Reproducible compression

[karel-brinda]

Hello,

I think GeCo3 is a fantastic work, but currently it's applicability beyond the scope of experimental compression works is currently zero due to the missing reproducibility of decompression. And this is really pity because it's a great method!

Is there a chance that there could be support for a reproducible mode of compression?

Ideally, the program should be able to fail if the decompression possibly incorrect, and probably even fail (unless there's some mechanism like md5 that would guarantee that the result is the same despite a slightly different instruction set).

[karel-brinda]

For instance, we're currently using it for compressing Masked Superstrings of k-mer sets, where we use it for compressing the actual superstring of input $k$-mers – ie. a perfect use case for GeCo3.

However, it's impossible to share such compressed superstrings with anyone as they wouldn't likely be decompressed correctly, we have to resort to xz2 in the end.

[pratas]

Hi Karel,

Thank you for your words.
If the purpose is to compress such sequence in FASTA format without reference, then I recommend JARVIS3.
Is this the purpose?

Cheers

[karel-brinda]

In this use case, we always have a single sequence – superstring.