Redesign content rule to support all possible values

[erykpiast]

It's, unfortunately, a breaking change. To support values other than strings, I had to introduce variant type, so existing rules need to be migrated like this:

-  contentRule("foo")
+  contentRule(`string("foo"))

Do you think that's okay? Can we afford another major version in such a short time? Maybe there is another way to model this?

If we eventually agree to break the API, does it make sense to rename contentRule to content? (and contentRules to contents I suppose). I was quite surprised that content function is not available, but I assume there was a good reason for this name. What was that?

Plus, I've added support for counter-* rules and matching functions.

[giraud]

Sorry, I had little time recently to work on bs-css. I'll review that pr in the next days

[giraud]

If we eventually agree to break the API, does it make sense to rename contentRule to content? (and contentRules to contents I suppose). I was quite surprised that content function is not available, but I assume there was a good reason for this name. What was that?

There was - unfortunately - already a content alias in the global namespace. contentRule was used to introduce that rule and not break existing code.

[erykpiast]

If we eventually agree to break the API, does it make sense to rename contentRule to content? (and contentRules to contents I suppose). I was quite surprised that content function is not available, but I assume there was a good reason for this name. What was that?

There was - unfortunately - already a content alias in the global namespace. contentRule was used to introduce that rule and not break existing code.

Alright, that makes sense. I can see there is an alias for a polymorphic variant. 👍

[giraud]

if possible, I'd like to inline the string_of_content

[erykpiast]

This way there will be duplication as I use the function in contentRules as well. What do you propose then?

[erykpiast]

Maybe I could put these functions (together with string_of_counter_*) into modules inside AtomicTypes? I can see you're avoiding this so far (like string_of_background), what was the reason of that?

[giraud]

just string_of_content. Css.re is already huge, I'm trying to not add useless code. In that case, creating a function doensn't add anything. For operations, it's needed because it's used in multiple locations

[erykpiast]

string_of_content is used twice as well. Am I missing something?

[erykpiast]

Thank you so much for such a detailed comments, @giraud!

[giraud]

Thank you

[erykpiast]

@giraud Hey! When can I expect a new version in NPM registry? :)