Instead of defining somewhat high level mouse actions, allow granular
control of mouse clicks and mouse down/up/move events. We will want to
simulate things like holding down a mouse button after double clicking
and then dragging the mouse to another position in the future, and this
enables that.
Instead of doing a full document style invalidation when a stylesheet is
dynamically added, we now analyze the new sheet's selectors to determine
which elements could potentially be affected, and only invalidate those.
This works by building an InvalidationSet from the rightmost compound
selector (the "subject") of each rule in the new stylesheet, extracting
class, ID, tag name, attribute, and pseudo-class features. We then walk
the DOM tree and only mark elements matching those features as needing a
style update.
If any selector has a rightmost compound that is purely universal (no
identifying features), or uses a pseudo-class not supported by the
invalidation set matching logic, we fall back to full invalidation.
When a pseudo-class state changed, we always walked the entire
document (or shadow root) tree to find affected elements, even
though only the subtree rooted at the old/new common ancestor
can be affected.
Narrow the tree walk to start from old_new_common_ancestor
instead of the root. To ensure ancestor-dependent selectors are
still correctly evaluated, we seed the style computer's ancestor
filter by walking up from the common ancestor to the root before
the invalidation walk.
This reduces the work from O(total elements) to
O(subtree elements) + O(tree depth), which is a large improvement
on pages where pseudo-class changes (hover, focus, active, target)
occur deep in the DOM.
This was extremely hot (10%+) when hovering mailboxes on GMail.
When a parent element's display property changes (e.g., to flex or
grid), children may need to be blockified or un-blockified.
Previously, children only received a recompute_inherited_style() call
which doesn't run the blockification logic.
This patch adds a parent_display_changed flag to the recursive style
update that forces children to get a full style recompute when their
parent's display change triggers a layout tree rebuild.
The main change here is that we now properly absolutize values which
means we now support `random()` and `sibling-{count,index}()`
We are also more consistent with how we handle computation for the other
font properties
Previously we would just set the attributes to the serialized
descriptors, even if they were the empty string.
We now apply defaults when we have empty descriptors and apply parsing
logic from the various `set_*` methods (only applicable to `font-family`
so far where we now extract the value from either a string or a
custom-ident)
Fixes an issue in some css/css-shapes WPT tests where we weren't
properly matching fonts.
Previously we would always try to load the font URL relative to the
document's base URL. This commit means that for CSS-connected
`FontFace`s we now try to load relative to the URL of the stylesheet
that contains the associated `CSSFontFaceRule`
The spec states that updates to descriptors in a `@font-face` rule
should be reflected in the connected `FontFace`'s attributes.
Previously this was achieved by directly accessing those descriptors
when calling the attribute getters, but this has a couple of issues:
a) The changes are only reflected if we use the accessors (i.e.
`FontFace::family()` rather than `FontFace::m_family`) which isn't
the case everywhere
b) The changes aren't persisted after the `FontFace` is disconnected
from it's `CSSFontFaceRule`
To fix these issues we now reparse and store the `FontFace`'s attributes
whenever the `CSSFontFaceRule`'s descriptors change.
Previously we would run it once within `set_src()` and then again within
`set_property()` which would lead to us creating two new `FontFace`s
instead of just one
Previously, we registered `@property` rules during parsing, and treated
them the same as `CSS.registerProperty()` calls. This is not correct
for a couple of reasons: One, the spec wants us to distinguish between
those two sources of registered custom properties, with
`CSS.registerProperty()` calls taking precedence. Two, we never removed
the registered property when its `@property` was removed from the
document.
This commit deals with this by iterating active CSSPropertyRules to find
which ones currently apply, and storing those in a cache. This cache is
invalidated whenever the Document's style is invalidated, which happens
whenever a CSSRule is added or removed from the Document.
The attached test demonstrates this now working as it should.
The `:has()` pseudo-class requires traversing descendants (or siblings)
to find matches.
With this change we cache results keyed by `(Selector*, Element*)`
pairs. The cache is stored in `StyleComputer` and cleared at the start
of each style computation pass in `Document::update_style()`.
When `:has()` uses a descendant combinator and we find a match, we also
cache that all ancestors between the matching descendant and the
anchor match. For example with `div:has(.target)`:
```html
<div id="A"> <!-- checking :has(.target) here -->
<div id="B">
<div id="C">
<span class="target"/>
</div>
</div>
</div>
```
When we find `.target` while checking `div#A`, we also cache that
`div#B` and `div#C` match `:has(.target)` since they also contain
`.target`. Later when styling these elements, we get cache hits and skip
traversal.
`EdgeStyleValues` which consist of an offset of a `calc()`s which
resolves to 50% should be considered "centered" for
`SerializationMode::ResolvedValue` for the purpose of omitting the
position value from gradient serialization.
When we invoke Tokenizer::consume_a_url_token, we have already consumed
the "url(" text from the source. Thus we would set the original source
text to the text starting just after those consumed code points. Let's
instead pass in the known starting position from the caller.
This fixes a bug seen on https://lichess.org, where they perform a
`substring(4)` on the property value to remove the "url(" text. This
would strip away the "http" part of the URL, and we would try to load
"s://lichess.org/image.svg". With this fixed, we can play chess games
on this site.
If a script on the page cancels a mousemove event, we would return early
and neglect to update the cursor. This is seen regularly on "Diablo Web"
where they set the cursor to "none" on the main canvas, and also cancel
mousemove events.
Previously we were doing a couple things wrong:
- Using the cascaded rather than computed value (so we didn't support
CSS-wide keywords)
- Only supporting the case where we had one animation-play-state
Previously we would only update these if:
a) We had a cascaded value for `transition-property`
b) The source of that cascaded value had changed since we last
registered transitions
This meant that there were a lot of changes we didn't apply:
- Changes exclusively to properties other than `transition-property`
(e.g. `transition-duration`, `transition-behavior`, etc)
- Removing the `transition-property` property
- Updating the `transition-property` property in a way that didn't
change it's source (e.g. setting it within inline-style)
Unfortunately this does mean that we now register transitions for all
properties on most elements since "all" is the initial value for
"transition-property" which isn't great for performance, but that can be
looked at in later commits.
Also renames the `clear_transitions` function to clarify this doesn't
affect the associated transition animations.
This fixes an issue where transitions weren't being cancelled when the
relevant transition-property entry was no longer present
This updates the `parse_text_decoration_line_value` to reject values
which non-exclusively include `none` e.g. `underline none`.
It also simplifies handling by always producing a Vector (except for
`none`) and adding VERIFY_NOT_REACHED in more places which shouldn't be
reachable.
We were doing this manually within `Document::update_layout()` and
`CSSStyleProperties::get_direct_property()` but we should do it for all
callers of `Document::update_style()`
Unrecognized property names should still be kept in the list to preserve
matching of indices (e.g. that the Nth property should associate with
the Nth duration)
This excludes `step-end` and `step-start` which are expected to be
converted to the equivalent function at parse time.
We are expected to serialize these as the explicit keywords - previously
we would parse as `EasingStyleValue` and serialize equivalent functions
as the keywords. This caused issues as we would incorrectly serialize
even explicit functions as the keyword.
This also allows us to move the magic easing functions to
`EasingFunction` rather than `EasingStyleValue` which is a bit tidier
We now:
- Serialize longhands in the correct order
- Support serializing multiple values
- Include default longhands where required (to distinguish
animation-name from that longhand).
