Feature request - support glob deps in interpolates when dep will always be a single file

[zachsa]

Currently I have a task for generating import maps that is run on changes to an html file:

[[task]]
name = 'jspm:client'
targets = ['.cache/#.html']
deps = [
    'client/html/#.html',
    '.cache'
]
invalidation = 'always'
run = "node jspm-client.js"

I would like to be able to change the location of the html files, so that the deps becomes client/pages/**/#.html. I think this is generally useful (although since this job is always re-run for all html files I can just remove the interpolation from this and discover HTML files manually)

In this case the output of .cache is a rollup build that contains an unknown number of chunks

[guybedford]

What is the reason you cannot just use client/html/##.html as the dep and .cache/##.html as the target?

[zachsa]

Does the double hash find nested HTML files?

[guybedford]

Yes that is what the double interpolate form is for.

[zachsa]

This doesn't work as I want to normalize the output directory to be flat.

[guybedford]

Right, but flattening is by definition a many to one mapping, which cannot be supported in interpolation as interpolation must be fully reversible in its analysis.

[zachsa]

I see - thank you