bcftools query; sample names missing in header line

[WimSpee]

Dear bcftools developers,

With older versions of bcftools query , sample names were used in the header line.
This makes it clear which sample the genotype attribute belongs to.

bcftools 1.9
bcftools query -H  -f '%CHROM\t%POS\t%REF\t%ALT[\t%GT]\n' input.vcf.gz | head

[1]CHROM	[2]POS	[3]REF	[4]ALT	[5]SAMPLE_1:GT	[6]SAMPLE_2:GT
chr02	125	T	C	0/0	1/1
chr02	126	C	A	0/0	1/1

With recent versions of bcftools, including v1.18, the sample names are not used anymore in the header line.

bcftools 1.18
bcftools query -H  -f '%CHROM\t%POS\t%REF\t%ALT[\t%GT]\n' input.vcf.gz | head

[1]CHROM	[2]POS	[3]REF	[4]ALT	[5]GT	[6]GT
chr02	125	T	C	0/0	1/1
chr02	126	C	A	0/0	1/1

It would be nice if the person receiving the output file of bcftools query can know which sample is which column.
Without having to go back to the input vcf, and without an extra step to add back the sample names.

Is it possible to (at least optionally) add the the sample names back again in the header line of the bcftools query output?

I know %SAMPLE can be used in the table, but that more than doubles the size of the ouptut file.
And sample meta data makes sense for me to be present once in the column header, not repetitive in all the records.

Thank you.

[pd3]

Thank you for the issue. This change was motivated by complexities arising from simultaneous usage of -H and filtering expression -i/-e. When both the -f formatting and filtering expression contains FORMAT fields, say -f '%POS [ %GT]\n' -i'GT="alt"', the samples that don't pass the expressions are normally not printed on output. This made the header incorrect.

In the commit a220acf this is addressed by adding the sample names as before whenever possible, for example when there is no danger of excluding samples, either because the filtering expression does not filter by sample or because the new option -F, -print-filtered was used.

I hope this helps.

[WimSpee]

Thank you for resolving the issue and the background information!