py3: work on unicode in lazy_import

[fchapoton]

solving the first problem met in #22945

CC: @jdemeyer @tscrim

Component: python3

Keywords: unicode

Reviewer: Frédéric Chapoton

Issue created by migration from https://trac.sagemath.org/ticket/22946

[fchapoton]

Commit: 48b2364

[fchapoton]

New commits:

48b2364

py3: work on unicode in lazy_import

[fchapoton]

Branch: u/chapoton/22946

[fchapoton]

Changed keywords from none to unicode

[sagetrac-git]

Changed commit from 48b2364 to b171480

[sagetrac-git]

Branch pushed to git repo; I updated commit sha1. This was a forced push. New commits:

b171480

py3: unicode_literal in lazy_import

[fchapoton]

comment:5

bot is essentially green

[tscrim]

comment:6

Could you explain how importing unicode_literal allows you to use unicode in Python3? I would think you should use unicode_literal.

[tscrim]

comment:7

Also, is the spacing in the slices PEP8?

[fchapoton]

comment:8

hmm, well. I have doubts, I must say.

Question 1: This is a pyx file. I nevertheless wonder whether cython understands "unicode" and "str" correctly in py2/py3 contexts

Question 2: I am sure for the +, not so sure about the :

I am only starting to try and touch the unicode question.. IMHO, this will be a large scale mess, maybe only marginally easier than the "cmp" mountain.

[tscrim]

comment:9

Question 1 is something Jeroen might be able to answer, I cannot. For Question 2, I would almost say not putting the spaces around the + as it looks somewhat strange. I more strongly disagree with the space after the : in a slice.

Hopefully the unicode problem won't be so pervasive as we don't use unicode in too many places.

[sagetrac-git]

Changed commit from b171480 to 0554525

[sagetrac-git]

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

015eebb	Merge branch 'u/chapoton/22946' in 8.0.b5
0554525	trac 22946 undo pep8 change

[fchapoton]

comment:11

Jeroen, any comment ?

[fchapoton]

comment:12

ping ?

[tscrim]

comment:13

Or could the import of unicode_literals be a change to unicode like print_function is to print?

[tscrim]

comment:14

Okay, so I've looked a little more into unicode_literals, and it is a global flag that makes all strings into unicode (i.e., is like doing u'foo'). I'm slightly worried about potential side effects (well, not on my system since I don't have unicode file paths), so is it strictly necessary to include it?

Then again, this is Cython code, so Python rules need not apply.

[fchapoton]

comment:15

so, is there still some objection to the current state of the branch ? patchbot is green..

[tscrim]

comment:16

I am okay with the branch fundamentally, but I am still a little nervous. So this is a positive review, but I don't want to merge it in such a late beta (maybe the next will be an rc). Instead I would want this merged in 8.1.beta0.

[jdemeyer]

comment:17

To answer the questions above:

1. from future import unicode_literals is like changing all strings in the sources from "foo" to u"foo".

2. Cython does understand str, bytes, unicode, basestring in all Python versions.

Note that this conflicts with #22755.

[sagetrac-git]

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

8b16e02

Merge branch 'u/chapoton/22946' in 8.0.b12

[sagetrac-git]

Changed commit from 0554525 to 8b16e02

[fchapoton]

comment:20

after merging with 8.0.b12, the branch is essentially empty. So I propose to set this to duplicate.

[jdemeyer]

Changed author from Frédéric Chapoton to none

[jdemeyer]

Reviewer: Frédéric Chapoton

[fchapoton]

Changed branch from u/chapoton/22946 to none

[fchapoton]

Changed commit from 8b16e02 to none