Cache canonical_matrix(::MonomialOrdering)

[lgoettgens]

Implements the proposal of #2498 (comment).

The speed change is shown in #2498 (comment).

Feel free to close this and propose an alternative to this @HechtiDerLachs.

[HechtiDerLachs]

I'd suggest not to go via attributes. This just entails another dictionary lookup and this is data structure is very low-level, i.e. it is called with high frequency.

[HechtiDerLachs]

And as @fieker said: Avoiding the lookup of a groebner basis in the first place is also possible and should be quicker.

But thanks for the very quick and constructive response!

[fingolfin]

This just hides over the actual problem, which is that canonical_matrix is being called far too often. See the discussion in #2099

[fingolfin]

Lest I be misunderstood: I still think this PR is useful and I am fine with merging it, but I removed the Resolves #2498 because I think it would be a bad idea to keep fixing these "leave issues" instead of attacking the root.

[simonbrandhorst]

I am glad for anything that helps my computations to actually terminate.

[codecov]

Codecov Report

Merging #2499 (f5cb257) into master (7b3acca) will decrease coverage by 0.01%.
The diff coverage is 75.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #2499      +/-   ##
==========================================
- Coverage   72.93%   72.93%   -0.01%     
==========================================
  Files         406      406              
  Lines       53901    53904       +3     
==========================================
+ Hits        39314    39315       +1     
- Misses      14587    14589       +2     

Impacted Files	Coverage Δ
src/Rings/orderings.jl	97.31% <75.00%> (-0.25%)	⬇️

... and 5 files with indirect coverage changes