gecko-dev/layout/base/FrameProperties.h

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

436 lines
15 KiB
C
Raw Normal View History

/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
2012-05-21 11:12:37 +00:00
* License, v. 2.0. If a copy of the MPL was not distributed with this
* file, You can obtain one at http://mozilla.org/MPL/2.0/. */
#ifndef FRAMEPROPERTIES_H_
#define FRAMEPROPERTIES_H_
#include "mozilla/DebugOnly.h"
#include "mozilla/MemoryReporting.h"
#include "mozilla/Unused.h"
#include "nsTArray.h"
#include "nsThreadUtils.h"
class nsIFrame;
namespace mozilla {
struct FramePropertyDescriptorUntyped {
/**
* mDestructor will be called if it's non-null.
*/
typedef void UntypedDestructor(void* aPropertyValue);
UntypedDestructor* mDestructor;
/**
* mDestructorWithFrame will be called if it's non-null and mDestructor
* is null. WARNING: The frame passed to mDestructorWithFrame may
* be a dangling frame pointer, if this is being called during
* presshell teardown. Do not use it except to compare against
* other frame pointers. No frame will have been allocated with
* the same address yet.
*/
typedef void UntypedDestructorWithFrame(const nsIFrame* aFrame,
void* aPropertyValue);
UntypedDestructorWithFrame* mDestructorWithFrame;
/**
* mDestructor and mDestructorWithFrame may both be null, in which case
* no value destruction is a no-op.
*/
protected:
/**
* At most one destructor should be passed in. In general, you should
* just use the static function FramePropertyDescriptor::New* below
* instead of using this constructor directly.
*/
constexpr FramePropertyDescriptorUntyped(
UntypedDestructor* aDtor, UntypedDestructorWithFrame* aDtorWithFrame)
: mDestructor(aDtor), mDestructorWithFrame(aDtorWithFrame) {}
};
/**
* A pointer to a FramePropertyDescriptor serves as a unique property ID.
* The FramePropertyDescriptor stores metadata about the property.
* Currently the only metadata is a destructor function. The destructor
* function is called on property values when they are overwritten or
* deleted.
*
* To use this class, declare a global (i.e., file, class or function-scope
* static member) FramePropertyDescriptor and pass its address as
* aProperty in the FrameProperties methods.
*/
template <typename T>
struct FramePropertyDescriptor : public FramePropertyDescriptorUntyped {
typedef void Destructor(T* aPropertyValue);
typedef void DestructorWithFrame(const nsIFrame* aFrame, T* aPropertyValue);
template <Destructor Dtor>
static constexpr const FramePropertyDescriptor<T> NewWithDestructor() {
return {Destruct<Dtor>, nullptr};
}
template <DestructorWithFrame Dtor>
static constexpr const FramePropertyDescriptor<T>
NewWithDestructorWithFrame() {
return {nullptr, DestructWithFrame<Dtor>};
}
static constexpr const FramePropertyDescriptor<T> NewWithoutDestructor() {
return {nullptr, nullptr};
}
private:
constexpr FramePropertyDescriptor(UntypedDestructor* aDtor,
UntypedDestructorWithFrame* aDtorWithFrame)
: FramePropertyDescriptorUntyped(aDtor, aDtorWithFrame) {}
template <Destructor Dtor>
static void Destruct(void* aPropertyValue) {
Dtor(static_cast<T*>(aPropertyValue));
}
template <DestructorWithFrame Dtor>
static void DestructWithFrame(const nsIFrame* aFrame, void* aPropertyValue) {
Dtor(aFrame, static_cast<T*>(aPropertyValue));
}
};
// SmallValueHolder<T> is a placeholder intended to be used as template
// argument of FramePropertyDescriptor for types which can fit directly into our
// internal value slot (i.e. types that can fit in 64 bits). This class should
// never be defined, so that we won't use it for unexpected purpose by mistake.
template <typename T>
class SmallValueHolder;
namespace detail {
template <typename T>
struct FramePropertyTypeHelper {
typedef T* Type;
};
template <typename T>
struct FramePropertyTypeHelper<SmallValueHolder<T>> {
typedef T Type;
};
} // namespace detail
/**
* The FrameProperties class is optimized for storing 0 or 1 properties on
* a given frame. Storing very large numbers of properties on a single
* frame will not be efficient.
