•

bwamem2/index

. I’d use

${prefix}/${prefix}

or

${prefix}/${fasta.getBaseName()}

•

bamcmp

. I’d use

${prefix}

with some neutral qualifiers like

_A

and

_B

(and rename the output channels too !)

I need prefix/something else

In what situation would you have multiple indexes that would share the same directory name, that couldn’t be solved by making the directory name more specific?

Fastqscreen

Ah, looking at it - you pass it a directory full of multiple e.g. bwa indexes?

So our implementation in #C06PHJM0Y3D slightly differs from the regular modules. https://github.com/nf-core/seqinspector/blob/dev/modules/nf-core/fastqscreen/fastqscreen/main.nf And I'm hoping to port some of the logic back

and you don't want to

nf-core modules patch

it? I'm not sure nf-core modules needs to cater to everything, given we can write patches

Pipeline wouldn't pass linting without patching

Please can we not just assume that patching fixes everything! Particularly when it is so easy to make this work in a reasonable way.

Yes

Question, apart from for tidiness reasons, why do we even put the index in a directory?

Good question. Historical reasons?

def config_content = ref_names.withIndex().collect { name, i -> "DATABASE ${name} ./${ref_dirs[i]}/${ref_basenames[i]} ${ref_aligners[i]}" }.join('\n')

Perhaps I'm missing something, but this approach doesn't preclude providing a bunch of references as

${prefix}/${fasta.baseName}

(would need a little modification)? Perhaps we could even just drop the directory entirely, emitting the raw index files with a simple

${prefix}

, and do:

input:
tuple val(basenames), path(index_files, stageAs: "ref/*"), val(aligners)

script:
def config_content = ref_names.withIndex().collect { name, i -> "DATABASE ${name} ./ref/${ref_basenames[i]} ${ref_aligners[i]}" }.join('\n')

alternative

script:
def config_content = [basenames, index_files, aligners].transpose().collect { name, idx, aligner -> "DATABASE ${name} ${idx} ${aligner}" }.join('\n')