Codecov Report

Merging #626 into dev will increase coverage by 7.35%.
The diff coverage is 72.11%.

@@            Coverage Diff             @@
##              dev     #626      +/-   ##
==========================================
+ Coverage   68.09%   75.45%   +7.35%     
==========================================
  Files          11       11              
  Lines        1984     2159     +175     
==========================================
+ Hits         1351     1629     +278     
+ Misses        633      530     -103     

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 129bc8f...d89850b. Read the comment docs.

OK, so when I create a group, name it, and then delete it, it doesn't disappear from the builder, and I can't access options (with the cog).
I can delete a group that I haven't named, but not one that is named.

Deleted group that doesn't disappear from the builder is actually not saved in the schema, but still disturbing to see it there.

I like the possibility to add an icon to a params, but do we really need all brands from font-awesome?

Given the huge amount of icons, is there any guidelines on which icons to use for some common params?
Or should we already hardcode the basics common params (--input, --help, --genome...)
Like should I get info, info-circle, hands-helping, question or question-circle for --help.

I am allowed to try to change the icon for a group, but it is actually not working.
Group does have the possibility to have an help message which is disturbing as it's naming it params..

I understood that nf-core schema build . --url ... is to change the url of the builder, which I guess was useful when developing.

I was hoping you actually could save your progress and access it again, which also explained the 24 hours expiration period.

I'm now guessing that if you want to continue working on a schema, you need to save it first, and then continue with the json.

Could be good to detail that in the docs.

Keyboards shortcuts for power users is a good idea.

But with a long enough list of params, if you're on the bottom and use shift + enter to go up, at some point the focus goes too up and you actually don't see it anymore.

EDIT: and similar behaviour with shift + ctrl + ↑

Also, it's shift + enter to go up, not tab + enter

If you’re able to make a gif of this kind of behaviour to explain what you mean, that would be fab 👍🏻 Screenshots also fine.

I agree that having defaults for --input for example and --help would be cool, as they will be very similar across pipelines and should be using the same font awesome icons as Maxime said :-)

Mentioned on slack - my idea for this is that if no schema is found it will auto generate one using the pipeline schema, then update this against the pipeline.

Should be relatively little code to add and will mean you start off with a nice boilerplate schema with all of the common options already done for you.

Sounds good to me :-)

#630

Ok, rewrote how the schema builder waits for a response from the website so that it no longer times out. It should wait forever now.

TODO:

• Convert answer variable types after prompt (number, integer, range)
  • Avoid ERROR: [✗] Input parameters are invalid: '5' is not of type 'integer'
• Give selection prompt for parameters with enum?
  • Only if number of options is small?

The json schema web interface looks amazing and works super smoothly and made me realize of a couple of params in Bcellmagic missing in the nextflow.config, so great stuff!

I have just one minor comment:

• Dragging a parameter into a new parameter group works, but it took me a while to make it work. Maybe making the box where the single parameters fit a bit bigger when it is empty makes it easier.

For the nf-core launch cli:

• When a parameter is required but not provided, there is no note for that but only an Error on the very end: ERROR: [✗] Input parameters are invalid: 'metadata' is a required property. A (required) note could be shown next to the parameter, to avoid having to repeat the process for all required parameters.

Thanks for the testing and feedback @ggabernet!

Regarding the last point, for me this already happens:

Could you give a bit more info about when this happens for you so that I can try to replicate it?

With the JSON schema builder I set metadata as required:

The JSON description for metadata in the builder is then the following:

                "metadata": {
                    "type": "string",
                    "description": "Input metadata sheet",
                    "fa_icon": "far fa-file"
                },

Edit: found that this is listed here:

            "fa_icon": "fas fa-terminal",
            "required": [
                "metadata",
                "cprimers",
                "vprimers"
            ],

Then I launch the pipeline with nf-core launch, it finds the schema and groups the parameters in the groups I defined in the builder, however I can press Continue through this group of params and it does not prompt me an error that metadata is required. Just in the end:

It does prompt me an error if I try to edit it and leave it empty, though, the same as you described:

Aha, ok - so it needs to validate at the group level as well as the parameter level 👍 Thanks, yes I should be able to do this..

And yes, in the template we have a default value set for reads but maybe it's a better idea to leave this without a default (in the schema at least).

One other small thing: when I run nf-core launch . in a directory where there's no pipeline, it still creates an nextflow_schema.json.

In the end it complains that it does not find the main.nf:

Maybe it could check for the main.nf already when creating the schema, or if nextflow.config is not there, just stop.

Hah, excellent! Apparently nextflow config works fine with an empty directory too. Ok I'll put in a check for that, good spot 👍

Ok, I think that all of your points are now addressed @ggabernet 👍