Dev 0.5.12

[noamteyssier]

No description provided.

[gemini-code-assist]

Code Review

This pull request introduces the revcomp command to reverse complement sequences in BINSEQ files and the verify command to compute order-independent checksums using xxh3-64. It also enhances grep and split with pattern reverse-complementation (--rc), adds header-matching capabilities to grep, and prevents binary data from being implicitly written to stdout. The reviewer feedback suggests improving error handling in revcomp by explicitly bailing out when --mate 2 is requested on single-channel files and reusing the already opened BinseqReader to avoid redundant file I/O. Additionally, the reviewer recommends suppressing noisy header warnings when verifying BQ files.

[gemini-code-assist]

If a user requests --mate/-M 2 on a single-channel file, the current implementation silently ignores the flag and writes the output file completely unmodified (since rc_extended is false and rc_primary is false), while still logging that it wrote reverse-complemented records. We should explicitly bail out in this case (similar to the verify command), and warn if -M 1 is passed redundantly.

pub fn run(args: &RevcompCommand) -> Result<()> {
    let reader = BinseqReader::new(args.input.path())?;
    if !reader.is_paired() {
        if args.mate == Mate::Two {
            bail!(
                "`--mate/-M 2` was requested but `{}` is single-channel (no extended/mate-2 channel)",
                args.input.path()
            );
        } else if args.mate == Mate::One {
            warn!("`--mate/-M 1` is redundant as only a single channel exists in `{}`", args.input.path());
        }
    }

    let builder = get_builder(&reader);

[gemini-code-assist]

Import bail from anyhow to support error handling when validating the --mate flag on single-channel files.

Suggested change

	use anyhow::Result;
	use anyhow::{bail, Result};

[gemini-code-assist]

Instead of re-opening and re-mapping the input file using MmapReader::new in get_builder, we can reuse the already opened BinseqReader by matching on its variants. This avoids redundant file I/O and simplifies the function signature by removing the Result wrapper.

fn get_builder(reader: &BinseqReader) -> BinseqWriterBuilder {
    match reader {
        BinseqReader::Bq(r) => BinseqWriterBuilder::from_bq_header(r.header()),
        BinseqReader::Vbq(r) => BinseqWriterBuilder::from_vbq_header(r.header()),
        BinseqReader::Cbq(r) => BinseqWriterBuilder::from_cbq_header(r.header()),
    }
}

[gemini-code-assist]

BQ files (.bq) never store header data by design. Printing a warning when verifying a BQ file is noisy and unnecessary. We should suppress this warning when the reader is a BQ reader.

        if fields.seq || fields.qual || fields.flags {
            if !matches!(reader, BinseqReader::Bq(_)) {
                warn!(
                    "`{}` has no header data; excluding headers from the checksum",
                    args.input.path()
                );
            }