Contradiction in docs about when to use multiple input channels

Ask for help
1

I was confused (and still am to be honest) about when to use multiple input channels vs tuple inputs. So I looked around in the docs and found Processes | Seqera Docs saying

Do not supply more than one channel when calling a process with multiple inputs. Invoking a process with multiple channels can lead to non-deterministic behavior. All additional inputs should be dataflow values.

“Ok, great!” I thought. But then I read Processes | Seqera Docs and saw the following example

process blast {
  input:
    val species
    path query

  output:
    tuple val(species), path('result')

  script:
    """
    blast -db nr -query $query > result
    """
}

workflow {
  ch_species = channel.of('human', 'cow', 'horse')
  ch_query = channel.fromPath('*.fa')

  blast(ch_species, ch_query)
}

Here we are passing two channels as separate inputs. Is this truly a contradiction or am I misunderstanding something? I also saw Multiple Inputs vs Tuple Inputs which indicates that you should indeed not pass two channels as separate inputs.

2

The second example needs to be updated to call the process with a single combined input:

blast(ch_species.combine(ch_query))

And the process should be updated to declare a single tuple input.

Thanks for reporting. Can you submit a GitHub issue so that I can track it properly?

3

Yep. Here you go: Outdated documentation using multiple inputs · Issue #7124 · nextflow-io/nextflow · GitHub