Not related to meta-omics, but to pipeline-chaining. I assume you have solved this: how do you generate a samplesheet for a downstream pipeline that contains the published output?

You could use the index from workflow outputs, but it requires some careful control and consideration over the structure of your publications

sarek: we create samplesheet in the pipeline: https://github.com/nf-core/sarek/blob/dev/subworkflows/local/channel_markduplicates_create_csv/main.nf

I see, I was trying something similar as sarek, but in a different (wrong) way ;D

different than sarek is not wrong, I do like different to be honest

Well, it didnt work, so...

checkout what I did in #CL88J906S https://github.com/nf-core/demultiplex/pull/411

My issue with a process approach is that you have to write the output paths twice, so there's a chance something can go wrong. I wish the index system was more flexible, at the moment it is not great

I am acutally not using a process

I do not like a process approach because it's wasting a process for just basic text processing that easily done in groovy

Here is what I am trying

ch_assembly_manifest = ch_main_scaffolded
        .map { it ->
            def meta = it.meta
                // A list of filenames, corresponding and the candidate output folder
                [
                [ meta.id, meta.scaffolds ? meta.scaffolds.ragtag ?: null : null,           'scaffold/ragtag'     ],
                [ meta.id, meta.scaffolds ? meta.scaffolds.hic ?: null : null,              'scaffold/hic/yahs'   ],
                [ meta.id, meta.scaffolds ? meta.scaffolds.longstitch ?: null : null,       'scaffold/longstitch' ],
                [ meta.id, meta.scaffolds ? meta.scaffolds.links ?: null : null,            'scaffold/links'      ],
                [ meta.id, meta.polished ? meta.polished.pilon ?: null : null,            'polish/pilon'        ],
                [ meta.id, meta.polished ? meta.polished.medaka ?: null : null,           'polish/medaka'       ],
                [ meta.id, meta.polished ? meta.polished.dorado ?: null : null,           'polish/dorado'       ],
                [ meta.id, meta.assembly, {
                    // Assembly publishdirs are a bit more specific
                    meta.strategy == "single" ? (
                        meta.assembler_ont == "flye" || meta.assembler_hifi == "flye"
                            ? 'assembly/flye'
                            : 'assembly/hifiasm'
                    ) :
                    meta.strategy == "hybrid"
                        ? 'assembly/hifiasm'
                        : 'assembly/ragtag'
                    }
                ]
                ]
        }

Personally, I think it's better that the downstream pipeline chases the upstream rather than the other way around so I'm not adding any samplesheet/param file generation to "my" pipelines. Moreover, nf-meta looks like a better way. Haven't tested that yet.

Hm, interesting perspective. Probably depends a bit on how obvious the link between the two pipelines is

I wish the index system was more flexible, at the moment it is not great

Hop onto

#nextflow-syntax

on the Nextflow slack with any suggestions!

I have proposed my request already! github.com/nextflow-io/nextflow/issues/6830