NAF files generated through the Python library can't be decompressed with `unnaf`

[apcamargo]

Hi @althonos,

I'm testing out nafcodec's Python library and I noticed that the files I write with it's encoder can't be decompressed with the unnaf command. See a reproducible example below.

import nafcodec

with nafcodec.open(
    "sequence.naf", "w", sequence_type="dna", sequence=True, id=True
) as e:
    e.write(nafcodec.Record(id="seq1", sequence="ATCG"))

$ unnaf -c sequence.naf
unnaf error: incomplete or truncated input

I can, however, read the sequence.naf within Python using nafcodec.

I noticed that in the README you mention that you still need to implement the encoder, so I'm wondering is this is indeed a bug or if you're still working on the encoder.

[althonos]

Ahah, I'll have a look 👍 No in theory the encoder is pretty much there, but I remember finding bugs in the decoder while doing that (e.g. the unnaf encoder couldn't handle DNA files in format version 2 because it expected those to always use format version 1, this kind of things) so I'll have to figure out if that's an unnaf bug or a nafcodec but

[althonos]

@apcamargo : So basically two bugs in one:

• The nafcodec encoder was writing the wrong decoded sequence length for DNA, as it was outputting the number of bytes instead of the number of characters (i.e. 2x shorter sequence length because of the 4-bit encoding). I made a fix for the next release.
• unnaf does not actually support archives without comment/record names, even though it should be totally fine; in https://github.com/KirillKryukov/naf/blob/2c8684c17de96f0b4da3b43393c230d3a39d0b67/unnaf/src/output.c#L613 it actually should be skipping names if they are not available, but right now it tries to always load names.

[althonos]

First bug fixed in v0.3.0, for the second one you have to do:

import nafcodec

with nafcodec.open(
    "sequence.naf", "w", sequence_type="dna", sequence=True, id=True, comment=True
) as e:
    e.write(nafcodec.Record(id="seq1", sequence="ATCG", comment=""))

if you really need unnaf compatibility.

[apcamargo]

Thank you, @althonos!