This currently only applies to property-level calculation contexts, more
work to be done to generate accepted ranges for other calculation
contexts (e.g. within transformation functions, color functions, etc)
This reverts 0e3487b9ab.
Back when I made that change, I thought we could make our StyleValue
classes match the typed-om definitions directly. However, they have
different requirements. Typed-om types need to be mutable and GCed,
whereas StyleValues are immutable and ideally wouldn't require a JS VM.
While I was already making such a cataclysmic change, I've moved it into
the StyleValues directory, because it *not* being there has bothered me
for a long time. 😅
- Omit calcs that are resolved to `0px` from the serialized value
- Allow CSV to be the 'Z' component in interpolated value.
- Allow calcs with mixed percentages in the first two arguments.
To achieve the third item above the concept of a "special" value parsing
context has been added - this will also be useful for instance for
different arguments of color functions having different contexts.
Gains us 23 WPT tests
These will be used for the mask-repeat property as well in an upcoming
commit, hence the more generic names. Also, this more closely matches
the names used in the spec.
This parses `anchor-size(..)` functions in CSS, but does not yet result
in a useful `Size`: we need style & layout interleaving similar to
container queries for this, since the resulting value depends on layout
results.
Not supported yet: `anchor-size()` appearing inside a `calc()` node.
Adds 4280 WPT subtest passes in `css/css-anchor-position`.
Before this change, whenever element's attributes changed, we would add
a flag to "pending invalidation", indicating that all descendants whose
style uses CSS custom properties needed to be recomputed. This resulted
in severe overinvalidation, because we would run invalidation regardless
of whether any custom property on affected element actually changed.
This change takes another approach, and now we decide whether
descendant's style needs to be recomputed based on whether ancestor's
style recomputation results in a change of custom properties, though
this approach adds a little overhead to style computation as now we have
to compare old vs new hashmap of custom properties.
This brings substantial improvement on discord and x.com where, before
this change, advantage of using invalidation sets was lost and we had
to recompute all descendants, because almost all of them use custom
properties.
Before this change, calc() would resolve to different types depending on
the nearest containing value context. This meant that rgb(calc(), ...)
by itself worked correctly due to fallbacks, but rgb(calc(), ...) inside
e.g a linear-gradient would create a calc() value that resolves to a
length, which subsequently got rejected by the color value parser.
Fixing this makes various little gradients show up on Discord.
UnresolvedStyleValue::create() has one user where we know if there are
any arbitrary substitution functions in the list of CVs, and two users
where we don't know and just hope there aren't any. I'm about to add
another user that also doesn't know, and so it seems worth just making
UnresolvedStyleValue::create() do that work instead.
We keep the parameter, now Optional<>, so that we save some redundant
work in that one place where we do already know.
The existing resolve methods are not to spec and we are working to
replace them with new ones based on the `simplify_a_calculation_tree`
method.
These are marked as deprecated rather than replaced outright as work
will need to be done on the caller side to be made compatible with the
new methods, for instance the new methods can fail to resolve (e.g.
if we are missing required context), where the existing methods will
always resolve (albeit sometimes with an incorrect value).
No functionality changes.
"Arbitrary substitution functions" are a family of functions that
includes var() and attr(). All of them resolve to an arbitrary set of
component values that are not known at parse-time, so they have to be
substituted at computed-value time.
Besides it being nice to follow the spec closely, this means we'll be
able to implement the others (such as `if()` and `inherit()`) more
easily.
The main omission here is the new "spread syntax", which can be
implemented in the future.
This has an extra parameter to allow stopping at the first comma token,
which we need for var() and attr()'s "argument grammar".
Co-authored-by: Tim Ledbetter <tim.ledbetter@ladybird.org>
Custom properties are required to produce a computed value just like
regular properties. The computed value is defined in the spec as
"specified value with variables substituted, or the guaranteed-invalid
value", though in reality all arbitrary substitution functions should be
substituted, not just `var()`.
To support this, we parse the CSS-wide keywords normally in custom
properties, instead of ignoring them. We don't yet handle all of them
properly, and because that will require us to cascade them like regular
properties. This is just enough to prevent regressions when implementing
ASFs.
Our output in this new test is not quite correct, because of the awkward
way we handle whitespace in property values - so it has 3 spaces in the
middle instead of 1, until that's fixed.
It's possible this computed-value production should go in
cascade_custom_properties(), but I had issues with that. Hopefully once
we start cascading custom properties properly, it'll be clearer how
this should all work.
