nextflow-authoring

Create one process per file with a tag, label, pinned conda and container directives, tuple val(meta), path(files) inputs, emit-named outputs, a versions.yml emission, and a mandatory stub block. Use task.ext.args and task.ext.prefix instead of hardcoded arguments so the module stays configurable and portable.

groupTuple hangs when it waits for a group that never completes, typically because a sample was dropped upstream and no size was specified. Fix it by passing size: N when group sizes are known, for example groupTuple(by: [0], size: params.n_replicates).

join performs an inner join on a key and drops unmatched items unless remainder: true is set. combine produces a cartesian product of two channels. mix merges channels with non-deterministic ordering, so it must never be used for order-dependent logic.

Define named profiles in nextflow.config setting process.executor to slurm or awsbatch, plus queue, singularity.enabled with autoMounts on HPC, and wave with fusion on AWS. Set executor.queueSize and submitRateLimit to avoid flooding the scheduler.

Create a main.nf.test file next to the module declaring the process, inputs in a when block, and assertions in a then block such as process.success, output channel sizes, and snapshot matching. Run it with nf-test test and include a stub test using the -stub option.

Do not use it for running existing nf-core pipelines on data; that execution-focused work belongs to the nextflow-development skill. This skill covers writing, structuring, and debugging DSL2 code, not pipeline operation.