This may not be enough for some applications (e.g. consider very small allele frequencies). Could we add an option to VCFEncoder that formats such values as %.9g (to keep full 32-bit float precision, which would nicely match the number format assumed by the VCF/BCL standard for other fields like QUAL)
I'm happy to prepare a PR, but I realize this touches some very mature parts of the codebase, so I'd like to get some feedback first.
Expected behaviour
VCFEncoder should be configurable to output annotation values with floating point precision.
Actual behaviour
VCFEncoder can't be configured in this way right now.
VCFEncoder.formatVCFDoublerestricts the precision of floating point values to three digits after the comma:https://github.com/samtools/htsjdk/blob/8f82871c167cca7f3f9921a251469f96dd1a2979/src/main/java/htsjdk/variant/vcf/VCFEncoder.java#L261-L279
This may not be enough for some applications (e.g. consider very small allele frequencies). Could we add an option to
VCFEncoderthat formats such values as%.9g(to keep full32-bit floatprecision, which would nicely match the number format assumed by the VCF/BCL standard for other fields likeQUAL)I'm happy to prepare a PR, but I realize this touches some very mature parts of the codebase, so I'd like to get some feedback first.
Expected behaviour
VCFEncodershould be configurable to output annotation values with floating point precision.Actual behaviour
VCFEncodercan't be configured in this way right now.