[css-typed-om] Should a type have a null percent hint to match `<number>`?

[cdoublev]

Before w3c/css-houdini-drafts@808c87a, a type always had to have a null percent hint to match <number>.

Now:

If the context in which the value is used does not allow <percentage> values, then the type must additionally have a null percent hint to be considered matching.

opacity: calc((1% + 1px) / 1px) is now valid if I am not mistaken, which seems unexpected because what is the resolved value of 1%?

• the context allows <number> and <percentage> (resolved to <number>)
• 1% has type [percent → 1]
• 1% + 1px has type [length → 1] with a length percent hint
• (1% + 1px) / 1px has type [length → 0] with a length percent hint
• [length → 0] with a length percent hint matches <number>

[cdoublev]

However, based from #10017 (comment):

things like height: calc(sqrt(50% / 1px) * 1px) should work

So to match <number>, maybe the percent hint should match the type that <percentage> resolves to, if the type has a percent hint.

---

Suggested text:

A type is said to match a CSS production in some circumstances:

• A type matches <length> if its only non-zero entry is «[ "length" → 1 ]». Similarly for <angle>, <time>, <frequency>, <resolution>, <flex>, and <percentage>.
• A type matches <number> if it has no non-zero entries.

If the type has a percent hint, its value must match the type that <percentage> resolves to in the context.

[tabatkins]

Ah, indeed, matching number needs some more nuance there.

[tabatkins]

Okay, tweaked the definitions a little in w3c/css-houdini-drafts@d384226 to pay more attention to the percent hint, when you have context for the value's use.

Your 'opacity' example is now properly invalid, as 'opacity' does not resolve percentages against any other type, so it requires the calculation to have a null percent hint.

[cdoublev]

Thanks!

I am aware that the algorithm has to account for inputs that cannot be produced in a style sheet but with handmade CSSNumericValues. For example, a type [length → 1] with angle percent hint matching <length> in a context defined with <angle-percentage>.

That said, would you say that all context/percentage conditions can be simplified to "if the type has a percent hint, this percent hint must match the type that <percentage> resolves to in the context"?

[tabatkins]

I think so, yes.