GitHub-issue-270: #270
Radicle-issue-0d6a513: https://app.radicle.xyz/nodes/seed.radicle.garden/rad:z23ZVuA1DWS99PDJ1rcarCtJi99x1/issues/0d6a51328f554f235c38a2a652b844c4ba21bba5

…d of set

Issue-267: #267

…ead of a sequence

Issue-268: #268

Issue-271: #271

…hin helpers

Issue-272: #272

Issue-272: #272

We observed that Griffe's internals/API were a bit confusing, so decided to refactor them.

Griffe has a concept of "exported member". A **module member** is considered:

- **explicitly exported** when its name appears in the module's `__all__`
- **implicitly exported** when the module's `__all__` is undefined, the member is available at runtime (for example not guarded behind `if TYPE_CHECKING`), and the member is an alias (defined elsewhere), *except* if it's a submodule, *unless* it was also imported in the current module

A **class member** is considered **implicitly exported**, the same way as module members are considered implicitly exported. It is never considered explicitly exported because we don't/can't use `__all__` in class bodies.

This "exported" marker is then used in a few places, 3 of of them being of interest to us here.

- if `__all__` is defined in the wildcard imported module, add all objects referenced in the imported module's `__all__` to the importing module
- else, add all objects that are implicitly exported by the imported module to the importing module

There is a first issue here: members starting with `_` are considered implicitly exported. This is wrong: actual wildcard imports will not import these objects at runtime.

Between the same old and same new module, if a member disappeared, and it was either a submodule, or it was exported (explicitly or implicitly), we consider its disappearance a breaking change. Again, it's wrong here to consider that a private object's disappearance makes a breaking change. This issue wasn't spotted before because the code that checks breaking changes has an earlier condition on the object names: it skips private objects anyway.

Griffe objects have an `is_public()` method that tell if an object is public or not. An object is:

- **strictly public** if:
  - its `public` attribute is true
  - or it is explicitly exported
- **public** if:
  - it is strictly public
  - or it's not private (name not starting with `_`) and it's not imported

Griffe gained this method later on, to be used by mkdocstrings-python.

---

"Exported" is usually used to say "public" (we export an object to make it public), but in Griffe it's rather used to say "the object is exposed in regard to the wildcard import runtime mechanism". So we renamed "exported" to "wildcard exposed".

Then we fixed the issue mentioned above about private members being considered implicitly exposed. And since wildcard imports are not configurable, we dropped the distinction between explicitly and implicitly exposed. An object is either exposed, or it's not.

Finally the code that checks breaking changes now uses the "public" API instead of the "exposed" one.

Issue-281: #281

…`Stats`

This change gives the Parameter class the `docstring` and `function` attributes. 

The docstring value is only populated for dataclasses.
This makes it possible to access this information for all dataclass parameters,
especially those set as `InitVar` where it was not possible to get it
since they are not members (attributes) of the class.

This slot can also be used to support things like PEP 727 (with extensions).
In the future, we might add such a docstring slot to yielded, received and returned values too.

Issue-286: #286
Related-to-mkdocstrings/griffe#252: #252
PR-288: #288
Co-authored-by: Timothée Mazzucotelli <[email protected]>

This doesn't solve the issue that `functools.partial` does not set the `__module__` attribute of the resulting partial to be the module of the wrapped function, which means that the inspector always treats partials as aliases (since it finds out they were defined in the `functools` module).

This issue cannot be fixed on our side, because special-casing partials to decide that they are always defined in the current module would not work for partials imported from elsewhere.

Therefore, devs *must* set the `__module__` attribute of partials themselves.