Remove giac as dependency, and add sagemath_giac as optional dependency

[tobiasdiez]

Not sure if this change is completely right. It seems to me that there is still a giac expect interface in use, so perhaps giac should now be a runtime dependency instead of build dependency of sagelib? Or is that interface obsolete and should/can be removed?

📝 Checklist

• The title is concise and informative.
• The description explains in detail what this PR is about.
• I have linked a relevant issue or discussion.
• I have created tests covering the changes.
• I have updated the documentation and checked the documentation preview.

⌛ Dependencies

[github-actions]

Documentation preview for this PR (built with commit 67b184f; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

[user202729]

Or is that interface obsolete and should/can be removed?

Certainly I don't need it, but maybe someone who just-inconveniently-happen to not be able to build sage with giac linked…

Of course that adds to maintenance cost, so I'd be okay with removing it. I think usual procedure requires a one-year deprecation period.

[dimpase]

deprecation period only applies to non-bugs. if there is a bug fixed along the way, it can get in right away

[tobiasdiez]

@orlitzky what is the status of the giac expect interface? Can this be removed/deprecated?

[orlitzky]

The pexpect interface is used by some other parts of sagelib, but only if you explicitly ask for it. Judging by my own comments, at least the following user-facing bits will still use pexpect rather than the faster library interface:

• sum(..., algorithm='giac')
• laplace(..., algorithm='giac')
• inverse_laplace(..., algorithm='giac')
• simplify(..., algorithm='giac')

But in all of these cases you have to explicitly ask for giac, and if you don't, giac isn't a dependency.

The answer to the question you asked is, no, it can't be completely dropped yet because those remaining functions need to be ported to sage.libs.giac (i.e. sagemath-giac) first.

But to answer the question you didn't ask, I think you could drop giac as a dependency of sagelib regardless. If someone wants to use algorithm='giac' with any of those functions they can just install giac and it will start working immediately. You'll get an error that suggests this:

TypeError: unable to start giac because the command 'giac --sage' failed: The command was not found or was not executable: giac.


In order to use the Giac interface you need to have Giac installed
and have a program called "giac" in your PATH. You need a giac version
supporting "giac --sage" ( roughly after 0.9.1 of march 2011). Some giac
instructions  and the help's langage depend of you LANG variable. To obtain
inline help for  giac commands, you also need to have the program "cas_help"
in your PATH.

[dimpase]

on the sage-distro level, should we relegate giac to the role of a dummy package, as we did for R a while ago? Anyhow, this looks good, and the "required" test-long fails due to some "no space left on device" error, not related to the changes here.

[user202729]

The error message looks… not too actionable? Doesn't help that the online-generated pages for spkg documentation isn't too descriptive on how to actually install the package either, but this particular one doesn't even have sage -i.

[dimpase]

sage -i might not even be available - e.g. with a meson build

[user202729]

sage -i might not even be available - e.g. with a meson build

That's another issue, no?

While I use meson myself, I feel it's better to at least be helpful in one circumstance than to be helpful in no circumstance. To avoid maintenance overhead we may just call the feature.requires() framework, then when the suggestion message sage -i is finally changed it will automatically improve.