Replace many @aliases by @deprecate_binding

[fingolfin]

Also drop two obsolete redundant uses of @alias. Also includes the fixes from PR #1556.

Companion to Nemocas/AbstractAlgebra.jl#1469

[thofma]

No objection to the change, but I am curious. Is this official API? Or could this be gone in some minor version?

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Comparison is base (8dff2d9) 84.17% compared to head (1df6741) 84.16%.
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1557      +/-   ##
==========================================
- Coverage   84.17%   84.16%   -0.01%     
==========================================
  Files          94       94              
  Lines       36658    36658              
==========================================
- Hits        30857    30854       -3     
- Misses       5801     5804       +3     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[fingolfin]

It is not official API, but it has been around since at least Julia 1.0, and there is an open issue requesting that it be made official.

All in all I think it is very unlikely that this is removed in a patch version of Julia. It might be removed in 1.11 or 1.12 or whatever (though I again think it is unlikely), but I hope that by then we'll have moved these one step further (either by switching them to use plain @deprecate, or by removing them outright).

[fingolfin]

Oh, and: this breaks the CI tests of our downstream packages as they test with depwarn=error. So I guess we should consider this a breaking change... In any case, we probably should first fix all the uses of deprecated API names in our various packages before we merge this...

[lgoettgens]

Deprecating stuff is not breaking!
But yeah, we should fix our packages beforehand

[fingolfin]

(Old comment follows which I wrote a week or so ago but apparently forgot to submit sigh):

The failures are due to the use of CyclotomicField; together with PR #1576, this PR now removes this if deprecations are set to error.

Indeed we use CyclotomicField as a kind of "fake type" in Oscar (and I now recall that @fieker pointed this out to me previously). Specifically:

experimental/GModule/GModule.jl:57:julia> C = gmodule(CyclotomicField, C);
experimental/GModule/GModule.jl:283:  return gmodule(AnticNumberField, gmodule(CyclotomicField, M))
experimental/GModule/GModule.jl:290:    a = gmodule(CyclotomicField, m)
experimental/GModule/GModule.jl:338:function irreducible_modules(::typeof(CyclotomicField), G::Oscar.GAPGroup)
experimental/GModule/GModule.jl:340:  return [gmodule(CyclotomicField, m) for m in z]
experimental/GModule/GModule.jl:345:  z = irreducible_modules(CyclotomicField, G)
experimental/GModule/GModule.jl:354:function gmodule(::typeof(CyclotomicField), C::GModule)
experimental/GModule/GModule.jl:1261:  C1 = gmodule(CyclotomicField, C)
experimental/GModule/GModule.jl:1262:  D1 = gmodule(CyclotomicField, D)
experimental/GModule/GModule.jl:1408:function Oscar.gmodule(T::Union{Type{CyclotomicField}, Type{AnticNumberField}}, chi::Oscar.GAPGroupClassFunction)
experimental/GModule/test/runtests.jl:111:  S = gmodule(CyclotomicField, R)
src/Rings/AbelianClosure.jl:426:function minimize(::typeof(CyclotomicField), a::AbstractArray{nf_elem})
src/Rings/AbelianClosure.jl:454:function minimize(::typeof(CyclotomicField), a::MatElem{nf_elem})
src/Rings/AbelianClosure.jl:455:  return matrix(minimize(CyclotomicField, a.entries))
src/Rings/AbelianClosure.jl:458:function minimize(::typeof(CyclotomicField), a::nf_elem)
src/Rings/AbelianClosure.jl:459:  return minimize(CyclotomicField, [a])[1]
src/Rings/AbelianClosure.jl:462:conductor(a::nf_elem) = conductor(parent(minimize(CyclotomicField, a)))
src/Rings/AbelianClosure.jl:684:hash(a::QQAbElem, h::UInt) = hash(minimize(CyclotomicField, a.data), h)

Note that even in OSCAR 0.13.0, this is just a generic function:

julia> CyclotomicField
cyclotomic_field (generic function with 9 methods)

How do we proceed? We can of course just keep this alias. But I wonder if this may not lead to other kinds of confusion down the line (after all, CyclotomicField is used here in a way that suggests it is a type, but it really isn't).

We could of course change Oscar to use cyclotomic_field. Or perhaps one could use QQab instead? I.e. minimize(QQab, a.entries) and

[thofma]

It could be replaced for minimize (these should be internal anyway), but for irreducible_gmodules(...) it would be wrong to replace it by QQab and irreducible_modules(cyclotomic_field, G) looks a bit odd.

[lgoettgens]

But I don't think that the name CyclotomicField needs to be provided by Nemo for some Oscar functionality. If Nemo does not use it, it shouldn't be there nor exported.

[fingolfin]

Discussed it with @fieker: we'll remove CyclotomicField here and re-introduce it in Oscar, as suggested by @lgoettgens .