Remove redundant Python 2 code #10865

hugovk · 2017-10-21T22:19:57Z

IPython 6 dropped support for Python 2.7, 3.0, 3.1 and 3.2 (eg. #9145, #9900, #10131).

This removes a lot of old Python 2 code still in the codebase.

TODO:

takluyver · 2017-10-23T09:51:02Z

I haven't checked in detail, but I'm not 100% comfortable removing all those cast_unicode calls. We have a cast_unicode_py2 function for when we know it's redundant on Python 3, so if we're using cast_unicode instead, there's a chance that that is necessary even on Python 3.

Also, as with your PR to my fork, can we ask you to do it with your editor configured to not strip whitespace? It can cause merge conflicts and problems with git blame.

hugovk · 2017-10-23T10:28:58Z

Sure, coming up.

…org/

hugovk · 2017-10-23T12:21:57Z

@takluyver This PR updated.

takluyver · 2017-10-24T10:07:18Z

.editorconfig

@@ -0,0 +1,19 @@
+# Top-most EditorConfig file


@Carreau @gnestor thoughts on adding this file? It seems reasonable to me, but I know people sometimes take issue with ever more config files in the repo root directory.

I'm ok adding it, but I don't know of a lot of people using that. It may likely just rot as the tox.ini file which is never used and still test on Python 27 as shown further in this PR.

takluyver · 2017-10-24T10:10:59Z

IPython/utils/_process_cli.py

@@ -12,23 +12,21 @@
 """

 # Import cli libraries:
-import clr


I think this might be a special thing that needs to be imported first for IronPython to work. Best to leave this module alone, I think.

takluyver · 2017-10-24T10:11:23Z

IPython/utils/_process_cli.py

 import System

 # Import Python libraries:
 import os

 # Import IPython libraries:
 from IPython.utils import py3compat
-from ._process_common import arg_split


This is there to be re-exposed by the module. Again, let's leave this as is.

takluyver · 2017-10-24T10:12:45Z

IPython/utils/py3compat.py

-def cast_bytes(s, encoding=None):
-    if not isinstance(s, bytes):
-        return encode(s, encoding)
-    return s


I'm a bit wary of removing things from py3compat in case other modules like ipykernel rely on them.

Sure, I'll make these changes. Seeing as Python 3 is now the only supported version, how about somehow marking py3compat as deprecated?

We can mark it as deprecated once we are not using it anymore. If we mark it as deprecated and still use it here and there people complain we don't even follow our own deprecations.

takluyver · 2017-10-24T10:14:12Z

Thanks, this mostly looks good now. A few comments inline.

takluyver · 2017-10-24T13:22:00Z

Yeah, I think it could have a 'deprecated' note in the docstring now, and then in the future when we think we've got rid of all uses of it, add a warning on import.

For now, let's leave all the functions exposed by py3compat that were there before (not just the one my previous comment was on the line of) :-)

Carreau · 2017-10-24T16:43:55Z

First thing I want to say is I'm worried about finishing up cleaning py2 compat code; this has been painful before for automatic backport of bug fixes, and lead to subtle bugs. We've stop removing things on purpose some time ago.

I'll look at it in more details.

Carreau · 2017-10-24T16:46:55Z

.gitignore

@@ -21,5 +21,6 @@ __pycache__
 .DS_Store
 \#*#
 .#*
+.cache


Should that also be in your system global .gitignore if you feel the need to add it ?

Perhaps, but it's in the common Python one --https://github.com/github/gitignore/blob/master/Python.gitignore#L43 -- and ignored in some other Python projects I just checked. And it showed up after running the tests for this project (iptest --coverage xml ), so it can show up for others too.

Yeah, I'm not sure exactly what creates it, but I see it quite often when running tests, so I'm fine with adding it to gitignore.

Carreau · 2017-10-24T16:50:51Z

IPython/utils/py3compat.py

@@ -1,22 +1,17 @@
 # coding: utf-8


if we remove the compatibility layer python2/python3, should this file still exists ? The way I see it either this file is / or is not. If there are function with use often they probably should be in utils. @takluyver thoughts ?

Thoughts: leave it in for now, except for the if/else branch that only runs on Py2. Add a deprecation note in the docstring. Then, over a succession of future releases:

Eliminate all remaining usage of it in IPython

Add a deprecation warning when it's loaded

(Eventually) get rid of it

It doesn't cost much to leave it sitting around unused for a while, and other code might be importing it. We did document it in earlier versions, though the current version tells people not to rely on it.

