Fix uses of Val inside Hecke

[lgoettgens]

This is similar to Nemocas/AbstractAlgebra.jl#1664.

Val{T} is a type and Val(T) its singleton object. This is the way it is described in the julia docs (https://docs.julialang.org/en/v1/manual/types/#%22Value-types%22).

I went through all the uses and corrected them.

For all exported functions where the parameter occurred in a docstring, I already added deprecations. If you want more deprecations, please let me know.

[lgoettgens]

I'll continue to add deprecations for everything that fails in the downstream tests

[lgoettgens]

Somehow the "fake" doctests in Oscar emit deprecation warnings that make them fail. Do you have an idea why depwarns are enabled there?

[thofma]

Hm, I don't remember how those fake doctests are run. @benlorenz do you remember from the top of your head?

[benlorenz]

These are run in a dummy-module with a dummy function:
https://github.com/oscar-system/Oscar.jl/blob/183cbe6a5e9c3dbdb7e851c00d083dbf3a059343/experimental/OrthogonalDiscriminants/test/data.jl#L79-L149

depwarn=yes is the julia-runtest default, we set it to depwarn=error for the extra tests.

[thofma]

Maybe add the old method back in @lgoettgens?

[benlorenz]

Deprecating exported functionality is considered breaking and should be done with a minor version bump instead of a patch-level bump.

[lgoettgens]

Deprecating exported functionality is considered breaking and should be done with a minor version bump instead of a patch-level bump.

This is just stated for versions after 1.0.0 (ref https://semver.org/spec/v2.0.0.html#spec-item-7), in the 0.y.z phase everything may change at any time (https://semver.org/spec/v2.0.0.html#spec-item-4). (And in this case here force_coerce_cyclo isn't even exported.)

To reduce friction here, let me add the deprecations as "real functions" (i.e. without a @deprecate) and we change this in the next minor release instead.

[thofma]

Julia does not follow SemVer for 0.*: https://pkgdocs.julialang.org/v1/compatibility/#compat-pre-1.0.

[benlorenz]

(Almost) everything that is exported by Hecke is also exported by Oscar and thus breaking for Oscar 1.0.x.

[lgoettgens]

(Almost) everything that is exported by Hecke is also exported by Oscar and thus breaking for Oscar 1.0.x.

Thanks for reminding me. However, this is no longer an issue due to my last changes. The deprecations will only be introduced in Hecke 0.31.0, which will then only be made compatible on the Oscar master branch, and not the release-1.0.

[benlorenz]

Thanks!

In this case it seems it is technically a bug in Oscar since it relies on the non-exported Hecke.force_coerce_cyclo.

But with all the re-exports we should be very careful with such things and moving to a new minor release soon seems like a good idea. We can still create a branch for 0.30 and do bugfixes there if necessary.