Instead of trying to be clever and detaching the paint tree lazily,
just detach all paintables from both DOM and layout tree when building
and committing respectively.
Switch over from deprecated_attribute to attribute, and reuse the result
of this lookup for both checking whether that attribute exists - and for
retrieving the value of that attribute when we need it.
Rather than resolving the text-shadow each time painting commands are
recorded, we can resolve it once during the layout commit and save the
resolved values in paintable fragments. This is also step towards
getting rid of layout node pointer in paintable fragment.
If the entire document is invalidated, we know a full style update is
coming soon, and there's no need to try and invalidate a smaller part.
This avoids a *lot* of work on some pages. As an example, we are able to
skip ~1.5 million style invalidations on https://html.spec.whatwg.org/
This is a part of refactoring towards making the paintable tree
independent of the layout tree. Now, instead of transferring text
fragments from the layout tree to the paintable tree during the layout
commit phase, we allocate separate PaintableFragments that contain only
the information necessary for painting. Doing this also allows us to
get rid LineBoxes, as they are used only during layout.
Use the [FlyString] extended attribute to allow these functions to take
an Optional<FlyString> directly, allowing us to tidy up some conversions
from Optional<String>.
`JsonValue::to_byte_string` has peculiar type-erasure semantics which is
not usually intended. Unfortunately, it also has a very stereotypical
name which does not warn about unexpected behavior. So let's prefix it
with `deprecated_` to make new code use `as_string` if it just wants to
get string value or `serialized<StringBuilder>` if it needs to do proper
serialization.
A bunch of users used consume_specific with a constant ByteString
literal, which can be replaced by an allocation-free StringView literal.
The generic consume_while overload gains a requires clause so that
consume_specific("abc") causes a more understandable and actionable
error.
Instead of spawning these processes from the WebContent process, we now
create them in the Browser chrome.
Part 1/N of "all processes are owned by the chrome".
This patch makes a few changes to the way we calculate line-height:
- `line-height: normal` is now resolved using metrics from the used
font (specifically, round(A + D + lineGap)).
- `line-height: calc(...)` is now resolved at style compute time.
- `line-height` values are now absolutized at style compute time.
As a consequence of the above, we no longer need to walk the DOM
ancestor chain looking for line-heights during style computation.
Instead, values are inherited, resolved and absolutized locally.
This is not only much faster, but also makes our line-height metrics
match those of other engines like Gecko and Blink.
Fetching the viewport rect is currently somewhat expensive, since it
requires finding the navigable the document is active in.
We can avoid the cost of repeated calls by simply allowing StyleComputer
to cache the viewport rect at the start of style computation.
A bit faster:
```
N Min Max Median Avg Stddev
x 50 0.97179127 1.0031381 0.98313618 0.98407591 0.0092019442
+ 50 0.95996714 0.99507213 0.96965885 0.97242294 0.0095455053
Difference at 95.0% confidence
-0.011653 +/- 0.00372012
-1.18415% +/- 0.378032%
(Student's t, pooled s = 0.0093753)
```
All ColorSpace subclasses converted to float anyways, and this
allows us to save lots of float->Value->float conversions during
image color space processing.
A bit faster:
```
N Min Max Median Avg Stddev
x 50 0.99054313 1.0412271 0.99933481 1.0052408 0.012931916
+ 50 0.97073889 1.0075941 0.97849107 0.98184034 0.0090329046
Difference at 95.0% confidence
-0.0234004 +/- 0.00442595
-2.32785% +/- 0.440287%
(Student's t, pooled s = 0.0111541)
```
According to ministat, a bit faster to render page 3 of 0000849.pdf:
```
N Min Max Median Avg Stddev
x 50 1.000875 1.0427601 1.0208509 1.0201902 0.01066116
+ 50 0.99707389 1.03614 1.0084391 1.0107864 0.010002724
Difference at 95.0% confidence
-0.00940384 +/- 0.0041018
-0.921773% +/- 0.402062%
(Student's t, pooled s = 0.0103372)
```
Microoptimization; according to ministat a bit faster:
```
N Min Max Median Avg Stddev
x 50 1.0179932 1.0561159 1.0315337 1.0333617 0.0094757426
+ 50 1.000875 1.0427601 1.0208509 1.0201902 0.01066116
Difference at 95.0% confidence
-0.0131715 +/- 0.00400208
-1.27463% +/- 0.387287%
(Student's t, pooled s = 0.0100859)
```
This is a normative change in the ECMA-262 spec. See:
22de374
The issue noted here has been fixed in the same way that we previously
worked around it. Update the spec notes to match.
We have two known PlatformObjects that need to implement some of the
behavior of LegacyPlatformObjects to date: Window, and HTMLFormElement.
To make this not require double (or virtual) inheritance of
PlatformObject, move the behavior of LegacyPlatformObject into
PlatformObject. The selection of LegacyPlatformObject behavior is done
with a new bitfield of feature flags instead of a dozen virtual
functions that return bool. This change simplifies every class involved
in the diff with the notable exception of Window, which now needs some
ugly const casts to implement named property access.
Without a key function, the vtable for this class can be emitted into
every shared object or executable that needs it. This can cause bugs and
bad behavior when trying to access the vtable or RTTI for the class.
This is most easily seen when trying to call ``is<JS::Date>``, which
currently will do a dynamic_cast. Based on compiler, linker and loader
choices about ordering, it's possible that the code checking the RTTI
and the code that created the object could have a different vtable and
type_info in mind, causing false negatives for the ``is`` check.
Reduces time spent rendering page 3 of 0000849.pdf from 1.32s to 1.13s
on my machine.
Also reduces the time to run Meta/test_pdf.py on 0000.zip
(without 0000849.pdf) from 56s to 54s.
Reduces time spent rendering page 3 of 0000849.pdf from 1.45s to 1.32s
on my machine.
Also reduces the time to run Meta/test_pdf.py on 0000.zip
(without 0000849.pdf) from 58s to 55s.
Reduces time spent rendering page 3 of 0000849.pdf from 1.85s to 1.45s
on my machine.
As determined by running
time Build/lagom/bin/pdf --page 3 --render out.png \
~/Downloads/0000/0000849.pdf
a few times and eyeballing the min time.
Also reduces the time to run Meta/test_pdf.py on 0000.zip
(without 0000849.pdf) from 1m7s to 58s.
