Trivially extends dataclass constructor hinting to attrs next-gen APIs.
This will stumble in cases where attrs extends beyond the standard
dataclasses API, such as complex use of defaults, converters, et al.
However, it likely covers the vast majority of cases which fall solidly
in the intersection of the two APIs.
Extension beyond these cases could use [PEP0681 dataclass_transforms],
however this is definitely a problem for another day.
[PEP0681 dataclass_transforms]: https://peps.python.org/pep-0681/https://github.com/davidhalter/jedi/issues/1835
On integration tests file collection,
the value of `environment.executable` can also be a symlink
(e.g. in a virtualenv) with a different name than,
but pointing to the same as `sys.executable`
(e.g. .../bin/python3.10 and .../bin/python, respectively).
That causes skipping the collection of `completion/pytest.py`
and `completion/conftest.py` a lot of times, depending on the environment.
(e.g. "60 skipped" before x "23 skipped" after, in a local virtualenv)
At present, .gitignore patterns not starting with '/' are classified
as "ignored names" (opposing to "ignored paths") and not used for
filtering directories. But, according to the spec [1], the situation
is a bit different: all patterns apply to directories (and those
ending with '/' apply to directories only). Besides that, there two
kinds of patterns: those that match only w.r.t the directory where
defining .gitignore is located (they must contain a '/' in the
beginning or in the middle), which we call "absolute", and those that
also match in all subdirectories under the directory where defining
.gitignore is located (they must not contain '/' or contain only
trailing '/'), which we call "relative".
This commit implements handling of both "absolute" and "relative"
.gitignore patterns according to the spec. "Absolute" patterns are
handled mostly like `ignored_paths` were handled in the previous
implementation. "Relative" patterns are collected into a distinct set
containing `(defining_gitignore_dir, pattern)` tuples. For each
traversed `root_folder_io`, all applicable "relative" patterns are
expanded into a set of plain paths, which are then used for filtering
`folder_io`s.
While at it, also fix some minor issues. Explicitly ignore negative
and wildcard patterns, since we don't handle them correctly
anyway. Also, use '/' as a path separator instead of `os.path.sep`
when dealing with .gitignore, since the spec explicitly says that '/'
must be used on all platforms.
[1] https://git-scm.com/docs/gitignore
There are small typos in:
- jedi/api/exceptions.py
- jedi/inference/base_value.py
- jedi/inference/compiled/mixed.py
- jedi/inference/value/dynamic_arrays.py
Fixes:
- Should read `usually` rather than `ususally`.
- Should read `modifications` rather than `modfications`.
- Should read `interpreters` rather than `interpreteters`.
- Should read `inferred` rather than `inferrined`.
- Should read `completable` rather than `completeable`.
Signed-off-by: Tim Gates <tim.gates@iress.com>
If multiple directories in sys.path provide a nested namespace
package, then jedi would only visit the first directory which
contained the package. Fix this by saving the remaining path list in
the ImplicitNamespaceValue and add a test for it.
This ensures that these quoted likely forwards references in
return type annotations behave like their non-quoted equivalents.
I suspect there may be other places which will need similar
adjustments, which may mean that we should push the conversion
a layer closer to the parsing (perhaps in `py__annotations__`?).
One case I know that this doesn't solve (but which likely needs
similar adjustment) is generics in return types of comment-style
annotations. They're less likely and may not be worth supporting
since all supported Python versions can use the in-syntax spelling
for annotations at this point.
