tree inference & data management

[rvosa]

It is probably a good idea to run the steps incrementally, by which I mean: writing the alignments is a task that you can kill and then it resumes where it left of, until all families have an alignment. Same for the constraint trees: you probably want to have a tree for every alignment and store this in a folder structure. What you want to avoid is having to run everything in one giant loop that can never fail.

With that in mind, it's probably better to not have a temporary tree that has a throwaway name here, but rather have a system where the raxml run assumes that there is a constraint tree for every alignment (and they probably have the same name but with different extensions).

[rvosa]

Think of it as a pipeline with different steps. You'd want to be able to do something like make alignments and it starts spitting out alignments, possibly in parallel. If you kill the job, it'll be able to pick up where you left off. If you do make subtrees it knows that it first needs to make alignments and make constraints, i.e. it needs to be aware of dependencies. This can be accomplished through a variety of workflow management tools, maybe the easiest one being SnakeMake.

[Naomivanes]

Changed the temporary constraint tree to one that gets saved in a folder. Will make a new issue to make a snakemake pipeline/rewrite the code to be able to run in a snakemake pipeline.