Hi all, I opened a PR in #CL88J906S migrating from publishDir to workflow outputs. If anyone's available for their review or opinion feel free to check it: https://github.com/nf-core/demultiplex/pull/379

THANKS a lot

Yes, the losses are because some outputs are not captured by the subworkflows, and therefore I cannot capture them on the workflow outputs (example: kraken not outputing fastq). The gains seem to be that some things that are being written to a single output in the module (e.g. all fastq) but we captured only a subset of them via publishDir (e.g. only fastp fastq). So it is kind of tricky to have exact the same outputs unless we modify some modules and subworkflows. There are also some inconsistencies with how the outputs are being written between different modules which also makes it difficult to separate. What do you suggest we do?

oh, let me check all that again then

thank youuuu

Aren't all snapshots in tests/ pipeline level tests?

i agree, i just needed an external opinion hehe ok i'll first focus on modifying the modules to rescue the missing files.

I feel like: demultiplexed_fastq, demultiplex_reports, demultiplex_stats and fastq_idx have the same output structure. similar for: checkqc_reports, fastp_reports, falco_reports amd md5_checksums

thank you so much for your feedback and your time!! i'll move it to draft and keep working

Anytime

Ooh, please can we only publish the fastp trimmed files when we trim 😄