Before adding more speedup functions for asyncio, I want to rename the module.

Attached patch is tested on Mac 10.11. I'll test it on Windows later.

Why do you need to rename the module ?

Because:

* The "futures" name is used in concurrent.futures too. This module is for asyncio.
* Currently, this module has only Future class. But there are several ideas about adding C speedup to improve asyncio performance. (e.g. buffer slicing and Task class).

See also: http://bugs.python.org/issue26081#msg278368

I have read and tested your patch (with the tests) and you can merge it. But I can't review it via retvield.

Since this patch renames file, I used --git option of hg diff.
But Rietvelt seems doesn't accept git format patch.

Attached patch is same to previous, but generated without --git option.

I have reviewed your patch, seems to be fine. And thank you for your new patch without the --git flag.

Any reason why this module doesn't use argument clinic?

> Any reason why this module doesn't use argument clinic?

No reason.

But I don't want to do in this issue, since I don't know how to
generate good patch file for file rename + change in file.

I'll do it in future issue.

I hit this issue27705, and I have no time to reinstall Visual Studio 2015 Community today.
I'll try test on Windows later.

I found _futures module is not used on Windows for now (without this patch).
Because `{"_future", PyInit__future},` is not in PC/config.c

But, when it added, test_windows_events cause infinite loop.
This should be another issue.

With this (asyncio-speedups2.patch), build on Windows succeeds,
and test succeeds (without speedup enabled).

I want to apply this patch before fixing the issue and adding _asyncio to PC/config.c
Is it OK?

Sure. The build-specific changes for Windows look fine to me.

There is also an argument that this should not be a built-in module but should be separate (like _ctypes). I don't know if that is just a Windows decision or not, but I'm guessing you weren't the one who made that decision originally. Issue for another time

> There is also an argument that this should not be a built-in module but should be separate (like _ctypes). I don't know if that is just a Windows decision or not, but I'm guessing you weren't the one who made that decision originally. Issue for another time

Since _overlapped is separated extension, I think _future should be too.
I'll try it after fixing issue on windows.

Can I commit this patch without NEWS entry?
Only notable change is module name, but I don't expect no one import
it directly.

If it is required, it would be like:

- Issue #28428: Rename _futures module to _asyncio. It will have more
  speedup functions or classes other than asyncio.Future.

No news entry is necessary for this.

New changeset 9d06fdedae2b by INADA Naoki in branch '3.6':
Issue #28428: Rename _futures module to _asyncio.
https://hg.python.org/cpython/rev/9d06fdedae2b

New changeset c2f3b7c56dff by INADA Naoki in branch 'default':
Issue #28428: Rename _futures module to _asyncio. (merge from 3.6)
https://hg.python.org/cpython/rev/c2f3b7c56dff