[QST] How does coderabbit measure doc coverage?

What is your question?

The CodeRabbit docstring coverage evaluation is unclear and appears inconsistent with the repository’s documentation requirements.

In this PR/commit:

https://github.com/NVIDIA/cuopt/pull/1405/commits/2a98a4f2f04c966fa6818ac6d8de5d86f819bffe

CodeRabbit reports:

Docstring coverage is 71.43% which is insufficient. The required threshold is 80.00%.

Since 71.43% = 5/7, it appears CodeRabbit is counting 7 modified symbols. However, the modified methods appear to already have documentation where required, and the report does not list which 2 symbols are considered undocumented.

I then tried adding inline documentation in this commit:

https://github.com/NVIDIA/cuopt/pull/1405/commits/4dc4a444cd89ca41ee52096574700cd263c7d154

Unexpectedly, the reported coverage became:

Docstring coverage is 62.50% which is insufficient. The required threshold is 80.00%.

That looks like 5/8, meaning the change increased the denominator but did not increase the documented-symbol count.

Could you please explain the docstring coverage rules used by CodeRabbit, specifically:

1. Which symbols are counted in the numerator and denominator?
2. Are C++ structs/classes counted, or only functions/methods?
3. Are changed existing symbols counted even if the declaration itself was not modified?
4. What comment formats are recognized for C++ documentation? For example, are ///, //! \brief, and /** ... */ all accepted?
5. Is there a way to make CodeRabbit list the exact symbols considered undocumented?

Without this information, it is difficult to act on the coverage warning without unnecessary trial-and-error commits.