Run on Nextflow >=26.04 (strict syntax)

[hoelzer]

Closes #125.

Important

Stacked on #124 — it contains the smaller strict-syntax fixes this builds on. Merge #124 first, then this PR shows only its own commit.

Problem

Nextflow >=26.04 parses scripts with the strict syntax, where a script file may only contain declarations. clean.nf mixed ~180 lines of top-level statements with its include and workflow declarations, so the pipeline did not compile at all:

Error clean.nf:13:1: Statements cannot be mixed with script declarations
ERROR ~ Script compilation failed

nextflow lint reported 140 errors, all in clean.nf.

Changes

clean.nf

• parameter validation moved into check_for_unknown_parameters() and validate_parameters()
• the terminal banner moved into print_run_info()
• the input channels moved into the entry workflow, with input_channel(), fasta_channel() and control_fasta() for the branching parts
• exit 1, "..." → error "...", and exit 0, helpMSG() → log the help and return; exit is no longer part of the syntax. helpMSG() returns the message instead of logging it, and its colour variables are declared with def.

The order of checks and prints is deliberately unchanged, so --help still works without the otherwise required input parameters.

Directives can no longer reference process inputs

This one is a runtime failure rather than a parse error, so it only showed up when actually running on 26.04 (ERROR ~ No such variable: map_target):

• modules/bbmap.nf: the publishDir path containing ${map_target} is now a closure, which is still resolved lazily
• modules/alignment_processing.nf: pattern accepts no closure, so the two "${name}*.bam{,.bai}" patterns became static globs. For filter_true_dcs_alignments that would newly publish dcs.bam, so saveAs filters it out — closures do get the process inputs.

lib_pairedness

params.lib_pairedness was assigned at the top level. Assigning it from the entry workflow is flagged (Params should be declared at the top-level), so modules/functions.nf provides lib_pairedness() and the process scripts call it.

Result

nextflow lint on Nextflow 26.04.6: 140 errors → 0.

The CI matrix goes back to latest.

Verification

• 18 stub runs — fasta, nano, illumina PE/SE, --bwa, --bbduk — on Nextflow 23.04.1, 25.10.0 and 26.04.6
• 4 real runs on 26.04.6 exercising every changed code path: --min_clip (soft-clipped publishing), --dcs_strict (strict-DCS publishing), --bbduk (the map_target path) and --keep
• the published output of those 4 runs was compared against the same runs on the pre-refactor code: identical file trees and every file byte-identical (24/24, 18/18, 11/11, 30/30). dcs.bam is still not published.
• --help exits 0 with the full message, and all four validation errors exit 1 with their original message, on both 23.04.1 and 26.04.6

Not included

34 lint warnings remain (mostly Channel. → channel. and implicit closure parameters). They are not fatal and touch every workflow file, so they would drown out this diff.