Make Python 3 the default

[videlec]

Because 3 x 3 = 9 and we are at 9.0.beta3 these are three reasons to do the switch... see

https://groups.google.com/forum/#!topic/sage-devel/fmQbbSctZ-I

Depends on #28657

CC: @sagetrac-tmonteil

Component: misc

Author: Frédéric Chapoton

Branch: 715e85a

Reviewer: John Palmieri, Eric Gourgoulhon

Issue created by migration from https://trac.sagemath.org/ticket/28660

[fchapoton]

Branch: public/python3_switch

[fchapoton]

comment:1

No idea how to do that properly. Any volunteer to do the job ?

---

New commits:

c9646f9

py3 switch : some changes in doc only

[fchapoton]

Commit: c9646f9

[fchapoton]

comment:2

Some of the involved stuff:

• SAGE_PYTHON_VERSION

• SAGE_PYTHON3

• --with-python

• WITH_PYTHON

[kiwifb]

comment:3

Changing https://github.com/sagemath/sage-prod/blob/master/configure.ac#L395 should be a good start.

[sagetrac-git]

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

4357742

python3 switch : change in configure.ac

[sagetrac-git]

Changed commit from c9646f9 to 4357742

[fchapoton]

comment:5

ok, I did that. Anything else ?

[dimpase]

Dependencies: #25837

[dimpase]

comment:6

I think it depends on #25837, as sagenb doesn't work with Python 3.

[fchapoton]

comment:7

Did you test ? Sagenb is already automatically put aside (i.e. not built) with python3.

[dimpase]

comment:8

well, this is OK as an experimental feature, but we do install all the standard packages, normally speaking, and sagenb is standard. Besides, with Python 3 the message on https://github.com/vbraun/ExportSageNB/pull/16/commits
is unclear (it does not say that Sage must be built with Python 2 for sagenb to function)

[sagetrac-git]

Changed commit from 4357742 to ad6e2f9

[sagetrac-git]

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

ad6e2f9

py3 switch more doc changes

[fchapoton]

comment:10

vbraun/ExportSageNB#17

[jhpalmieri]

comment:11

If a prerequisite for building from source becomes "Python >= 3.6", then we need #28657 as a dependency.

[jhpalmieri]

Changed dependencies from #25837 to #25837 #28657

[jhpalmieri]

comment:12

In my opinion, #25837 is a hybrid ticket: one part should be simple — make sagenb optional, perhaps with a better error message, as in comment:10. The other part is to add to Jupyter some functionality which sagenb has but Jupyter does not. I am willing to sacrifice the second of these goals (for now) in order to make the switch to Python 3.

[egourgoulhon]

comment:13

Replying to @jhpalmieri:

In my opinion, #25837 is a hybrid ticket: one part should be simple — make sagenb optional, perhaps with a better error message, as in comment:10. The other part is to add to Jupyter some functionality which sagenb has but Jupyter does not. I am willing to sacrifice the second of these goals (for now) in order to make the switch to Python 3.

+1

[sagetrac-git]

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

1b42e43

py3 switch more details

[sagetrac-git]

Changed commit from ad6e2f9 to 1b42e43

[sagetrac-git]

Changed commit from 1b42e43 to 3efeee9

[sagetrac-git]

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

3efeee9

py3 swithc more details

[fchapoton]

comment:16

May I remove the dependency to #25837 ?

[jhpalmieri]

comment:17

Replying to @fchapoton:

May I remove the dependency to #25837 ?

I certainly think so.

[jhpalmieri]

comment:18

And #28657 doesn't need to be a dependency for this ticket, as long as it remains a blocker for the next release.

[jhpalmieri]

comment:19

I tried this and it works: after running make, I see

configure:7614: checking Python version to install
configure:7631: result: 3

[sagetrac-git]

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

715e85a

make sagenb optional

[fchapoton]

Changed dependencies from #25837 #28657 to #28657

[fchapoton]

comment:22

ok, we can turn sagenb to optional here if you prefer.

