CAT_FASTQ: symlink fastq files when no merging needed

[funnell]

Description of feature

Related to #69, I had a look at the CAT_FASTQ output. Most of my files have just one set of fastqs per sample so the cat step unecessarily duplicates data.

Maybe this process can be updated to check if there are only 1 fastq in each of the input1 and input2 folders, and if so, just symlink the file to the output instead. Might need to symlink from the original fastq path rather than symlink the working dir fastq symlink.

[nickp60]

It should only be running the CAT_FASTQ process aliased as MERGE_RUNS for samples with multiple runs. For paired data, it runs CAT_FASTQ to combine R1 and R2 reads into a single file, which is the required input for all of the tools here except RGI which takes pairing into account. Are you seeing different behavior?

[funnell]

No, my apologies, I misunderstood what was happening.
That said, do you think process substitution could be useful here to avoid duplicating the fastq data?
Something like:

humann --input <(cat R1.fastq.gz R2.fastq.gz) ...

[nickp60]

I asked around on slack, didn't get many replies but I don't see any issues with that on my end! I only see one example of this in nf-core/modules here. If you want to submit PRs for the modules that use the merged R1/R2 file, I'm happy to review! Once those are accepted we can drop the CAT_FASTQ process here.