[css-values-5] should 'calc-size(0)' work?

[dbaron]

In https://issues.chromium.org/341564374 @yisibl pointed out that calc-size(0) doesn't work. I think this matches the current spec which defines the behavior to work like calc(), where it was intentional that calc(0) doesn't work.

I think despite the parallels with calc() we may want to consider making calc-size(0) work correctly, because:

• calc-size() is explicitly for lengths only so there's less concern about unit algebra making sense
• it would be nice to preserve the principle that "wrap an animation endpoint in calc-size()" works in general, and 0 is one of the most common length animation endpoints.

[Loirooriol]

To clarify, this is about unitless lengths. calc(0) works fine as an <integer>.

[tabatkins]

Yes, I think allowing a bare 0 is fine; as you say, we have the necessary context to infer the type (the calculation must be a length).

We probably want to do it as a grammar branch, so only a literal 0 with nothing else works. (Such a calculation would otherwise be invalid since it would type as <number>, so this is safe.) If the calculation is any more complex than that, we continue to treat it as a number.

[dbaron]

I think now that we've dropped the single-argument form in #10478 (and recommended interpolate-size as the way to opt in to animations/transitions rather than wrapping values with calc-size()), we should probably just keep the current spec for this one and keep calc-size() consistent with calc().

[tabatkins]

Yeah, this is just covered by the existing rules now that the arguments aren't mingled any longer - calc-size(0, ...) is valid (equivalent to 0px, as it's a <length-percentage>), while calc-size(..., 0) isn't legal (it's a calculation, so a literal zero is taken as a <number>, so it resolves to the wrong type).

[dbaron]

With the current spec I think 0 is valid in neither part, since the basis is defined as [ <intrinsic-size-keyword> | percentage | <calc-size()> | any | <calc-sum> ]. (That's also what Chrome implements -- 0 is valid for neither part.)

@tabatkins do you think that should change?

[tabatkins]

Ah, you're right, we allow a calculation there currently.

I think we should either disallow bare 0 (as the current spec says) or change the basic from calc-sum to length-percentage (thus requiring a calc() if you want to mix units). I lean towards leaving the spec as-is, so both arguments of the function act like calc() as much as possible, since the function's name is calc-size().