Default spacetype and sectortype implementations in type domain

[ogauthe]

PEPSKit.jl overload sectortype and spacetype for many objects (see QuantumKitHub/PEPSKit.jl@af3f140 or QuantumKitHub/PEPSKit.jl#218). Currently one has to repeat 4 lines of code for every struct in order to dispatch on type and use sectortype(spacetype).

This PR proposes to set defaults in TensorKit such that a single line

spacetype(::Type{MyType}) = ...

is sufficient to use sectortype(x::MyType) and spacetype(x::MyType).

I also defined a default for field as it is very similar. I kept the specializations for AbstractTensorMap to conserve easy access to the documentation, I do not know if there is a better way.

[lkdvos]

As we discussed, the docs can perhaps be migrated to a centralized location through:

@doc """
...
""" spacetype

Otherwise definitely looks like a nice quality of life addition to me, also relevant for MPSKit.

[ogauthe]

I generalized the doc. Also generalized sectortype to allow sectortype(SU2Irrep) == SU2Irrep.

[codecov]

Codecov Report

Attention: Patch coverage is 70.00000% with 3 lines in your changes missing coverage. Please review.

Project coverage is 82.66%. Comparing base (0c73bd2) to head (2f46fdd).
Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
src/spaces/vectorspaces.jl	70.00%	3 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##           master     #248   +/-   ##
=======================================
  Coverage   82.65%   82.66%           
=======================================
  Files          43       43           
  Lines        5593     5583   -10     
=======================================
- Hits         4623     4615    -8     
+ Misses        970      968    -2     

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[lkdvos]

I definitely think this is a QOL improvement, @Jutho how do you feel about this?

Thanks @ogauthe !

[Jutho]

Yes this is indeed much more convenient. The only small comment is that I would streamline:

sectortype(::Type{S}) where {S<:Sector} = S
spacetype(S::Type{<:ElementarySpace}) = S

so that they take the same form (both seem to do the job).

[ogauthe]

Yes this is indeed much more convenient. The only small comment is that I would streamline:

sectortype(::Type{S}) where {S<:Sector} = S
spacetype(S::Type{<:ElementarySpace}) = S

so that they take the same form (both seem to do the job).

I found there are subtle differences in dispatch behavior between these 2 versions (see https://docs.julialang.org/en/v1/manual/performance-tips/#Be-aware-of-when-Julia-avoids-specializing). I now impose specialization.

[lkdvos]

This is true, but often not relevant since these functions get inlined and constant folded most of the time, but I fully agree not relying on these implicit compiler details is annoying since it's hard to verify, so might as well be explicit about it