Python 3 preparation: Change names of some function attributes

[wluebbe]

Only the modern syntax like f.__doc__ is accepted by Python 3.

Changes according to lib2to3/fixes/fix_funcattrs.py:

f.func_x -> f.__x__
for 
('func_closure' | 'func_doc' | 'func_globals' | 'func_name' | 
'func_defaults' | 'func_code' | 'func_dict')

This ticket is tracked as a dependency of meta-ticket ticket:15980.

CC: @fchapoton

Component: distribution

Author: Wilfried Luebbe, R. Andrew Ohana

Branch/Commit: c03d421

Reviewer: Frédéric Chapoton, R. Andrew Ohana

Issue created by migration from https://trac.sagemath.org/ticket/15983

[wluebbe]

Commit: 3b0d1f7

[wluebbe]

New commits:

3b0d1f7

Change names of some function attributes

[wluebbe]

Branch: u/wluebbe/ticket/15983

[fchapoton]

comment:3

maybe you also have to correct the names inside the 'hasattr' ?

[wluebbe]

comment:4

Yes, seems suspicious. I will look into it.

[wluebbe]

comment:5

From the Python docs (http://docs.python.org/2/reference/datamodel.html):

//Changed in version 2.6: The double-underscore attributes !closure!, !code!, !defaults!, and !globals! were introduced as aliases for the corresponding func_* attributes for forwards compatibility with Python 3.//

!dict!, !doc! and !name! were already available earlier.

Looking at the code with hasattr it should probably be manually refactored, keeping in mind that the __XX__ are always available as alias for the func_XX.

Would you agree?

[sagetrac-git]

Changed commit from 3b0d1f7 to a97c3ad

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. New commits:

a97c3ad

added manual changes that were not found by 2to3 tool (in .pyx and hasattr)

[wluebbe]

comment:8

The modules with hasattr() might still benefit from some refactoring, as the code could now be a bit simplified.

But I leave this for later ...

[fchapoton]

comment:9

looks good, I have checked that every instance was found and corrected. But one needs to make sure that nothing is broken and that all tests pass..

[wluebbe]

comment:10

doctests fail in

• src/sage/sets/set_from_iterator.py (line 438)
• src/sage/misc/cachefunc.pyx (line 315 and 318)

[ohanar]

comment:11

Actually this looks like a bug fix -- the previous version wouldn't work on Python functions/methods defined in Cython code (well they really just duck-type Python functions), since Cython doesn't include the func_* aliases.

The doctests just need updating to reflect the bug fix.

[ohanar]

comment:12

Actually, nevermind. This is supposed to be stripped, but there are now two instances of this line information, and only one is getting stripped. I should have a patch soon to fix it.

It seems that the 'func_doc' was being used to separate cython functions from python functions.

[ohanar]

comment:13

Ok, I believe this should be a fix for the issue, for now.

---

New commits:

9357384

remove hack that used func_doc to distinguish cython and python functions

[ohanar]

Changed commit from a97c3ad to 9357384

[ohanar]

Changed branch from u/wluebbe/ticket/15983 to u/ohanar/remove_func_star

[wluebbe]

Changed commit from 9357384 to c03d421

[wluebbe]

Changed branch from u/ohanar/remove_func_star to u/wluebbe/ticket/15983

[wluebbe]

comment:14

I did a rebase on develop to test with a fresh 6.2.beta6.

MAKE='make -j4' make
./sage -tp 4 --all --long >logs/sage-ptestlong-develop
All tests passed!
----------------------------------------------------------------------
Total time for all tests: 2379.2 seconds
#####
git fetch trac u/ohanar/remove_func_star:u/ohanar/remove_func_star
git co u/ohanar/remove_func_star
git co -b u/wluebbe/ticket/15983
git rebase develop
#####
./sage -b
./sage -tp 4 --all --long >logs/sage-ptestlong-15983
All tests passed!

My questions (as still being pretty new to Sage):

• Is it OK to build and test as above?
• When is it necessary to do make clean; make lib-clean; make instead of ./sage -b?

---

New commits:

2908c63	Change names of some function attributes
108eccc	added manual changes that were not found by 2to3 tool (in .pyx and hasattr)
c03d421	remove hack that used func_doc to distinguish cython and python functions

[vbraun]

comment:15

author/reviewer names please

[wluebbe]

Reviewer: Frédéric Chapoton, R. Andrew Ohana

[wluebbe]

Author: Wilfried Luebbe, R. Andrew Ohana

[ohanar]

comment:17

Replying to @wluebbe:

I did a rebase on develop to test with a fresh 6.2.beta6.

In general, you should probably merge in the latest beta rather than rebase (this is a change from how sage use to do things to match standard git practices). That said, extra merges are unnecessary, so you should try to avoid making merges unless there is a merge conflict or an issue is only popping up on a more recent version of sage.

My questions (as still being pretty new to Sage):

• Is it OK to build and test as above?

Yes, although part of a review is to look at the changes in the code and make sure you agree with them.

• When is it necessary to do make clean; make lib-clean; make instead of ./sage -b?

Use make if there have been upgraded packages in the distribution (so between most betas/releases), you can use ./sage -b if the only changes between the last time you built have been to the sage library (useful if you are working on a feature branch that only touches the sage library).

[wluebbe]

comment:18

Thanks for the explanation!

Next time I will try a merge when there is a merge conflict.

[vbraun]

Changed branch from u/wluebbe/ticket/15983 to c03d421