query_unmapped fails about noodles

[zaeleus]

Are you able to update to noodles 0.50.0 / noodles-bam 0.44.0? Reader::query_unmapped now requires the header, which should fix this issue.

• bam/reader: Require a SAM header when querying for unmapped records (Reader::query_unmapped).

  It's possible for a chunk to include mapped records, which are subsequently filtered out, but they do require the associated header to decode.

[zaeleus]

So by my reading of the spec, I think it is allowable, for an unmapped read, to have RNAME set, but POS as 0.

This is not allowed in by the spec. An unmapped record with no coordinate (i.e., POS = 0) does not have a reference sequence name (i.e., RNAME = *). See § 1.4.3 (2022-08-22): "An unmapped segment without coordinate has a * at [RNAME]."

Does the query range allow 0 to be part of the interval?

No, 0 is not a coordinate that can be queried from the index. To query a full region, use an unbounded range, e.g., Region::new("sq0", ..).

[drtconway]

Great! That appears to work, thank you!

While I have your attention, can I ask a related question?

So by my reading of the spec, I think it is allowable, for an unmapped read, to have RNAME set, but POS as 0.

Does the query range allow 0 to be part of the interval? My understanding is that Position starts at 1.

I guess the simplest form of query to achieve my goals, would be to do a query that just iterates over anything with the given RNAME.