*/
class FrameProperties {
public:
template <typename T>
using Descriptor = const FramePropertyDescriptor<T>*;
using UntypedDescriptor = const FramePropertyDescriptorUntyped*;
template <typename T>
using PropertyType = typename detail::FramePropertyTypeHelper<T>::Type;
explicit FrameProperties() = default;
~FrameProperties() {
MOZ_ASSERT(mProperties.Length() == 0, "forgot to delete properties");
}
/**
* Return true if we have no properties, otherwise return false.
*/
bool IsEmpty() const { return mProperties.IsEmpty(); }
/**
* Set a property value. This requires a linear search through
* the properties of the frame. Any existing value for the property
* is destroyed.
*/
template <typename T>
void Set(Descriptor<T> aProperty, PropertyType<T> aValue,
const nsIFrame* aFrame) {
uint64_t v = ReinterpretHelper<T>::ToInternalValue(aValue);
SetInternal(aProperty, v, aFrame);
}
/**
* Add a property value; the descriptor MUST NOT already be present.
*/
template <typename T>
void Add(Descriptor<T> aProperty, PropertyType<T> aValue) {
MOZ_ASSERT(!Has(aProperty), "duplicate frame property");
uint64_t v = ReinterpretHelper<T>::ToInternalValue(aValue);
AddInternal(aProperty, v);
}
/**
* @return true if @aProperty is set. This requires a linear search through
* the properties of the frame.
*
* In most cases, this shouldn't be used outside of assertions, because if
* you're doing a lookup anyway it would be far more efficient to call Get()
* or Take() and check the aFoundResult outparam to find out whether the
* property is set. Legitimate non-assertion uses include:
*
* - Checking if a frame property is set in cases where that's all we want
* to know (i.e., we don't intend to read the actual value or remove the
* property).
*
* - Calling Has() before Set() in cases where we don't want to overwrite
* an existing value for the frame property.
*/
template <typename T>
bool Has(Descriptor<T> aProperty) const {
Bug 1479996 - Combine nsTArray::IndexOf and element access into lambda-friendly functions - r=froydnj In many places, nsTArray::IndexOf is followed by accessing that element (hopefully with `Elements() + index`, which skips unnecessary bounds checks.) But this pattern introduces operations that could be avoided: - IndexOf converts the address of the found element into an index, - The caller must test for a special `NoIndex` value, - On success, accesses convert that index back into the original address. This patch introduces new 'ApplyIf...' functions that combine all these operations in a more efficient ensemble: If the sought element is found, it is passed by reference to a given callable object (usually a lambda); if not found, another callable is invoked. Depending on what they need, the first callable may take one of these parameter lists: (), (size_t), (maybe-const elem_type&), (size_t, maybe-const elem_type&). On top of removing the pointer->index->pointer operations in most cases, invoking callables directly from ApplyIf is safer, as the array is guaranteed to be untouched at this time. Also, being templates taking function objects, in most cases the compiler should be able to inline and optimize the search and its callables' code. This patch gives example uses in nsTArray::Contains, and in FrameProperties::GetInternal, SetInternal. And FrameProperties::Has now calls Contains, which is more efficient because the former code would compute an index (or NoIndex), and then convert that index to a bool; whereas the new code directly produces a bool from within the search algorithm. Differential Revision: https://phabricator.services.mozilla.com/D2758 --HG-- extra : moz-landing-system : lando
2018-08-28 22:04:09 +00:00
return mProperties.Contains(aProperty, PropertyComparator());
}
/**
* Get a property value. This requires a linear search through
* the properties of the frame. If the frame has no such property,
* returns zero-filled result, which means null for pointers and
* zero for integers and floating point types.
* @param aFoundResult if non-null, receives a value 'true' iff
* the frame has a value for the property. This lets callers
* disambiguate a null result, which can mean 'no such property' or
* 'property value is null'.
*/
template <typename T>
PropertyType<T> Get(Descriptor<T> aProperty,
bool* aFoundResult = nullptr) const {
uint64_t v = GetInternal(aProperty, aFoundResult);
return ReinterpretHelper<T>::FromInternalValue(v);
}
/**
* Remove a property value, and return it without destroying it.
*
* This requires a linear search through the properties of the frame.
* If the frame has no such property, returns zero-filled result, which means
* null for pointers and zero for integers and floating point types.
