Treat haploid VCFs as phased in vg autoindex

[jeizenga]

Changelog Entry

To be copied to the draft changelog by merger:

• vg autoindex recognizes haploid VCFs as phased

Description

Some people have been trying to use VCFs from minigraph-cactus in vg autoindex, which are often output as haploid VCFs (because the input haplotypes have not been grouped into diplotypes). Haploid VCFs are trivially phased, but the phase-checking code wasn't written to recognize this case.

Resolves #3886. Also mentioned in jonassibbesen/rpvg#43.

[adamnovak]

This is failing the same Giraffe tests on Mac and on Gitlab's Linux build:

#30 456.8 Test Summary Report
#30 456.8 -------------------
#30 456.8 t/50_vg_giraffe.t   (Wstat: 0 Tests: 45 Failed: 3)
#30 456.8   Failed tests:  43-45

[adamnovak]

I think I have a fix for this.

[adamnovak]

Those tests use autoindex with test/1mb1kgp/z.vcf.gz, which has no samples. We're looking at it and deciding it must be a phased haploid VCF because it doesn't have any non-haploid genotypes (because it has no genotypes, because it has no samples). And then we don't do something we need to do (probably generating haplotypes) for the tests to work.

[jeizenga]

Thanks!