recommended way to build conda envs

[wwood]

Hi,

Kinda newbie thread of thought here.

I'm wanting to run several aviary runs in a snakefile in parallel, but don't want them all building conda envs y themselves. What is the recommendation for a separate step that builds the conda envs? Does aviary recover --build build everything necessary if the argument to build is missing?

I guess I'm looking for something like this to work

aviary recover --conda-prefix ~/e/aviary-conda-prefix --build

or specificalyl for short reads

aviary recover -1 dummy.fastq -2 dummy.fastq --build --conda-prefix ~/e/aviary-conda-prefix

But those both error out with missing input exceptions. Maybe --build should have a default?

Oh, it works when the input read files actually exist (maybe needs to handle that more gracefully).

Also is a bit confusing that there's a [BUILD] afterwards in the help. I interpreted that to mean a specific snakemake rule needs specifying. Maybe [yes|no]?

What about changing the help to

Build conda environments necessary to run the pipeline, and then exit. Equivalent to "--snakemake-cmds '--conda-create-envs-only True ' ". Other inputs should be specified as if running normally so that the right set of conda environments is built.

?

[rhysnewell]

Yeah, ideally the build command would just build all of the environments without needing the read files but I unfortunately couldn't get snakemake to co-operate in that regard. The help message can be updated

[AroneyS]

I just used placeholder files to get Aviary conda env built in Bin Chicken (https://github.com/AroneyS/binchicken/blob/823bd135a945142cfb7c26d951616068d8e45e32/binchicken/binchicken.py#L1006)