Refactor fq_default to use gr generics

[fredrik-johansson]

A start of reimplementing fq_default so that it's just syntactic sugar over gr generics. In particular, fq_default_ctx is now just an alias for gr_ctx. This PR also adds some missing functionality for nmod, fmpz_mod etc. gr contexts to make this work.

A lot of fq_default methods remain as if-statements; everything should be changed to gr method calls eventually, but this will require more work.

I've tried to make the documented interface completely backwards compatible, but there is possibly going to be some unanticipated breakage in Nemo, e.g. (there will certainly be breakage if some external code attempts to directly access the internals of the fq_default context object).

[fredrik-johansson]

Unfortunately fq_default is so poorly tested that it's hard to know if refactoring will break anything. Fingers crossed...

[thofma]

Bad timing, since we just changed everything to use fq_default[_ctx] and relied on it being a union over the various things.

Is the plan to do more incompatible changes or will these slow down at some point? I guess we just stick with 2.10 and wait till things settle down.

[albinahlback]

Is the plan to do more incompatible changes or will these slow down at some point? I guess we just stick with 2.10 and wait till things settle down.

I haven't talked to Fredrik about this, but my hope is that most of the breaking changes are between 3.0 and 3.1.

[fredrik-johansson]

Bad timing, since we just changed everything to use fq_default and relied on it being a union over the various things.

fq_default_t is still just a union over all the things. The only thing that has changed is the internal structure of the context object.

[thofma]

So can I still pass fq_default_ctx_t into a function expecting a fq_t if I know that the type is FQ_DEFAULT_FQ?

Edit: Sorry, I was not precise enough with my previous comment. I meant both the elements as well as the context object.

[fredrik-johansson]

So can I still pass fq_default_ctx_t into a function expecting a fq_t if I know that the type is FQ_DEFAULT_FQ?

You can pass a fq_default_t with type FQ_DEFAULT_FQ into a function expecting fq_t and vice versa.

fq_default_ctx_t and fq_ctx_t are not interchangeable, but that was already the case.

[thofma]

Is there a way to get to the fmpz_mod_ctx_t given a fq_default_ctx_t if I know that it has the right type?

[fredrik-johansson]

Fixed the 32-bit test failure which wasn't actually my fault (there was a missing init in the test code).

Seems like nothing tested broke in Nemo, so are there any objections to merging this?

[fredrik-johansson]

Is there a way to get to the fmpz_mod_ctx_t given a fq_default_ctx_t if I know that it has the right type?

Yes, gr_ctx_data_as_ptr(ctx) returns a pointer to the internal context object.

With the exception of nmod where the nmod_t is stored inline in the context object and gr_ctx_data_as_ptr(ctx) may be used.

Of course, we could add a less cryptic interface in the fq_default module.

[thofma]

The Nemo run has errors: https://github.com/flintlib/flint/actions/runs/7758309026/job/21159771932?pr=1765#step:7:3333

Given that it apparently does not catch all errors, the error might not be related to this PR.

[fredrik-johansson]

The Nemo run has errors: https://github.com/flintlib/flint/actions/runs/7758309026/job/21159771932?pr=1765#step:7:3333

Given that it apparently does not catch all errors, the error might not be related to this PR.

That error also shows up in the last test run on main.

It is surely a remnant of #1754 since it refers to _fmpz_mod_mat_clear_fn.

[albinahlback]

The Nemo run has errors: https://github.com/flintlib/flint/actions/runs/7758309026/job/21159771932?pr=1765#step:7:3333

Given that it apparently does not catch all errors, the error might not be related to this PR.

Hopefully I fixed it. I am rerunning the test.

[thofma]

I don't understand why it is still working, but if it works, it works.

[fredrik-johansson]

Added fq_default_ctx_inner to access the internal fq_*_ctx.

[thofma]

Merci