Maintenance: linting

[Magellol]

Invoke the linting too through intlc-internal

A "faulty" json file can be found here https://github.com/unsplash/unsplash-web/blob/master/app/routes/Photos/lang/en-US.translations.json

A good one here: https://github.com/unsplash/unsplash-web/blob/master/app/routes/Press/lang/en-US.translations.json

e.g

intlc-internal lint app/routes/Photos/lang/en-US.translations.json
TooManyInterpolations
# exit 1

intlc-internal lint app/routes/Press/lang/en-US.translations.json
Success
# exit 0

TODO

• [x] Make linting fail on Failure msg

Nice to have

• [ ] Accumulate linting errors. Right now it will stop as soon as it sees one

[Magellol]

@samhh I've got a start on invoking the Linting module in the CLI af9013c — I think it might be worth looking at accumulating errors already? 🤔 (I wonder if we could print lines and columns from source as well)

[samhh]

I think it might be worth looking at accumulating errors already?

Certainly! It'd be great to get output something like this:

"msgA": too many interpolations
"msgB": non-ASCII, too many interpolations

I wonder if we could print lines and columns from source as well

This needs some research. At the moment our AST doesn't hold onto source information after parsing is complete. As with nicer parser errors this could come later, it's not a blocker.

[Magellol]

@samhh Yes! I think 0742643...f79160c does the trick

[Magellol]

(pre-)nested interpolations yet

Hmm. Just so I understand, if something is nested that means it got flattened right? But we're going to lint against english translations and we don't commit flattened format in english files I think 🤔

[samhh]

They can show up prior to flattening, for example:

{hasCollection, boolean, true {<collectionEl></collectionEl>} false {one of their collections}}

[Magellol]

@samhh Wasn't easy to figure out! Mind taking a look at https://github.com/unsplash/intlc/commit/cd03f4e3e4620f4b5f14f41d4f9a41bde66e101c? I didn't carry on with pattern matching the Plural case because it is actually quite complex and before I attempt to do it I was curious if my approach made sense? I think it's sound and my tests are passing but perhaps there is an easier method to this madness?

[samhh]

Pushed some updates. A lot of commits but @Magellol already did all the heavy lifting. :slightly_smiling_face: Summary:

• Finishes what @Magellol started - the pattern matching you didn't like is indeed something I'd like to improve that's come up before. Lenses might be one way. See: #48
• Simplifies the core implementation via a getStreams abstraction. Fixes laziness in nested streams as well.
• Adds support for multiple lint rules so there's no friction for @nim442.

[samhh]

One more thing we need is to add CI output for the "internal" binary. That can wait for another PR.