Typed processes are still a preview feature, no?

For a few more weeks, yes

I think it's on our road map anyway. Might be jumping the gun somewhat, but good to get it done

There's a converter built into the language server, right?

Yeah, I'm not saying it's rocket science or anything...

Problems I forsee: • Would need to bump the minimal Nextflow version • Could require a lot of concomitant changes to subworkflows, e.g. where

[]

is used in place of a null input

[11:52 AM]

Mostly, it looks like it might make most sense to do workflow outputs together with record types

Yes, that's my plan for rnaseq soonish

My thinking is: try to get as many pipelines up to date as possible (see my hackathon project), then work on module updates (project) and then go to town

Records are a whole other kettle of fish, I'm not sure we've even figured out how to use them in an nf-core context, or whether they're desireable? • how do we manage sample metadata in records (keep a meta map/record, or move the fields to the base map? in which case, how do we emit them from the process) • defining best as possible a standard vocabulary for field names to avoid endless renaming. i think this could be really thorny, there is always going to be a conflict between what's useful to name fields in a pipeline vs what's useful to name them in a generic module

yeah I think static typing is probably the easier of all of these

IMO the new output syntax is impractical in rnaseq without records. There will just have to be .... discussions....

Better lineage, pipeline chaining, etc. etc.

Workflow outputs are /so good/, I agree, I've been using them in my pipelines

honestly I don't really mind how we get there, as long as we get outputs 😅

Currently I'm collating sample-level information into an output record/map in each subworkflow and just passing them down

Will the next version be out before the hackathon? Otherwise I would be happy to set up a group to start experimenting and writing some guidelines for static processes. Record types could be part of that but those might need more experimenting than we can do in 2 days

workflow outputs are great, except for the fact you have to manually pass them between subworkflows/workflows

Will the next version be out before the hackathon?

Yes, aiming for the week before

Okay I'll submit a PR for a project later today 😉

Can probably count me in! I'll be dragged off to training rooms at times, but I'm obviously keen to work on it

Yeah I'll put myself as the main contact point, feel free to drop by as much as you like 😉

Typed processes are still a preview feature, no?

---

For a few more weeks, yes

Should be stable in 26.10

With the same syntax ? I remember we got bitten by the last preview feature when it turned stable with a different interface and the reason was that “preview features are not stable, use at your own risk”. I don’t want the same to happen again

@Ben Sherman I don't think the syntax is changing, is it?

correct, no syntax change this time

I completed a working version of the rnaseq migration. Ben has promised to review and may call me silly, but thought it would be good to have something concrete to argue about at the hackathon: github.com/nf-core/rnaseq/pull/1945