Add script recommendations

[pinin4fjords]

Propose some lines about module scripts.

@netlify /docs/guidelines/components/modules

[netlify]

✅ Deploy Preview for nf-core-main-site ready!

Name	Link
🔨 Latest commit	c438115
🔍 Latest deploy log	https://app.netlify.com/sites/nf-core-main-site/deploys/6759db8f6566850008f6b011
😎 Deploy Preview	https://deploy-preview-2888--nf-core-main-site.netlify.app/docs/guidelines/components/modules
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[netlify]

✅ Deploy Preview for nf-core-docs ready!

Name	Link
🔨 Latest commit	c438115
🔍 Latest deploy log	https://app.netlify.com/sites/nf-core-docs/deploys/6759db8fdbef930008982d13
😎 Deploy Preview	https://deploy-preview-2888--nf-core-docs.netlify.app/docs/guidelines/components/modules
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[pinin4fjords]

@nf-core-bot fix linting

[jfy133]

Ugh, escaping in suggestions is hard... hope you undersatnd it...

I also decided to be more strict, and require a stadnardised tempalte file naming structure (Which would require an update to deseq2/differential, let me know what you think...

[jfy133]

A few further tweaks

[jfy133]

Suggested change

	#### Script template stub
	#### Module template stub

[pinin4fjords]

@nf-core-bot fix linting

[christopher-hakkaart]

Can we reference the reason why not to use module binaries? Personally I don't really like the interface to template, and I hate to see R scripts written with base R and the horribly clunky code to allow arguments in that way. I would suggest that we encourage the use of a seqera container with an appropriate interface package (e.g. argparse). We could do with more explanation as to how to actualy use template, because I don't think we have that here (and the Seqera docs are almost non-existent on the topic).

A PR is in progress to address the gap in the Nextflow docs. This doesn't mean that nf-core can't have its own standards/recommendations, but it will hopefully add something that can be easily referenced.

[SPPearce]

Can we reference the reason why not to use module binaries? Personally I don't really like the interface to template, and I hate to see R scripts written with base R and the horribly clunky code to allow arguments in that way. I would suggest that we encourage the use of a seqera container with an appropriate interface package (e.g. argparse). We could do with more explanation as to how to actualy use template, because I don't think we have that here (and the Seqera docs are almost non-existent on the topic).

A PR is in progress to address the gap in the Nextflow docs. This doesn't mean that nf-core can't have its own standards/recommendations, but it will hopefully add something that can be easily referenced.

Great, always happy to see more documentation on the Nextflow page itself.

[jfy133]

Can we reference the reason why not to use module binaries?

Apparently you can only use binaries in cloud contexts if you use wave (not other containers), which I feel is overly restrictive.

Personally I don't really like the interface to template, and I hate to see R scripts written with base R and the horribly clunky code to allow arguments in that way. I would suggest that we encourage the use of a seqera container with an appropriate interface package (e.g. argparse). We could do with more explanation as to how to actualy use template, because I don't think we have that here (and the Seqera docs are almost non-existent on the topic).

Agree, more Nextflow docs/example on how to use would be nice, and again, would rather not 'force' Seqera Containers.

I much prefer the use of in-line scripts where possible, what is the justification behind saying it can only be 20 lines?

It's a rule of thumb suggested by @pinin4fjords (I've changed to 'approx' though)

Oh, and some examples pointing to other modules rather than deseq2/differential repeatedly would be good.

Suggest some then! 😝

[christopher-hakkaart]

I suggest removing the repeated "approximately 20 lines" and instead describing this once: something like, "When a script extends beyond a readable length (approximately 20 lines)..."

Then just refer to a readable length for the rest of the text.

Readable length could be "Manageable size", "too verbose", "excessive length", or something like that

[jfy133]

I suggest removing the repeated "approximately 20 lines" and instead describing this once: something like, "When a script extends beyond a readable length (approximately 20 lines)..."

Then just refer to a readable length for the rest of the text.

Readable length could be "Manageable size", "too verbose", "excessive length", or something like that

Done!

[christopher-hakkaart]

Much better - thanks!

[pinin4fjords]

@SPPearce to be clear, I want module binaries to be the way to do this, with argparse baked into the container to avoid all that junk I have to put in templates, I'm no template fanboy.

But module binaries are not entirely portable (see the provisos in the linked docs), so templates are the least worst way of doing this I've found - overjoyed if someone tells me a better way.

My crappy outdated R is another issue entirely ;-)

[christopher-hakkaart]

Is everyone reasonably happy with the current PR? Are there any blockers to merging?

[Joon-Klaps]

Nothing is mentioned on the R_sessionInfo.log that is created in the template there.

Also an alternative suggestion for writing the R version:

paste(version[['major']],version[['minor']], sep = ".")

Think it's more easier to read then:

strsplit(version[['version.string']], ' ')[[1]][3]

[jfy133]

Nothing is mentioned on the R_sessionInfo.log that is created in the template there.

What do you mean? I don't follow, sorry.

[jfy133]

For the second comment @Joon-Klaps please feel free to update the moduel/example :)

[Joon-Klaps]

In the mentioned example lines of code there is a small section dedicated into exporting the R.sessionInfo().

sink(paste(opt\$output_prefix, "R_sessionInfo.log", sep = '.'))
print(sessionInfo())
sink()

However, nothing is mentioned of this in these guidelines. So is it necessary or not? I'm guessing so given it was included in the referenced lines but would also explicitly mention it then in the guidelines. Although not certain about the added value of the R.session() info.

[Joon-Klaps]

nf-core/modules#7243

[jfy133]

Ah got you. I don't think that makes sense putting 'how' to generate the yaml file in the guideline per se - we only care that it exists.

We can point to other examples though if it helps (feel free to make an additional PR). Maybe a dedicated tutorial page how to write such a template module would make sense due to the intricacies of the variable escaping? What do you think @pinin4fjords ?

[bentsherman]

The Nextflow docs discourages the use of templates for non-bash scripts: https://nextflow.io/docs/latest/process.html#template

Template scripts are recommended only for Bash scripts. Languages that do not prefix variables with $ (e.g. Python and R) can’t be executed directly as a template script.

The best practice for using a custom script is to embed it in the process definition at first and move it to a separate file with its own command line interface once the code matures.

[bentsherman]

But I guess the module binaries feature requires Wave when the work directory is in object storage

[jfy133]

Yes, the requirement of wave is why we opted not to do so as then it's not portable :/ (and use the examples from Jon)

[pinin4fjords]

@bentsherman any chance of module binaries becoming a universal solution in future, like workflow bin? My last discussion with our lord and master suggested not, but this is is the unavoidable consequence (as far as I can tell), unless we embed massive chunks of code in the process definition, which I don't think is a better way.