Test for some special cases in DecomposeCyclotomicAlgebra

[olexandr-konovalov]

Attempts to find examples with increased coverage. Trying suggestions by @drallenherman from the discussion in #76:

# Split simple components of dimension 16^2
gap> QG:=GroupRing(Rationals,SmallGroup(1920,136971));;
gap> W:=WedderburnDecompositionInfo(QG);;
gap> Length(W);
93
gap> A:=W[Length(W)];
[ 16, GaussianRationals ]
gap> LocalIndicesOfCyclotomicAlgebra(A);
[  ]
gap> QG:=GroupRing(Rationals,SmallGroup(1920,204811));;
gap> W:=WedderburnDecompositionInfo(QG);;
gap> Length(W);
105
gap> A:=W[Length(W)];
[ 16, NF(20,[ 1, 3, 7, 9 ]) ]
gap> LocalIndicesOfCyclotomicAlgebra(A);
[  ]

[codecov[bot]]

Codecov Report

Merging #81 into divalg will increase coverage by 0.71%. The diff coverage is n/a.

@@            Coverage Diff             @@
##           divalg      #81      +/-   ##
==========================================
+ Coverage   81.59%   82.30%   +0.71%     
==========================================
  Files          11       11              
  Lines        5193     5193              
==========================================
+ Hits         4237     4274      +37     
+ Misses        956      919      -37     

Impacted Files	Coverage Δ
lib/div-alg.gi	89.53% <0.00%> (+2.21%)	:arrow_up:

[olexandr-konovalov]

So, new examples suggested by @drallenherman do not cover more lines, but they take ~20s in total, so if these cases are interesting and may be used to check performance or memory regressions, or may check some further modifications of code, treating special cases, we still can add them. Any thoughts?

[drallenherman]

So, new examples suggested by @drallenherman do not cover more lines, but they take ~20s in total, so if these cases are interesting and may be used to check performance or memory regressions, or may check some further modifications of code, treating special cases, we still can add them. Any thoughts?

I think we should find examples that increase coverage. I'll look at the specific lines not being covered and try to come up with better ones.

[olexandr-konovalov]

@drallenherman ok, then please post examples here, and I will update and force-push this PR to see if they have any effect.

[drallenherman]

So, new examples suggested by @drallenherman do not cover more lines, but they take ~20s in total, so if these cases are interesting and may be used to check performance or memory regressions, or may check some further modifications of code, treating special cases, we still can add them. Any thoughts?

I think we should find examples that increase coverage. I'll look at the specific lines not being covered and try to come up with better ones.

Quite a few lines of div-alg that lack code coverage are in DecomposeCyclotomicAlgebra. If we apply that to the last component of QG for SG(240,96) it will improve coverage. Especially after branch #80 gets merged.

[olexandr-konovalov]

Nice @drallenherman !

# Some special cases in DecomposeCyclotomicAlgebra (PR #81)
gap> QG:=GroupRing(Rationals,SmallGroup(240,96));;
gap> W:=WedderburnDecompositionInfo(QG);;
gap> Length(W);
18
gap> A:=W[Length(W)];
[ 1, Rationals, 30, [ [ 2, 11, 0 ], [ 4, 7, 0 ] ], [ [ 15 ] ] ]
gap> DecomposeCyclotomicAlgebra(A);
[ [ Rationals, CF(3), [ 1 ] ], [ Rationals, CF(5), [ 0 ] ] ]

hits 37 more lines! Going to merge it.

[drallenherman]

So, new examples suggested by @drallenherman do not cover more lines, but they take ~20s in total, so if these cases are interesting and may be used to check performance or memory regressions, or may check some further modifications of code, treating special cases, we still can add them. Any thoughts?

I think we should find examples that increase coverage. I'll look at the specific lines not being covered and try to come up with better ones.

Quite a few lines of div-alg that lack code coverage are in DecomposeCyclotomicAlgebra. If we apply that to the last component of QG for SG(240,96) it will improve coverage. Especially after branch #80 gets merged.

Nice @drallenherman !

# Some special cases in DecomposeCyclotomicAlgebra (PR #81)
gap> QG:=GroupRing(Rationals,SmallGroup(240,96));;
gap> W:=WedderburnDecompositionInfo(QG);;
gap> Length(W);
18
gap> A:=W[Length(W)];
[ 1, Rationals, 30, [ [ 2, 11, 0 ], [ 4, 7, 0 ] ], [ [ 15 ] ] ]
gap> DecomposeCyclotomicAlgebra(A);
[ [ Rationals, CF(3), [ 1 ] ], [ Rationals, CF(5), [ 0 ] ] ]

hits 37 more lines! Going to merge it.

All right. We need to merge #80 next to bring in the improved version of DecomposeCA. When you do that you will need to change the [0] at the end of this one to a [9]. It's likely coverage will go down slightly it's hard to cover all of the cases in the new DecomposeCA.

[olexandr-konovalov]

We need to merge #80 next to bring in the improved version of DecomposeCA. When you do that you will need to change the [0] at the end of this one to a [9].

No - you will need to rebase your #80 (what you need to do anyway) and update the test there.