Per the warning <here> I have found that in my `EN...
# modules
a
Per the warning here I have found that in my
ENSEMBLVEP_VEP
processes I am getting all my extra
vcf.gz
files that are passed as
extra_files
copied back into my task
workDir
. This is not great when running on AWS with S3
workDir
as it increases copy time and generally costs more. Is there any reason not to change the output channel designations from
path("*.vcf.gz")
to
path("${prefix}.vcf.gz")
?
m
I don't see any atm 🤔 feel free to fix!
m
yeah go for it, I did this for several other modules
it does make total sense
r
shared a GIF
should we generally add this to the guidelines to avoid wild cards in the output? We shoot ourselves in the foot with this a little too often
party this 3
a
I was just about to ask the same haha. Given that advice, and the willingness here for that change, I wonder if nf-core should push for using more qualified output globs? It obviously requires more knowledge of what a tool outputs and can break more if different versions do things more differently, but in general I have found
${prefix}*.vcf
(or whatever extension) works well with my local modules
spiderman pointing 2
r
given it should be caught with the tests on a module change, the blast radius feels pretty contained to me if a tool changes the naming structure on update. <!subteam^S044DU0NG68> ☝️ wdyt?
👍🏻 1
s
Yes, this comes up regularly
e
If it is for the better..always do it!
a
Do it!
j
shared a GIF
m
Update it. There's also a related issue on Nextflow: https://github.com/nextflow-io/nextflow/issues/3995