A draft of the new guidelines for modules: hackmd.io/YmVPL55sTomtFIpNKlRKBA I'm still working on it but feel free to comment if I did something stupid

I would argue strongly that

prefix

should be a required input as well as

args

. Otherwise subworkflows still have to depend on config

Yeah we're currently also discussing this 😛

Otherwise this looks really nice <3

The one main contention we have is for example chaining modules e.g. samtools sort -> samtools view and if say

samtools_args

is set, we don't want that propagating to samtools view when it's set for samtools sort

I'd say we use

samtools_sort_args

and

samtools_view_args

in that case

I like that better

Either that, or arguments as a list, if we want to maintain the existing pattern better?

record(
   ...
   arguments: List<String?>
)

...

tool \\
${arguments[0]} |
tool2 ${arguments[1]}

I'm against indexed args. It's very easy to get stuff mixed up

Yeah I'd like it to be more descriptive

Generally agreed, but there are cases it would be necessary: github.com/nf-core/modules/pull/11550

That's definitely not a standard module 😅

but there it would still be more readable to have the tool name in the args

Though perhaps in the case that there are multiple tools,

arguments

itself should be a record with keys? That would be easier for linting

record(
    ...
    arguments: Record
)

And we shouldn't make anything else less clear just for this

but there it would still be more readable to have the tool name in the args

Yes, until you run samtools view twice 🥹

samtools_view1_args

and

samtools_view2_args

?

We need a method to deal with position placement anyway. There are tools that require global and tool specific options for example

I'd say we use samtools_sort_args and samtools_view_args in that case

Re this, maybe you could specify the naming SHOULD(?) match the same rules for naming of nf-core modules/subworkflows etc?

We're thinking of multi-tool commands here too like those with bwa and samtools

And then only add numbers when you have repeated use of the same command twice in one pipe (like

samtools_view

)

Yeah that one still needs some work, I might remove it in general

record MyProcessArgs {
    tool1_args: String
    tool2_args: String
}

process MYPROCESS {
  record(
      ...
      arguments: MyProcessArgs
  )

But what about putting these all in a record rather than as flat record inputs?

I think it will generate more adapter code with little benefit. In the end we just need a few strings inputs

I think a stageAs example would be good (plus perhaps demonstrating how it can be used for "module binaries" by staging a script?).

Do you have a module in mind? I thought module binaries were wave only so we never used them.

I'm not sure module binaries are possible using stageAs, at least not in a portable way

Basically the suggestion from Ben was you could put a script in a

moduleDir

and stage it into the work directory using stageAs: https://nfcore.slack.com/archives/C04QR0T3G3H/p1773668639446329?thread_ts=1769678574.070939&amp;cid=C04QR0T3G3H

oooh I like this!

Need to confirm that works since that was back in march right?

do it eyebrows