[et_xmlfile] Complete the writer_cm annotation

[eddiesr93]

xmlfile.writer_cm was Incomplete. It is initialised to None in __init__ and assigned the result of the @contextmanager-decorated incremental_tree._get_writer in xmlfile.__enter__, so its type is _GeneratorContextManager[tuple[ET._FileWrite, str]] | None.

The context manager yields (write, encoding): the first element is passed straight to _IncrementalFileWriter, whose output_file parameter is already typed as ET._FileWrite, and the second is the encoding str.

Verified locally with stubtest, mypy, and pyright.

[srittau]

I think we can actually type the first tuple element a bit more accurate by using:

Suggested change

	writer_cm: _GeneratorContextManager[tuple[ET._FileWrite, str]] | None
	writer_cm: _GeneratorContextManager[tuple[SupportsWrite[str], str]] | None

(SupportsWrite must be imported from _typeshed.) _FileWrite is a type alias, of which one element is SupportsWrite. But the other elements of this type alias don't make much sense if you look at how its used:

https://foss.heptapod.net/openpyxl/et_xmlfile/-/blob/branch/default/et_xmlfile/xmlfile.py?ref_type=heads#L91

This will actually just call the writer with a str.

[eddiesr93]

_get_writer yields the bound write method itself, and _IncrementalFileWriter calls it directly (self._file(...)), so I went with Callable[[str], object] instead of SupportsWrite[str]. Aligned _IncrementalFileWriter.__init__ too, it gets the same value.

[github-actions]

According to mypy_primer, this change has no effect on the checked open source code. 🤖🎉