I'm currently fixing the default arguments passed through config to the main imputation algorithm of phaseimpute. I'm wondering if it could be a good idea to add some nextflow arguments that the user could pass to the ext.args of each imputation process. Something like "opt_args_glimpse1" that would be added to

ext.args   = { "--impute-reference-only-variants --seed ${params.seed} ${params.opt_args_glimpse1}" }

This would allow users to fine tune algorithm without having to dive into all the configs files. Is it something that you do in other pipeline ?

Yes, we do it in mag: github.com/nf-core/mag/blob/…/nextflow_schema.json#… github.com/nf-core/mag/blob/…/nextflow_schema.json#…

I also do it in genomeassembernf-co.re/genomeassembler/2.0.0/parameters#…

I didn't knew but you can provide default in

nextflow.config

that call themselves previously define parameters. Something like

params {
    seed = 1
    glimpse1_args = "--impute-only-ref --seed ${params.seed}"
}

Doing so allows to expose all default arguments passed down to the software without risking the user defining twice some options. However sample level infos cannot be injected (e.g. meta.id).

I think that only works because the "${params.seed}" substitution happens later (in .command.sh?), something like

params {
   seed = 1
   glimpse_args = "--impute-only-ref -seed " + params.seed
}

will fail. I am not completely sure if I get what you are going for, are you suggesting this is an alternative for ext.args?

It seems that

nf-core pipeline lint

doesn't like the hack 😅

nextflow_config: Config default value incorrect: params.options_shapeit5 is set as --seed ${params.seed} --hmm-ne ${params.effective_population_size} in nextflow_schema.json but is --seed 1 --hmm-ne 100000 in nextflow.config.

It seems it resolve the default when comparing config and schema.

Copy code

nextflow_config: Config default value incorrect: params.options_shapeit5 is set as --seed ${params.seed} --hmm-ne ${params.effective_population_size} in nextflow_schema.json but is --seed 1 --hmm-ne 100000 in nextflow.config.

It seems it resolve the default when comparing config and schema.

This you could probably solve by just lying in nextflow_schema.json by manually substituting 😉

Yeah, but the aim was to properly show what and how was set the default to the user. I will look into it 🙂