Add initial types to repo/fun.py #1192
Conversation
|
Finally fixed all automated issues! All functions in base.py and fun.py now typed and passes mypy and pyright (and flake8) |
There was a problem hiding this comment.
Thanks a lot for this tremendous work!
Admittedly I only got to fly through about 2/3 of all changes and think there is a high likelihood of me having missed something.
I will see to finish the rest tomorrow but would hope that another maintainer could also have a look.
|
|
||
| # Subclass configuration | ||
| # Subclasses may easily bring in their own custom types by placing a constructor or type here | ||
| GitCommandWrapperType = Git | ||
|
|
||
| def __init__(self, path=None, odbt=GitCmdObjectDB, search_parent_directories=False, expand_vars=True): | ||
| def __init__(self, path: Optional[PathLike] = None, odbt: Type[GitCmdObjectDB] = GitCmdObjectDB, |
There was a problem hiding this comment.
There is an actual base class for object databases which is more general.
There are two implementations for it.
There was a problem hiding this comment.
I think it's best to 'make up' a type and call it ObjectDB, as LooseObjectDB is too specific and kind of wrong here, too. Honestly I don't really know what I was thinking and this clearly is an example of OOP being gone wrong.
Otherwise you could dig into GitDB and see how this DB is composed of various types which provide a bundle of methods which may or may not be that useful in practice.
I do recommend to not get into types of GitDB and instead make them up (inside of GitPython) based on what's common usage in GitPython instead.
| """Create a new head within the repository. | ||
| For more documentation, please see the Head.create method. | ||
|
|
||
| :return: newly created Head Reference""" | ||
| return Head.create(self, path, commit, force, logmsg) | ||
|
|
||
| def delete_head(self, *heads, **kwargs): | ||
| def delete_head(self, *heads: HEAD, **kwargs: Any) -> None: |
There was a problem hiding this comment.
I think HEAD is too specific - any symbolic reference can be created that way.
| @@ -429,11 +468,16 @@ def _get_config_path(self, config_level): | |||
| elif config_level == "global": | |||
| return osp.normpath(osp.expanduser("~/.gitconfig")) | |||
| elif config_level == "repository": | |||
| return osp.normpath(osp.join(self._common_dir or self.git_dir, "config")) | |||
There was a problem hiding this comment.
It's probably nicer to do it like this:
if not (self._common_dir or self.git_dir):
raise …
return osp.normpath(osp.join(self._common_dir or self.git_dir, "config"))That way there is no duplicate path transformation.
|
As this PR touches a lot of files I am merging it nonetheless even without the requested changes. These can be contributed separately and in the meantime what's there should already be a considerable improvement. Thanks again, and I am looking forward to your future contributions :). |
|
Sorry for slow reply, I was sent to work somewhere with no internet (shocking common in the UK!) I'll make the suggested changes on next PR. |
Hi,
This adds types to base.py and fun.py.
All annotations are consistant according to both mypy and pyright (but will need refining once whole lib is typed).
I've used an alias (TBD, 'to be determined') for Any to show places where I was unsure of the type, but i'm sure it should be more specific than Any.
i've avoided changing any runtime code as much as possible, except:
i've used Union[str, os._pathlike] to annotate path arguments. Where this showed errors due to use of string-only methods, i've then wrapped the arg in str() within the function. This should mean they are all compatible with pathlib.Paths and pathlib.Purepaths, but that will require tests writing before its official.
Adding typeguards: e.g. where git_dir is initialised as None, then assigned later to a string, mypy showed places that needed to be checked that it was actually a string and not still None. This could probably be improved by initializing as "" instead but it touched too much code.
changed Except (OSError, IOError) to just Except OSError, as typechecking says it was redundant for python 3.
Separate tuple unpacking to seperate lines to provide specific types (unpacked tuples cant be typed yet)