Carreau · 2017-10-24T16:51:56Z

IPython/utils/tokenize2.py

-    StringPrefix = r'(?:[bB][rR]?|[rR][bB]?|[uU])?'
-else:
-    StringPrefix = r'(?:[bB]?[rR]?)?'
+StringPrefix = r'(?:[bB][rR]?|[rR][bB]?|[uU])?'


if I remember correctly this file is base on Python 3.2 stdlib, does it make sens to check whether the bugs have been fixed upstream on 3.4+ instead of having it diverge more ?

I suspect this change is reverting a divergence that was previously introduced. Ultimately it would be good to get rid of it, but that's a more complex job for another day.

Carreau · 2017-10-24T16:52:14Z

setupbase.py

-        exec(compile(open(fname).read(), fname, "exec"), globs, locs)
+def execfile(fname, globs, locs=None):
+    locs = locs or globs
+    exec(compile(open(fname).read(), fname, "exec"), globs, locs)


Carreau · 2017-10-24T16:52:40Z

tox.ini

@@ -19,7 +19,7 @@ toxworkdir = /tmp/tox_ipython
 ; Other IPython/testing dependencies should be in setup.py, not here
 deps =
  pypy: https://bitbucket.org/pypy/numpy/get/master.zip
-  py{36,35,34,33,27}: matplotlib
+  py{36,35,34,33}: matplotlib


Considering if we should just remove the .tox file as it is never used.

Carreau · 2017-10-24T16:57:36Z

I don't have much to say, we'll need to do that at some point (and that will simplify the code. My guts guts just make me worried that there will be subtle bugs, and that it may create conflicts later.

I've been willing to start a 7.x branch at some point; and would be more comfortable to get that into a major release than a minor. @takluyver thoughts ? You've been more active than me on IPython in the last few weeks.

takluyver · 2017-10-25T10:21:05Z

I think that if we make this a bit more conservative, it's fine for a 6.x release. I already had a look over it and I'm happy with most of the changes. The main thing is that I'd preserve the py3compat API at least through the 6.x series, and probably longer.

hugovk · 2017-10-25T13:27:55Z

Updated to:

retain all of py3compat.py as is, except for the if/else that only runs on Python 2
update py3compat.py docstring to note it is deprecated and will be removed in a future version
update pycompat.rst to note it will be removed in a future version
delete unused tox.ini

takluyver · 2017-10-25T20:07:05Z

This looks OK to me as it stands. I'm sure someone will complain about removing tox.ini, but it's evidently not in use, or people would have already removed Python 2 from the list. So I'd say that not having it is better than having it inaccurate.

takluyver · 2017-10-31T16:56:57Z

Thanks @hugovk !

hugovk force-pushed the rm-python2 branch from 505b60a to 43340b7 Compare October 21, 2017 22:39

Add EditorConfig to preserve trailing whitespace http://editorconfig.…

3de8810

…org/

hugovk force-pushed the rm-python2 branch 3 times, most recently from a7e65e3 to f05ac36 Compare October 23, 2017 12:09

takluyver added this to the 6.3 milestone Oct 24, 2017

takluyver reviewed Oct 24, 2017

View reviewed changes

Carreau reviewed Oct 24, 2017

View reviewed changes

Remove redundant Python 2 code

77b36a9

hugovk force-pushed the rm-python2 branch from f633d27 to 77b36a9 Compare October 25, 2017 13:14

hugovk added 2 commits October 25, 2017 16:20

Note py3compat is deprecated and will be removed in the future

9db6129

Delete unused tox.ini

fdcb05d

takluyver merged commit 5b2b7dd into ipython:master Oct 31, 2017

hugovk deleted the rm-python2 branch October 31, 2017 17:32

Uh oh!

Remove redundant Python 2 code #10865

Remove redundant Python 2 code #10865

Uh oh!

Conversation

hugovk commented Oct 21, 2017 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

takluyver commented Oct 23, 2017

Uh oh!

hugovk commented Oct 23, 2017

Uh oh!

hugovk commented Oct 23, 2017

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

takluyver commented Oct 24, 2017

Uh oh!

takluyver commented Oct 24, 2017

Uh oh!

Carreau commented Oct 24, 2017

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Carreau commented Oct 24, 2017

Uh oh!

takluyver commented Oct 25, 2017

Uh oh!

hugovk commented Oct 25, 2017

Uh oh!

takluyver commented Oct 25, 2017

Uh oh!

takluyver commented Oct 31, 2017

Uh oh!

Uh oh!

hugovk commented Oct 21, 2017 •

edited

Loading