BCF header contig IDX about noodles

[zaeleus]

Thanks for the detailed report. I see the issue here.

It was an unfortunate decision to entangle the format/spec with an implemntation detail. A few changes will have to be made to provide compatibility with htslib.

• bcf::header::StringMap needs to be reworked to allow nonsequential entries. This means it can no longer be backed by an IndexSet. (4291d88 and 4e39a77)
• A StringMap will need to be created for contig header records. Given that absolute positions can override relative positions, the order of vcf::header::Contigs cannot be used. (17397d7)
• Document that a record chromosome ID (chrom) is not an index into vcf::header::Contigs but an index into the contig StringMap (6b73f93).
• Writing a VCF record needs to use the contig StringMap instead of the VCF header contigs (9da12e7).
• Converting a BCF record to BCF record needs to the contig StringMap instead of the VCF header contigs (9c97ba8).
• Parse IDX field from raw record in vcf::header::Contig (22046f0).

I will work on these.

[zaeleus]

noodles 0.17.0 includes the changes that resolve this issue, mainly parsing the IDX field on VCF header contig records and also building a string map for those records.

The implementation is a breaking change. The user now has to parse a bcf::header::StringMaps, which includes both the dictionary of strings and dictionary of contigs. Typically, if any method will need both maps (e.g., Record::try_into_vcf_record and Writer::write_vcf_record), use the collection; otherwise, use the respective string map (StringMaps::strings or StringMaps::contigs). For example,

let raw_header = reader.read_header()?;
let header = raw_header.parse()?;
let string_maps: StringMaps = raw_header.parse()?;

let record = bcf::Record::default();
let dp = record.info().get(&header, string_maps.strings(), Key::TotalDepth);

let chromosome_id = usize::try_from(record.chromosome_id()).expect("invalid chrom");
let chromosome_name = strings_map.contigs().get_index(chromosome_id);

writer.write_vcf_record(&header, &string_maps, &record)?;