[css-values-4] mix() can not represent retargeted transitions

[andruud]

Say you are transitioning from A to B, and then in the middle of that transition, you retarget to C. If the interpolated values can only be represented with mix(), then we'd have mix(mix(A; B; x%); C; y%), which isn't valid.

Can we drop the weird "must be the sole value" and actually give mix() a type?

Note: See w3c/css-houdini-drafts#425 for a real example of something that can be transitioned, but not represented.

[Loirooriol]

I think that these functions which accept a list of values and behave as one of them should allow the individual values to be things that would normally require to be the sole value.

[tabatkins]

Yes, I haven't gotten around to specifying this correctly, but it really should be written as something like <top-level-value>, which is only allowed in places that explicitly allow it (such as in mix() or as the sole value of a property.

[tabatkins]

Okay, defined <whole-value> and switched mix() and toggle() to using it.

[Loirooriol]

@tabatkins You forgot to replace some <declaration-value>:

csswg-drafts/css-values-5/Overview.bs

Line 419 in 46b0c7d

interpolates between the first <<declaration-value>> (its [=mix start value=])

csswg-drafts/css-values-5/Overview.bs

Line 420 in 46b0c7d

and the second <<declaration-value>> (its [=mix end value=]).

csswg-drafts/css-values-5/Overview.bs

Line 430 in 46b0c7d

if the two <<declaration-value>>s being interpolated by ''mix()''

csswg-drafts/css-values-5/Overview.bs

Line 441 in 46b0c7d

and its <<declaration-value>>s (if provided)

[tabatkins]

Ah thanks, fixed.