Bug in running function within publishDir block

[gevro]

Hi,
In nextflow 25.04.3, this worked:

Defined functions at the top of the nextflow script:

sampleBaseDir = { individual_id, sample_id -> "${params.analysis_output_dir}/${params.analysis_id}.${individual_id}.${sample_id}" }
dirProcessReads = { individual_id, sample_id -> "${sampleBaseDir(individual_id, sample_id)}/processedReads" }

Then in a process do:

publishDir { dirProcessReads(individual_id, sample_id) }, mode: 'link'

But in nextflow 25.10.0, this is giving an error:

ERROR ~ Unexpected error while finalizing task 'mergeAlignedSampleBAMs (mergeAlignedSampleBAMs: SampleID' - cause: No such property: individual_id for class: nextflow.script.dsl.ProcessDslV1

[gevro]

F/u investigation suggests that task variables are not available anymore inside publishDir closures. I figured out a workaround using the publishDir saveAs option whose closures do have access to task variables.

[bentsherman]

Can you share the full process definition? Or at least the inputs?

[gevro]

Inputs:

input:
      tuple val(individual_id), val(sample_id)

[bentsherman]

Thanks

So far, confirmed that the errors still happens:

• with the strict syntax (NXF_SYNTAX_PARSER=v2)
• with typed processes (ProcessDslV2)
• when rewriting sampleBaseDir and dirProcessReads to functions

So there are no workarounds, except perhaps to use workflow outputs instead of publishDir, but this error likely also happens for other process directives

It's almost certainly a regression caused by #6368 . We had to break up the old monolithic process DSL class in order to support the new typed process syntax

[gevro]

Thanks. Will this be fixed in a future version?

[bentsherman]

So it looks like the error only happens for a few directives: accelerator, arch, publishDir. Which makes sense because all of these directives are weird in the same way

I think your example used to work due to some under-the-hood Groovy DSL magic, but now it doesn't. Honestly I was expecting something like this to break 😄

Anyway, I should be able to come up with a fix pretty quickly. We'll try to patch 25.10, hopefully 25.10.1, with a fix included

[gevro]

Thanks! Which version will the fix be deployed in?