noodles_cram: write_record fails for reads from circular contig about noodles

[zaeleus]

noodles 0.20.0 / noodles-cram 0.13.0 fixes the main issue of building the substitution matrix and using the new substitution codes (plus a slew of CRAM I/O fixes). It passes your sample input, but do tell if there are any other issues. Some additional notes:

• If you're not using Reader::records but decoding raw records directly from the slice (Slice::records), you must resolve the CRAM records (Slice::resolve_records) before sending them to the writer. (This concept will likely be removed in the future.)

• The CRAM reader and writer now use fasta::Repository, a concurrent sequence cache that uses a storage adapter to look up and load sequence data. With an indexed FASTA file, you no longer have to load all the reference sequences beforehand, e.g.,

  let reader = File::open(fasta_src).map(BufReader::new).map(fasta::Reader::new)?;
  let index = fai::read(fasta_index_src)?;
  let adapter = IndexedReader::new(reader, index);
  let repository = fasta::Repository::new(adapter);

[zaeleus]

Thanks for the sample input; it's very useful!

The major issue here is that the writer builds and uses the new substitution matrix too late in the write process and is actually producing incorrect counts. Furthermore, new substitution codes are never replaced in the input record.

I'll work on this this week.