feature-request: allow specificity calculation of AST's of popular parsers

[bartveneman]

TIL that this semi-copy exists (14M weekly downloads) for the sole reason that PostCSS AST's can be used to calculate specificity instead of CSSTree. Unfortunately I ran into the same issue for Project Wallace where I re-implemented specificity to cater for @projectwallace/css-parser's AST.

It'd be nice to have a more generic calculator, so I want to use this issue for some bikeshedding and gathering ideas. I'm going to make some proposals as a starting point:

calculate(selector: string): Specificity
// which calls CSSTree by default, as it does now
import type { Selector } from 'csstree'
calculateForAST(node: Selector): Specificity

// But then we could also expose

import type { Node } from 'postcss'
calculatePostCSS(node): Specificity

import type { Selector } from '@projectwallace/css-parser'
calculateProjectWallace(node: Selector): Specificity

The hard-to-design part here is to make one function that handles the specificity calculation, regardless of parser or node types:

// all above calculate() methods call this under the hood:

calculateGeneric()

[bramus]

Yes, this is something that has been on my mind for quite a while and apparently great minds think alike (or was it “fools seldom differ”? 😛).

I haven’t thought about specifics here, but some thoughts are:

• Being able to call calculate with any AST — instead of having separate functions for each — and have the package figure out which one it is would be a very nice feature
• This could be done by creating a bunch of conversion utilities that converts the CSSTree/Wallace/PostCSS ASTs to a generic format, upon which the calculate function would only accept this generic format.
  • Advantage is that we can drop the bundling of the csstree in the package, but a disadvantage is that we then would put the load on the user.
  • To retain the “calculate for a string” functionality, new packages can be created that bundle the core + CSSTree/Wallace/PostCSS, so we could plug it that way

[bartveneman]

I'll give this a shot, let's see what comes out of that. I already have a draft PR against my own fork to play around with. I think it's doable. Key is finding the right tradeoffs without blowing up bundle size. Glad you support the idea!