Although we know that the animation properties will always be filled in for
a transition in the cases where we need to query them (going forward we will
have a situation where an animation may only have frames, not properties, but
that will only happen when the animation isn't attached to an element or the
element is not attached to a document, but we don't run animations in that case
and cancel existing ones when we enter that state so although they *can* enter
that state, we'll never run these methods on them when they do), we still want
to move towards making frames the primary unit for interacting with animation
values since frames always exist and represent the public interface.
Ultimately it would be good to make the properties array on
a KeyframeEffect(ReadOnly) an encapsulated detail so that we can freely change
their structure (e.g. segments might not be the best setup, it might be better
to just have arrays of free-standing values to avoid the duplication of
values when segments are continuous).
This patch removes or encapsulates a few references to properties and
simplifies the code at the same time.
MozReview-Commit-ID: 3II36SYVoRE
When we go to switch CSS Animations over to using
KeyframeEffectReadOnly::SetFrames we will need a way to represent any filled-in
from/to values as nsCSSValue objects. These objects are built from the current
computed style. We currently use StyleAnimationValue::ExtractComputedValue for
this which returns a StyleAnimationValue. In order to convert this to an
nsCSSValue we can use StyleAnimationValue::UncomputeValue. However, in some
cases, the nsCSSValue objects returned by that method are dependent on the
passed-in StyleAnimationValue object.
This patch adds an overload to UncomputeValue that takes an rvalue
StyleAnimationValue reference and produces an nsCSSValue that is independent
of the StyleAnimationValue through a combination of copying data and
transferring ownership of data.
This patch also adjusts the return value for the case of filter and shadow
lists when the list is empty so that we return a none value in this case.
These are the only list types which are allowed to have a null list value.
Not only does this produce the correct result when these values are serialized
(the initial value for 'filter', 'text-shadow', and 'box-shadow' is 'none') it
also means that UncomputeValue should never return an nsCSSValue whose unit is
null which is important because when we later pass that value to BuildStyleRule
it will treat a null nsCSSValue as an error case (specifically, "longhand failed
to parse").
MozReview-Commit-ID: 4RoCn39ntiJ
In the next patch in this series, we would like to update the error handling of
the call to StyleAnimationValue::Interpolate in
KeyframeEffectReadOnly::ComposeStyle. Using AnimValuesStyleRule::AddEmptyValue
there, however, makes handling the error case difficult because we need a means
of clearing the allocated StyleAnimationValue.
However, simply using AnimationValuesStyleRule::AddValue means we will end up
doing needless allocations for StyleAnimationValue objects (the copy
constructor for which can result in performing potentially expensive heap
allocations, such as when lists are deep-copied).
Instead, we add a Move constructor to StyleAnimationValue and add an overload
of AnimValuesStyleRule::AddValue that takes an rvalue reference. This
provides a more consistent interface to AnimValuesStyleRule and avoids the
unnecessary allocations from copying StyleAnimationValue objects.
MozReview-Commit-ID: CaP1uPAgNnm
The view is not guaranteed to have a frame (and won't during the early parts of the presshell's existence). It will have a view manager, and that will have a presshell during this time period.
Anytime a view has a frame it will also have a view manager and a pressshell so this is strictly better.
