Gathering contig information from CSI files

[mliezun]

We have been trying to implement a tool for reading portions of a VCF file using a CSI index. We want to implement something like -r chr:beg-end|chr, but we see that there is no contig information in the CSI file as per this spec https://samtools.github.io/hts-specs/CSIv1.pdf

Could you help me figure this out? Or point me in the right direction?

[jmarshall]

It is a long-standing problem (see e.g. #70) that the index formats are under-documented.

In particular, the contents of l_aux/aux (“Auxiliary data”) are not described by CSIv1.pdf.

In practice, CSI — at least, as used by htslib/bcftools to index a BGZF-compressed VCF file — does store contig name information within this auxiliary data. You can see it if you view the index file with gunzip -c foo.vcf.gz.csi | od -c | head. See also vcf_idx_init() and hts_idx_tbi_name() (the latter's comments may be misleading) in the HTSlib source code.

When CSI is used by htslib/bcftools to index a BGZF-compressed VCF file, this aux block contains the format…names fields described in tabix.pdf.

[mliezun]

Thank you John, I will take a closer look to the tabix spec and the source code to see if we can solve this

[jkbonfield]

Closing as duplicate of #70.