Using noodles_tabix for reading tbi-indexed tsv.bgz files

[holtgrewe]

Hi @zaeleus - could you give an example how I could use the noodles_tabix module for jumping in bgzip-ed and tabix indexed TSV files?

[zaeleus]

The general strategy to querying is to

1. find the relevant chunks in the index (csi::Index::query),
2. read the data between those chunks (csi::io::Query),
3. parse the records within that data, and
4. filter the records that intersect the region.

There is currently no way to query an arbitrary tabix-indexed file, but it is planned as part of #158. You can, however, implement 3 and 4. I added an example of this in noodles-tabix/examples/tabix_query_generic.rs.

[zaeleus]

noodles 0.39.0/noodles-csi 0.18.0 now supports reading generic indexed files. It introduces a record parser (IndexedRecords) and filter (FilterByRegion), e.g.,

let records = Query::new(&mut decoder, chunks)
    .indexed_records(&header)
    .filter_by_region(&region);

for result in records {
    let record = result?;
    // ...
}

This pattern is wrapped by tabix::io::IndexedReader, which simplifies noodles-tabix/examples/tabix_query_generic.rs.

[brentp]

Hi @zaeleus , I'm trying to further understand this.
Is there an abstraction / trait that unifies reading from an indexed BCF or VCF where the index is either CSI or TBI?

[zaeleus]

Not currently, but I'll add one! Tracked in #188.