SALMON_INDEX runs when using --aligner star_rsem even if samples have explicit strandedness

[drpatelh]

Description of the bug

nextflow run main.nf \
    --input https://raw.githubusercontent.com/nf-core/test-datasets/rnaseq/samplesheet/samplesheet_test.csv \
    --fasta 'https://github.com/nf-core/test-datasets/raw/rnaseq/reference/genome.fasta' \
    --gtf 'https://github.com/nf-core/test-datasets/raw/rnaseq/reference/genes.gtf' \
    -profile docker \
    --aligner star_rsem \
    --outdir results \

Even though we have specified reverse explicitly for all samples in the samplesheet the following process is still run to create the Salmon index in order to subsample the reads to infer the strandedness:

[3c/a91c08] process > NFCORE_RNASEQ:RNASEQ:FASTQ_SUBSAMPLE_FQ_SALMON:SALMON_INDEX (genome.transcripts.fa)             [100%] 1 of 1, cached: 1 ✔

This process shouldn't be run, however, it is being triggered since this logic evaluates to true:

rnaseq/workflows/rnaseq.nf

Line 239 in 6e1e448

!params.salmon_index && !('salmon' in prepareToolIndices)

We need to find a way to add to that logic somehow to also check that these channels are empty:

rnaseq/workflows/rnaseq.nf

Line 234 in 6e1e448

ch_strand_fastq.auto_strand,

If you have no intention of running pseudo-alignment, for now, the easiest workaround is to provide a dummy path that exists to --salmon_index /my/random/existing/folder/ as this won't actually be used or validated elsewhere by the pipeline and will save creating the index.

[drpatelh]

Fixed in #978