GitLab

Gather should and shouldvdj rules. Used for coverage.

Those sequences have colinear alignments on the genome.
They are not V(D)J recombinations, even if some kmer are shared with germlines.

depends on 1 - index_load, not the opposite

+1 missing for taking into account the number of potential kmers.
<= n for taking into account the (rare) case where all the affectations
would be set.

Just consists of what was previously done in operator>>

Once again this raises the need for browser tests.

(and use '-E' for old option on forced edges)

Avoid using pow in the loop. Call it once and then multiply or divide.

We have a 500x500 array (half of it is useless…) which takes about 2MB
but avoid lots of computations when computing the e-value.

This should not change very much the probability for small index loads.

Doc updated accordingly.

Should take into account the total number of sequences but we don't know it at first.
Maybe the value should be computed afterwards…