[stdlib] tempfile and tempdir

[artemiogr97]

Implementation for #2018

[artemiogr97]

I opened the pull to have some feedback, some points to discuss:

• Are the read methods useful in NamedTemporaryFile?
  • personally I don't know in which case I would create a temporary file and read from it at the same time
• should I implement what I need in the os module directly in this pull ?
  • TemporaryDirectory is not implemented yet because os.mkdir and os.rmdir don't exist, I guess shutil.rmtree will be necessary too (implemented in [stdlib] Implement mkdir and rmdir #2430 and [stdlib] Implementation of rmtree #2439)
  • For NamedTemporaryFile there is a slight difference with respect to the python implementation, the name attribute could end up being relative but the functions os.abspath, and os.normpath still don't exist
    • also how different platforms are going to be handled in mojo, in python there are dedicated modules posix and ntpath, I guess the same should be done in mojo?

[artemiogr97]

#2439 must be merged first

[JoeLoser]

#2439 must be merged first

Sorry for the delay here. Is it possible to decouple this PR from #2439? I think temporary file things would be useful even outside of the rmtree functionality, so I'd rather not couple them.

FYI @laszlokindrat @Dan13llljws @ConnorGray as we were talking about temp files today.

[laszlokindrat]

Sorry for the delay here. Is it possible to decouple this PR from #2439? I think temporary file things would be useful even outside of the rmtree functionality, so I'd rather not couple them.

I completely agree. @artemiogr97 Can you please rebase this on the latest nightly, and maybe start splitting it up? Also one concrete suggestion: could you please start with the low level tempfile APIs, i.e. gettempdir and mkdtemp, and maybe mkstemp, and then we can build the rest on it.

This super cool btw, we're very happy that you are pushing this forward!

[artemiogr97]

could you please start with the low level tempfile APIs, i.e. gettempdir and mkdtemp, and maybe mkstemp, and then we can build the rest on it.

@laszlokindrat I started working on this, I'm having some issues with mkstemp, in python it returns a tuple with the file object and the name, I'm not sure how to do this in mojo, the problem is that when I do the following I have an error:

tmp_file_and_name = msktemp()
tmp_file = tmp_file_and_name[0]  # error here: 'FileHandle' is not copyable because it has no '__copyinit__'
file_name = tmp_file_and_name[1]

I tried doing tmp_file = tmp_file_and_name.get[0, FileHandle]() but then I have a double free at runtime.
I guess that the best option is to return only the file handle for now? but we wont have access to the file name

Edit: I just realized that returning only the file handle is not useful, there would not be a way to delete the file after using it

[artemiogr97]

I opened a new PR (#2742) that implements gettempdir and mkdtemp, and a second PR ( #2743) for TemporaryDirectory using mkdtemp

[artemiogr97]

@JoeLoser, @laszlokindrat I opened another PR #2762 for NamedTemporaryFile seems so be easier to implement than mkstemp for now, I opened it as a draft since I'm having some issues, I'll leave you all the details in that PR.

Also I guess I can close this PR and also #2439 since everything is now splitted in #2742, #2743 and #2762