alignment input

[rneher]

The rust implementation currently follows a pattern similar to nextclade where the sequence input is read from positional arguments or stdin, while tree and metadata are keyword arguments.

However, the nature of the sequences is different here from what it is in nextclade:
Sequences are not independent entities that can be processed from one or multiple files. Instead, there is a mapping from leaves of the tree to sequences in the alignment (it is ok if some sequences are missing). Sequences also have to be aligned, so it is reasonable to expect that all sequences are in one file. If there are multiple files, it would be natural to treat these as different partitions.

One could imagine flags like --nt-aligments partition1.fasta partition2.fasta and --aa-alignments HA.fasta NA.fasta PB2.fasta.

[rneher]

We anyway had on the ToDo a rethink/abstraction of the partitions such that we can flexibly handle partitions with different alphabets and different types. Probably better to tackle this after that re-org.

The primary partition types I see are sequences with different alphabets (nuc, aa, maybe binary at some point) and single discrete traits with arbitrary alphabet. The former can be either sparse or dense.

For more complex specifications (like different partitions for different parts of a sequence, 1st, 2nd, 3rd codon position etc, we might need config files).