Previously the type argument in attr() could be the name of a CSS type
on its own. This has changed, and now only `raw-string`
(previously `string`) or the name of a dimension unit is allowed. Other
types and more complex grammar use the `type()` function, which we
don't yet support.
I've updated the syntax comment, but not the algorithm itself, which
will be reimplemented in a later commit.
This brings parsing of grid-row-* and grid-column-* properties (and
their associated shorthands) more inline with spec.
Changes:
- Only set omitted properties for combined-value shorthands (e.g.
`grid-row: a` rather than `grid-row: a / b`) if the single value is
`<custom-ident>`.
- `[ [ <integer [-∞,-1]> | <integer [1,∞]> ] && <custom-ident>? ]`:
- Properly resolve `calc`s for `<integer>` that rely on compute-time
information.
- `[ span && [ <integer [1,∞]> || <custom-ident> ] ]`
- Allow `calc`s for `<integer>`
- Allow `<custom-ident>`
There is still work to be done to properly use these parsed values.
Gains us 46 WPT tests.
Our floating point number parser was based on the fast_float library:
https://github.com/fastfloat/fast_float
However, our implementation only supports 8-bit characters. To support
UTF-16, we will need to be able to convert char16_t-based strings to
numbers as well. This works out-of-the-box with fast_float.
We can also use fast_float for integer parsing.
Previously, we would allow calc values such as `calc(min(1 2))`, which
would be simplified to `calc(3)` because we assumed that numbers not
separated by an operator represented a sum. We now validate that the
number of operators we see is as we would expect before collecting
these values into a sum node.
Reimplements `grid`, `grid-template`, `grid-template-rows`, and
`grid-template-columns` in a way that uses a separate function for each
grammar rule defined in the specification. This change results in many
additional passing tests from the already imported WPT suite. Most of
the remaining test failures are related to incorrect serialization of
grid properties.
url() has some limitations because of allowing unquoted URLs as its
contents. For example, it can't use `var()`. To get around this, there's
an alternative `src()` function which behaves the same as `url()` except
that it is parsed as a regular function, which makes `var()` and friends
work properly.
There's no WPT test for this as far as I can tell, so I added our own.
A couple of differences from before:
- Only the fixed set of strings are allowed. Some formats can only be an
ident (eg, svg).
- We don't allow these foo-variations values in ident form.
- The comparison is done case-insensitively. It's unclear if this is
more or less correct, but as most things in CSS are insensitive,
including idents, it makes sense that these would be too.
"format(woff-variations)" and pals are supposed to expand like so:
"format(woff) tech(variations)".
However, since we don't support tech() yet, this patch just adds a small
hack where we still treat "woff-variations" as "woff" so that fonts
load and get used, even if we don't make use of the variations yet.
Previously, we would just assign the UnresolvedStyleValue to each
longhand, which was completely wrong but happened to work if it was a
ShorthandStyleValue (because that's basically a list of "set property X
to Y", and doesn't care which property it's the value of).
For example, the included `var-in-margin-shorthand.html` test would:
1. Set `margin-top` to `var(--a) 10px`
2. Resolve it to `margin-top: 5px 10px`
3. Reject that as invalid
What now happens is:
1. Set `margin-top` to a PendingSubstitutionValue
2. Resolve `margin` to `5px 10px`
3. Expand that out into its longhands
4. `margin-top` is `5px` 🎉
In order to support this, `for_each_property_expanding_shorthands()` now
runs the callback for the shorthand too if it's an unresolved or
pending-substitution value. This is so that we can store those in the
CascadedProperties until they can be resolved - otherwise, by the time
we want to resolve them, we don't have them any more.
`cascade_declarations()` has an unfortunate hack: it tracks, for each
declaration, which properties have already been given values, so that
it can avoid overwriting an actual value with a pending one. This is
necessary because of the unfortunate way that CSSStyleProperties holds
expanded longhands, and not just the original declarations. The spec
disagrees with itself about this, but we do need to do that expansion
for `element.style` to work correctly. This HashTable is unfortunate
but it does solve the problem until a better solution can be found.
ParsedFontFace and FontLoader now both keep track of which
CSSStyleSheet (if any) was the source of the font-face, so the URLs can
be completed correctly.
Math functions like abs(), clamp(), round(), etc, can be used by
themselves in property values, without wrapping them in calc().
Before this change, we were neglecting to run calc simplification on the
generated calculation node trees. By doing that manually after parsing a
standalone math function, we score at least a couple hundred WPT points.
