Remove broken experimental package gambit

[mkoeppe]

It is not compatible with python 3.

https://github.com/mkoeppe/sage/runs/876312037

  [gambit-15.1.1.p0]   make[2]: Leaving directory '/sage/local/var/tmp/sage/build/gambit-15.1.1.p0/src'
  [gambit-15.1.1.p0]   Traceback (most recent call last):
  [gambit-15.1.1.p0]     File "setup.py", line 32, in <module>
  [gambit-15.1.1.p0]       m.Extension.__dict__ = m._Extension.__dict__
  [gambit-15.1.1.p0]   AttributeError: attribute '__dict__' of 'type' objects is not writable
  [gambit-15.1.1.p0]   Error installing Python API
  [gambit-15.1.1.p0]   

Another failure report: cygwin linking trouble - https://groups.google.com/g/sage-devel/c/HisDSCo7zlk/m/dAs5K5OrBwAJ

Upstream looks unattended. The last releases are from 2017.

https://github.com/gambitproject/gambit/issues

http://www.gambit-project.org/

gambitproject/gambit#242

https://repology.org/project/gambit-game-theory/versions

Previous tickets: #24840, #24012

CC: @dimpase @theref @kcrisman @nathanncohen @drvinceknight @tscrim @jhpalmieri

Component: packages: experimental

Issue created by migration from https://trac.sagemath.org/ticket/30156

[dimpase]

comment:3

I've also emailed Vince Knight, who might know more about Gambit status.

[dimpase]

comment:4

no reply - so let's just remove it.

[dimpase]

comment:5

however there is a lot of stuff in src/sage/games/ that needs it.

[mkoeppe]

comment:6

Let's give people a bit of time to react. This is not urgent. When we have collected a bit more information on Meta-ticket #29900 (Fix optional and experimental packages for Sage 9.2), we can perhaps also send a message to sage-users

[tscrim]

comment:7

This is our one current use of game theory in Sage. I think a better solution is to extract PR242 as a patch to Sage. (I forget why this was downgraded to experimental in the first place.) We might also consider forking this to include some of the updates, something I would consider doing, but I probably wouldn't be able to do anything with it long-term.

[kcrisman]

comment:8

I agree we should wait on this - perhaps contacting the maintainer of Gambit itself.

[dimpase]

comment:9

The primary author got back to me saying that's fine, he'll want to do a better Python API anyway (he considers the current one to be unstable), but not before finishing his term as a dean... That is yes, we ought to remove it.

[mkoeppe]

comment:10

remvove "it" = the python interface, or the whole package?

[dimpase]

comment:11

the whole package, I suppose

[kcrisman]

comment:12

What about the current doctests etc that rely on it? It would be unfortunate to remove them completely - but perhaps need to have some warnings on those files that the Gambit tests are currently broken and to use a previous version of Sage if you need that functionality for now.

[jhpalmieri]

comment:13

If no one can install the package, they won't easily be able to run the doctests, so they can stay. If you want to add some warnings where it says "Run sage -i gambit ...", that would probably be a good idea.

[mkoeppe]

comment:14

Are there parts of the Sage game theory code depend on the Gambit executables, or does everything depend on the broken Gambit Python library?

[fchapoton]

comment:16

Gambit 16 does not support Python 3. The current master branch does.

[kcrisman]

comment:17

Gambit 16 does not support Python 3. The current master branch does.

Interesting. Maybe it would be worth simply packaging as such? Unfortunately it wouldn't appear to currently be pip-installable.

[mkoeppe]

comment:18

Sage development has entered the release candidate phase for 9.3. Setting a new milestone for this ticket based on a cursory review of ticket status, priority, and last modification date.

[mkoeppe]

It's 2024 and gambit is back: http://www.gambit-project.org/, including pygambit, supporting Python >= 3.8 - https://pypi.org/project/pygambit/#history

[mkoeppe]

Closing in favor of:

• gambit: Re-create as a pip package pygambit (help wanted!) #37809