A community effort to collect a curated set of analysis pipelines built using Nextflow.

nf-core

Not completely sure I'm in love with it? In the bwa example I don't see why we can't go `${prefix}/${fasta.getBaseName()}` - I don't really see the value in being able to manipulate both parts of that output. For the bamcmp example, I don't really see why you wouldn't just hard-code `_contaminant` and `_primary` and just have `${prefix}`?

There are some case very minor I admit in which hardcoded values aren't working

this whole replacing fasta.getBaseName by a suffix started with this

For something like fastqscreen, you might have a whole bunch of input genomes.
These may well be in folders like:
`GENOMENAME/genome.fa` (I have seen plenty of `genome.fasta` files)

The bamcmp example, in my previous workplace we hardcoded `_human` and `_mouse` via the prefix.

(because that is more informative than the generic `contaminant` )

As much as possible I would avoid hardcoding anything that can be setup via ext.whatever

but we can have a sensible default if no ext.whatever is found

&gt; The bamcmp example, in my previous workplace we hardcoded `_human` and `_mouse` via the prefix.
So in this module, you provide two BAMs - the same sample ID mapped to two references?

Musing on it - there seem to be three options:

1. `prefix`/`prefix2` - not necessarily very programmatic, but quite functional
2. `meta.id`, `meta.ref1_id`,`meta.ref2_id` - someone will murder me if I propose more meta keys. 
3. reference names are also process inputs. Perhaps a bit annoying to set up. With record types, functionally the same as the `meta` option.
I think that this is possibly a situation where I could acquiesce to a `prefix2` - but I think the process inputs option would be the most "technically correct" here.

Yes, this tool takes the same reads mapped to two references, and finds reads which are "better" mapped to one than the other

Yeah, we don't want more meta keys, so prefix/prefix2 works nicely here

Basically I feel in terms of provenance - the metadata for a BAM file needs to include both the sample metadata and the reference metadata. So that should all be available as part of the process inputs in one way or another.

But OK, that might be a job for "the onus is on you to make sure that's in `meta` and shove it in the `prefix` ".

If we had free for all on the meta, you could do that

currently the way we can is via setting prefix

For this tool, I don't really care myself, I'm no longer using it (and I would use different tools if I had to do the same job in the future), and would accept hardcoding the output if that is really what is necessary.
fastqscreen is more important to be honest.

For the aligners, I definitely think we can find a simpler solution that works with fastqscreen with a single prefix.