Re: [csswg-drafts] [css-syntax] Review requested of new Parsing text (#8834)

> I think my point was that I find a bit inconsistent that the algorithms for consuming rules no longer process one level at a time, leaving lower levels as a list of component values.

The *only* reason the old spec did so was because I didn't have a unified parsing model for all possible block contents. You had to *know* whether to parse a block's contents as "at-rules and declarations" or "at-rules and qualified rules", and that depended on what type of rule the block was in; it was easier to write the spec to have it delay this decision until later, when doing grammar-checking.

The new rules don't actually require having any grammar knowledge, because we've baked some restrictions into the grammar itself, so I *was* able to unify the two algorithms into one. (Knowing the grammar just lets you parse more efficiently; it doesn't change the end result.) So there was no longer any need to delay the parsing, and we could just get everything done in one pass.

> But it expects a list of component values, as it is only intended for an independent block, eg. style.cssText. It cannot handle a block resulting from parse a rule.

Correct, but I'm not sure why that's relevant.  The "parse" algorithms are the high-level algos called by other specs, while the "consume" algorithms are the low-level things that actually do the parsing. "Parse" algorithms call into "consume" ones; not vice versa.

> so there may no longer be a need for consume a declaration, consume an at-rule, consume a qualified rule, to require that the declaration or rule be valid in the current context.

*In theory*, you can indeed remove the check from all of these places. You could just check, in "consume a block's contents", whether the returned declaration contains an improperly-used {}-block, and then do all the rest of the validity checking after parsing. But that wouldn't be very efficient; every top-level rule that starts like `div:hover` would consume *the entire rest of the stylesheet* as a declaration, then the parser would realize it contains a {}-block in the wrong place, and re-parse a single rule from it. Then do the same thing again, and again, and again. The end result would be identical to one that checks validity as it goes, as the spec currently says, it would just be massively less efficient.

So, it's better to write the spec in a way that guides implementations to do the better thing. This also ensures that if we *did* accidentally make a change that would cause the "check validity now" and "check validity later" impls to diverge, the "check validity now" impls would stay correct; the spec wouldn't be accidentally requiring impls to change to the less efficient method.

-- 
GitHub Notification of comment by tabatkins
Please view or discuss this issue at https://github.com/w3c/csswg-drafts/issues/8834#issuecomment-1877730084 using your GitHub account


-- 
Sent via github-notify-ml as configured in https://github.com/w3c/github-notify-ml-config

Received on Thursday, 4 January 2024 20:38:46 UTC