The spec has a general rule for this, which is roughly that "If it's not
a falsey value, it's true". However, a couple of media-features are
always false, apparently breaking this rule. To handle that, we have an
array of false keywords in the JSON, instead of a single keyword. For
those always-false media-features, we can enter all their values into
this array.
Gets us 2 more WPT subtest passes.
Previously, for `foo < 30px` ranges, we'd flip them and store them as
`30px > foo` instead. That worked fine, but means the serialization is
wrong. So instead, keep them in their original form.
I experimented with giving Range two optional sub-structs instead of 4
optional members, thinking it would be smaller - but it's actually
larger, because the two Optional<Comparison>s fit snugly together. So,
the slightly-goofy all-Optionals remains.
This gets us 2 WPT passes that I'm aware of.
Instead of rejecting invalid media-feature values at parse time, we are
expected to parse them, and then treat them as Unknown when evaluating
their media query. To implement this, we first try to parse a valid
value as before. If that fails, or we have any trailing tokens that
could be part of a value, we then scoop up all the possible
ComponentValues and treat that as an "unknown"-type MediaFeatureValue.
This gets us 66 WPT passes.
Functionally this is the same before, as result is always True or False
before this point, and `True && Foo` evaluates to `Foo`. But this is
more clearly correct, instead of correct by coincidence.
This commit on its own has no observable behaviour changes, as we still
only return True or False, but for the next commit, we'll need to be
able to return the Unknown state here, and without this change we'd get
regressions.
The current spec defines this simply as `<ident>`, but does apparently
serialize as lowercase.
Because of this change, we no longer need to care about the deprecated
media types, as they all behave the same as unknown ones.
We still keep an enum around for KnownMediaType, to avoid repeated
string comparisons when evaluating it.
Gets us 2 WPT passes.
A MediaFeature either has a MediaFeatureValue, or a Range, or nothing.
Combining these into a Variant reduces the size from 176 bytes to 128,
and also makes constructing these a little less awkward.
CSS Values 5 now defines a `<boolean-expr[]>` type that is used in place
of the bespoke grammar that previously existed for `@media` and
`@supports` queries. This commit implements some BooleanExpression
types to represent the nodes in a `<boolean-expr[]>`, and reimplements
`@media` and `@supports` queries using this.
The one part of this implementation I'm not convinced on is that the
`evaluate()` methods take a `HTML::Window*`. This is a compromise
because `@media` requires a Window, and `@supports` does not require
anything at all. As more users of `<boolean-expr[]>` get implemented in
the future, it will become clear if this is sufficient, or if we need
to do something smarter.
As a bonus, this actually improves our serialization of media queries!
To be properly compatible with calc(), the resolved() methods all need:
- A length resolution context
- To return an Optional, as the calculation might not be resolvable
A bonus of this is that we can get rid of the overloads of `resolved()`
as they now all behave the same way.
A downside is a scattering of `value_or()` wherever these are used. It
might be the case that all unresolvable calculations have been rejected
before this point, but I'm not confident, and so I'll leave it like
this for now.
The length resolution context might be needed even when resolving an
integer value, since it might contain a sign() function with length
values inside. This fixes a WPT subtest.