[fchapoton]

comment:24

Please edit and enhance the wiki page about the switch https://wiki.sagemath.org/Python3-Switch

[videlec]

comment:25

Replying to @fchapoton:

Please edit and enhance the wiki page about the switch https://wiki.sagemath.org/Python3-Switch

I would have thought that url link with space in them would better be avoided. You created a page Python3-compatible code. Am I allowed to fix that? I also open a page for end users who will suffer much more the transition than developers.

[fchapoton]

comment:26

something else ? We should try to move forward.

[fchapoton]

Author: Frédéric Chapoton

[jhpalmieri]

comment:27

As I said before, it works for me. Anyone else with feedback?

[egourgoulhon]

comment:28

Builds like a charm on Ubuntu 18.04 running on a i7-6700HQ + 16 GB RAM computer: having merged this ticket branch in Sage 9.0.beta4 sources (from a fresh git clone + pull develop), running
MAKE="make -j8" make yielded a fully working Python3 Sage. In particular sage -python gives

Python 3.7.3 (default, Nov  4 2019, 17:58:44) 
[GCC 7.4.0] on linux
Type "help", "copyright", "credits" or "license" for more information.

[jhpalmieri]

Reviewer: John Palmieri, Eric Gourgoulhon

[jhpalmieri]

comment:30

Let's proceed with this.

[mwageringel]

comment:31

This works for me as well on CentOS.

[kcrisman]

comment:32

In my opinion, #25837 is a hybrid ticket: one part should be simple — make sagenb optional, perhaps with a better error message, as in comment:10. The other part is to add to Jupyter some functionality which sagenb has but Jupyter does not. I am willing to sacrifice the second of these goals (for now) in order to make the switch to Python 3.

"For now" are dangerous words, though, in particular regarding matrix input and automatic_names. I'm certainly glad that both 3d viewers seem to work in Jupyter now, though I haven't yet built with py3 support.

[jhpalmieri]

comment:33

The switch to Python 3 needs to happen. To get these issues fixed, people need to work on them, and if no one is working on them (as seems to be the case), that isn't reason enough to delay the switch to Python 3.

[vbraun]

Changed branch from public/python3_switch to 715e85a

[jhpalmieri]

Changed commit from 715e85a to none

[jhpalmieri]

comment:35

See #28742 for a followup: after building with Python 2 and then upgrading to a version of Sage including this ticket, the version of Python changes to 3. This causes problems, so we should detect the version mismatch and either keep using Python 2 or abort the make process, telling the user what the issue is.

[jhpalmieri]

comment:36

Even though sagenb is now optional, it is still always installed because it is a dependency for building the documentation. Look at line 4 in the following, from build/make/deps:

# Building the documentation has many dependencies, because all
# documented modules are imported and because we use matplotlib to
# produce plots.
DOC_DEPENDENCIES = sagelib $(inst_sphinx) $(inst_sagenb) \
	| $(SAGERUNTIME) $(inst_maxima) $(inst_networkx) $(inst_scipy) $(inst_sympy) \
	$(inst_matplotlib) $(inst_pillow) $(inst_mathjax) $(inst_mpmath) \
	$(inst_ipykernel) $(inst_jupyter_client) $(inst_conway_polynomials) \
	$(inst_tachyon) $(inst_jmol) $(inst_thebe) $(inst_ipywidgets) $(inst_typing)

Should we create another followup to deal with this? I can do it, but it may take me a few days, because I have other higher priority things I have to do first.

[kiwifb]

comment:37

Definitely need a follow up. We disentangled that bit a while ago now. I can take care of it, if that's the only place we need to fix for this.

[kiwifb]

comment:38

It is now #28805.

[jhpalmieri]

comment:39

Thanks.

[jhpalmieri]

comment:40

Should we delete the file src/ext/doctest/python3-known-passing.txt and the corresponding parts of Makefile?

[dimpase]

comment:41

yes, sure, let's do this.

[jhpalmieri]

comment:42

I opened #28835.