILD.gn:

pw_doc_gen("docs") {
  conf = "conf.py"
  index = "index.rst"
  output_directory = target_gen_dir
  deps = [
    "$dir_pw_bloat:docs",
    "$dir_pw_docgen:docs",
    "$dir_pw_preprocessor:docs",
  ]
}

A documentation tree is created under the output directory. Each of the sources and inputs in the target’s dependency graph is copied under this tree in the same directory structure as they appear under the root GN build directory ($dir_pigweed in this case). The conf.py and index.rst provided directly to the pw_doc_gen template are copied in at the root of the tree.

out/gen/docs/pw_docgen_tree/
├── conf.py
├── index.rst
├── pw_bloat
│   ├── bloat.rst
│   └── examples
│       └── simple_bloat.rst
├── pw_docgen
│   └── docgen.rst
└── pw_preprocessor
    └── docs.rst

This is the documentation tree which gets passed to Sphinx to build HTML output. Imports within documentation files must be relative to this structure. In practice, relative imports from within modules’ documentation groups are identical to the project’s directory structure. The only special case is the top-level index.rst file’s imports; they must start from the project’s build root.

Viewing Documentation#

pw_docgen includes a web server that serves locally-generated documentation at pw_docgen.docserver. It supports hot-reloading, so the rendered docs in your browser will refresh as you make changes to the source files.

In most cases, you will not need to run the docs server directly. Instead, it will be run via pw_watch.

Sphinx Extensions#

This module houses Pigweed-specific extensions for the Sphinx documentation generator. Extensions are included and configured in docs/conf.py.

module_metadata#

Per SEED-0102, Pigweed module documentation has a standard format. The pigweed-module Sphinx directive provides that format and registers module metadata that can be used elsewhere in the Sphinx build.

We need to add the directive after the document title, and add a class to the document title to achieve the title & subtitle formatting. Here’s an example:

.. rst-class:: with-subtitle

=========
pw_string
=========

.. pigweed-module::
   :name: pw_string
   :tagline: Efficient, easy, and safe string manipulation
   :status: stable
   :languages: C++17, Rust
   :code-size-impact: 500 to 1500 bytes

   Module sales pitch goes here!

Directive options#

  • name: The module name (required)

  • tagline: A very short tagline that summarizes the module (required)

  • status: One of experimental, unstable, and stable (required)

  • is-deprecated: A flag indicating that the module is deprecated

  • languages: A comma-separated list of languages the module supports. If the language has API docs (Rust), they will be linked from the metadata block.

  • code-size-impact: A summarize of the average code size impact

Canonical URL configuration#

module_metadata fixes the canonical URLs for */docs.html pages. By default Sphinx assumes that a page’s canonical URL is its full URL. E.g. the default canonical URL for //pw_string/docs.rst is https://pigweed.dev/pw_string/docs.html. The pigweed.dev server treats https://pigweed.dev/pw_string/ as the canonical URL however. This problem is not limited to module homepages; it occurs on any page that ends in /docs.html such as https://pigweed.dev/third_party/emboss/docs.html. module_metadata fixes this problem by ensuring that the <link rel="canonical" href="..."/> tag generated in the HTML is aligned with the server’s configuration.

After building the docs, the canonical URLs for all HTML pages can be verified by running the following command in a terminal from the root directory of the upstream Pigweed repo:

grep '<link rel="canonical' out/docs/gen/docs/html/* -R

Context: b/323077749

google_analytics#

When this extension is included and a google_analytics_id is set in the Sphinx configuration, a Google Analytics tracking tag will be added to each page of the documentation when it is rendered to HTML.

By default, the Sphinx configuration’s google_analytics_id is set automatically based on the value of the GN argument pw_docs_google_analytics_id, allowing you to control whether tracking is enabled or not in your build configuration. Typically, you would only enable this for documentation builds intended for deployment on the web.

Debugging Pigweed’s Sphinx extensions#

To step through your Pigweed extension code with pdb:

  1. Set a breakpoint in your extension code:

    breakpoint()
    
  2. Build python.install to install the code change into the bootstrap venv (environment/pigweed-venv/lib/python3.8/site-packages/pw_docgen):

    ninja -C out python.install
    
  3. Manually invoke Sphinx to build the docs and trigger your breakpoint:

    cd out
    sphinx-build -W -b html -d docs/gen/docs/help docs/gen/docs/pw_docgen_tree docs/gen/docs/html -v -v -v
    

    You should see build output from Sphinx. The build should pause at your breakpoint and you should then see pdb’s prompt ((Pdb)).