sphinx-rustdoc-postprocess

Post-process sphinxcontrib-rust RST output, converting leftover markdown fragments (code fences, tables, links, headings, inline code) to proper RST via pandoc.

The problem

sphinxcontrib-rust generates RST files from Rust crates, but rustdoc doc-comments are written in markdown. The generated RST ends up with markdown fragments embedded verbatim inside directive bodies, which Sphinx cannot render correctly.

For a real-world example, see rgpot (source), which uses this extension to document its Rust core library alongside C++ API docs.

Before (raw sphinxcontrib-rust output)

Given Rust doc-comments like these in lib.rs:

//! ## Module Overview
//!
//! | Module | Purpose |
//! |--------|---------|
//! | [`types`] | `#[repr(C)]` data structures for force/energy I/O |
//! | [`tensor`] | DLPack tensor helpers |

sphinxcontrib-rust produces RST with the markdown still intact inside directives:

.. py:module:: rgpot_core

   ## Module Overview

   | Module | Purpose |
   |--------|---------|
   | [`types`] | `#[repr(C)]` data structures for force/energy I/O |
   | [`tensor`] | DLPack tensor helpers |

This renders incorrectly in Sphinx: headings inside directives break the document structure, markdown tables appear as literal pipe characters, and single-backtick code is not valid RST.

After (postprocessed output)

After this extension runs, the same file becomes:

.. py:module:: rgpot_core

   **Module Overview**

   +------------+----------------------------------------------------+
   | Module     | Purpose                                            |
   +============+====================================================+
   | ``types``  | ``#[repr(C)]`` data structures for force/energy IO |
   +------------+----------------------------------------------------+
   | ``tensor`` | DLPack tensor helpers                              |
   +------------+----------------------------------------------------+

Similarly, markdown code fences:

```c
rgpot_status_t s = rgpot_potential_calculate(pot, &input, &output);
if (s != RGPOT_SUCCESS) {
    fprintf(stderr, "rgpot error: %s\n", rgpot_last_error());
}
```

become proper RST code-block directives:

.. code-block:: c

   rgpot_status_t s = rgpot_potential_calculate(pot, &input, &output);
   if (s != RGPOT_SUCCESS) {
       fprintf(stderr, "rgpot error: %s\n", rgpot_last_error());
   }

And markdown links like [metatensor](https://docs.metatensor.org/) become `metatensor <https://docs.metatensor.org/>`_, while rustdoc intra-doc links like [`types`] become types .

Installation

pip install sphinx-rustdoc-postprocess

Pandoc must be available on your PATH. See pandoc.org for installation instructions.

Usage

Add to your Sphinx conf.py:

extensions = [
    "sphinxcontrib_rust",
    "sphinx_rustdoc_postprocess",
]

The extension hooks into builder-inited at priority 600 (after sphinxcontrib-rust's default 500), so it automatically runs on the generated RST files before Sphinx reads them.

Configuration

Full configuration example

# conf.py
import os

extensions = [
    "sphinxcontrib_rust",
    "sphinx_rustdoc_postprocess",
]

# sphinxcontrib-rust settings
rust_crates = {
    "my_crate": os.path.abspath("../../my-crate/"),
}
rust_doc_dir = os.path.join(os.path.dirname(os.path.abspath(__file__)), "crates")
rust_rustdoc_fmt = "rst"

# Inject a toctree entry for the Rust docs into an existing index page
rustdoc_postprocess_toctree_target = "api/index.rst"
rustdoc_postprocess_toctree_rst = """

Rust API
--------

.. toctree::
   :maxdepth: 2

   ../crates/my_crate/lib
"""

What gets converted

Development

pixi install
pixi run test

A pre-commit job is setup on CI to enforce consistent styles, so it is best to set it up locally as well (using uvx for isolation):

# Run before committing
uvx pre-commit run --all-files

License

MIT