atomize option to bcftools norm removes FORMAT String tags with Number=A

[briankegerreis]

I have a VCF that includes a String FORMAT tag like this:

##FORMAT=<ID=GT,Number=1,Type=String,Description="Genotype">
##FORMAT=<ID=VF,Number=A,Type=Float,Description="Variant Frequency">
##FORMAT=<ID=STR,Number=A,Type=String,Description="String that goes missing">
#CHROM  POS     ID      REF     ALT     QUAL    FILTER  INFO    FORMAT  sample
chr1    150     .       TAT     CCC     100     PASS    AC=1;AN=2;TYPE=MNP     GT:VF:STR        0/1:0.1:DATA

When I do bcftools norm -a in.vcf, it removes the STR tag. Is this intended behavior, and am I being cavalier with the FORMAT field? TMI: The reason I'm not using INFO/STR is because this feeds into Illumina's hap.py engine, which adds INFO/TRUTH_FMT and INFO/QUERY_FMT for each variant to the output VCF.

Thanks!

[pd3]

Uh, string tags defined as Number=A,R,G are not implemented yet, sorry.

These are the places to plug these:

bcftools/abuf.c

Line 304 in 7a530fa

if ( type==BCF_HT_STR && len!=BCF_VL_FIXED && len!=BCF_VL_VAR ) return; // todo: Number=A,R,G for strings

bcftools/abuf.c

Line 452 in 7a530fa

if ( type==BCF_HT_STR && len!=BCF_VL_FIXED && len!=BCF_VL_VAR ) return; // todo: Number=A,R,G for strings

[pd3]

The support for Number=A,R string annotations has been now added.