Add missing mode property to fake file wrapper

[mrbean-bremen]

• conforms to the IO protocol
• fixes Bug: FakeFileOpen objects don't implement the mode property required by typing.io.IO #1162

Tasks

• Unit tests added that reproduce the issue or prove feature is working
• Fix or feature added
• Entry to release notes added
• Pre-commit CI shows no errors
• Unit tests passing

[evgenii-moriakhin]

to clarify how I discovered this bug - it might help you better handle different values for the mode parameter:I came across this code in the imp module (imp.load_source) (Python 3.9):

class _HackedGetData:
    """Compatibility support for 'file' arguments of various load_*()
    functions."""

    def __init__(self, fullname, path, file=None):
        super().__init__(fullname, path)
        self.file = file

    def get_data(self, path):
        """Gross hack to contort loader to deal w/ load_*()'s bad API."""
        if self.file and path == self.path:
            # The contract of get_data() requires us to return bytes. Reopen the
            # file in binary mode if needed.
            if not self.file.closed:
                file = self.file
                if 'b' not in file.mode:
                    file.close()
            if self.file.closed:
                self.file = file = open(self.path, 'rb')

            with file:
                return file.read()
        else:
            return super().get_data(path)

This code relies on checking for "b" (binary) mode.
For completeness, it might make sense to add these values too

[mrbean-bremen]

Thanks, good point - I forgot that!

[mrbean-bremen]

Ok, added the check for binary mode, please check!

[evgenii-moriakhin]

from doc https://docs.python.org/3/library/functions.html#open

Modes 'r+' and 'r+b' open the file with no truncation.

in the current code here

       if self.open_modes.truncate:
            if self._binary:
                return "rb+" if self.open_modes.can_read else "wb"

this is probably a typo

although I see a separate test on r+b below, and maybe I haven't studied how the truncate flag works in the fake file

[mrbean-bremen]

All these tests are run both in the real and in the fake fs, so no, this is not a typo - I just reproduced the behavior in the real file system, though that was unexpected to me, too.

[mrbean-bremen]

I also tested this locally, though I haven't found an explanation for this behavior.

[mrbean-bremen]

Ok, I'm merging this then.

[mrbean-bremen]

BTW, I've been thinking about the r+/w+ behavior, and I think that this is because while there is a difference on opening with r+ and w+ (w+ will truncate the file, r+ will not), after the fie has been opened, the behavior is exactly the same for both modes (e.g. both allow reading and writing). Returning the mode in both cases as "r+" may be a design decision.