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

nf-core

I think `ext.when` is deprecated, so I'm not sure about that.  I also doubt that this will have the desired effect, because disabling one process would not magically produce files that are not normally published in a given workflow.
Sorry, I use sarek in specific ways and am not familiar with all the options, so I can't give any good advice on your situtation.  I am pretty sure, though, the per-sample VCFs are _always_ being produced (since they would be the input for any later joining process, unless you don't run the variant caller at all), but they may not be _published_ with joint calling.  So they might be recoverable from the `work` directory, but you would have to find them there.
Maybe <@UE6D8290F> or <@UPC8CHDKQ> know of a way to do what you want.  Either way, for testing it might be good to prepare some small input files (FASTQ, or even mapped BAMs to use with `--step varinat_calling`) that allow you to run through the pipeline quickly and see what gets produced. :crossed_fingers:

Hi <@UE6D8290F> / <@UPC8CHDKQ>  Thomas suggested I ping you. *Question: in Sarek 3.8.1, is there a supported way to get per-sample gVCFs + per-sample VCFs both published for the same run?*
Use case: 500-sample WGS cohort. I need gVCFs as the archive format (for eventual cohort joint genotyping) AND per-sample VCFs. The two built-in modes seem mutually exclusive: default gives per-sample VCF only, `--joint_germline=true` gives per-sample gVCF + cohort joint VCF (no per-sample VCFs).
<@UPC8CHDKQ>, I tried your recipe `withName:GATK4_GENOMICSDBIMPORT { ext.when = false }` together with `--joint_germline=true`. In 3.8.1 the workflow aborts before `MERGE_HAPLOTYPECALLER` publishes:
```Hjava.lang.NullPointerException: Cannot invoke method getName() on null object
    at nextflow.extension.CollectFileOp.processItem(CollectFileOp.groovy:140)
    at Script_57ee94b0d1ea8856$_runScript_closure1$_closure2$_closure3.doCall(Script_57ee94b0d1ea8856:16)```
The 21 per-interval `.g.vcf.gz` files for the test sample DO exist in `work/` — they just never get merged/published because the session aborts. HC declares an optional output (`fileoutparam&lt;2:1&gt;`) which is missing in this config, and a downstream `collectFile { ... it.getName() ... }` doesn't null-check it. Cascading `ext.when = false` to `VCF_QC_BCFTOOLS_VCFTOOLS:.*`, `GATK4_GENOTYPEGVCFS`, `GATK4_VARIANTRECALIBRATOR`, `GATK4_APPLYVQSR`, `GATK4_VARIANTFILTRATION` doesn't help — the NPE is in a workflow-level Channel operator, not a process, so `ext.when` can't gate it.
Two questions:
1. Is there a supported config to get per-sample gVCF + per-sample VCF both published in 3.8.1?
2. Is the`CollectFileOp`NPE a known bug?