* @param aFoundResult if non-null, receives a value 'true' iff
* the frame had a value for the property. This lets callers
* disambiguate a null result, which can mean 'no such property' or
* 'property value is null'.
*/
template <typename T>
PropertyType<T> Take(Descriptor<T> aProperty, bool* aFoundResult = nullptr) {
uint64_t v = TakeInternal(aProperty, aFoundResult);
return ReinterpretHelper<T>::FromInternalValue(v);
}
/**
* Remove and destroy a property value. This requires a linear search through
* the properties of the frame. If the frame has no such property, nothing
* happens.
*/
template <typename T>
void Remove(Descriptor<T> aProperty, const nsIFrame* aFrame) {
RemoveInternal(aProperty, aFrame);
}
/**
* Call @aFunction for each property or until @aFunction returns false.
*/
template <class F>
void ForEach(F aFunction) const {
#ifdef DEBUG
size_t len = mProperties.Length();
#endif
for (const auto& prop : mProperties) {
bool shouldContinue = aFunction(prop.mProperty, prop.mValue);
MOZ_ASSERT(len == mProperties.Length(),
"frame property list was modified by ForEach callback!");
if (!shouldContinue) {
return;
}
}
}
/**
* Remove and destroy all property values for the frame.
*/
void RemoveAll(const nsIFrame* aFrame) {
nsTArray<PropertyValue> toDelete = std::move(mProperties);
for (auto& prop : toDelete) {
prop.DestroyValueFor(aFrame);
}
MOZ_ASSERT(mProperties.IsEmpty(), "a property dtor added new properties");
}
size_t SizeOfExcludingThis(mozilla::MallocSizeOf aMallocSizeOf) const {
// We currently report only the shallow size of the mProperties array.
// As for the PropertyValue entries: we don't need to measure the mProperty
// field of because it always points to static memory, and we can't measure
// mValue because the type is opaque.
// XXX Can we do better, e.g. with a method on the descriptor?
return mProperties.ShallowSizeOfExcludingThis(aMallocSizeOf);
}
private:
// Prevent copying of FrameProperties; we should always return/pass around
// references to it, not copies!
FrameProperties(const FrameProperties&) = delete;
FrameProperties& operator=(const FrameProperties&) = delete;
inline void SetInternal(UntypedDescriptor aProperty, uint64_t aValue,
const nsIFrame* aFrame);
inline void AddInternal(UntypedDescriptor aProperty, uint64_t aValue);
inline uint64_t GetInternal(UntypedDescriptor aProperty,
bool* aFoundResult) const;
inline uint64_t TakeInternal(UntypedDescriptor aProperty, bool* aFoundResult);
inline void RemoveInternal(UntypedDescriptor aProperty,
const nsIFrame* aFrame);
template <typename T>
struct ReinterpretHelper {
static_assert(sizeof(PropertyType<T>) <= sizeof(uint64_t),
"size of the value must never be larger than 64 bits");
static uint64_t ToInternalValue(PropertyType<T> aValue) {
uint64_t v = 0;
memcpy(&v, &aValue, sizeof(aValue));
return v;
}
static PropertyType<T> FromInternalValue(uint64_t aInternalValue) {
PropertyType<T> value;
memcpy(&value, &aInternalValue, sizeof(value));
return value;
}
};
/**
* Stores a property descriptor/value pair.
*/
struct PropertyValue {
PropertyValue() : mProperty(nullptr), mValue(0) {}
PropertyValue(UntypedDescriptor aProperty, uint64_t aValue)
: mProperty(aProperty), mValue(aValue) {}
// NOTE: This function converts our internal 64-bit-integer representation
// to a pointer-type representation. This is lossy on 32-bit systems, but it
// should be fine, as long as we *only* do this in cases where we're sure
// that the stored property-value is in fact a pointer. And we should have
// that assurance, since only pointer-typed frame properties are expected to
// have a destructor
void DestroyValueFor(const nsIFrame* aFrame) {
if (mProperty->mDestructor) {
mProperty->mDestructor(
ReinterpretHelper<void*>::FromInternalValue(mValue));
} else if (mProperty->mDestructorWithFrame) {
mProperty->mDestructorWithFrame(
aFrame, ReinterpretHelper<void*>::FromInternalValue(mValue));
}
}
UntypedDescriptor mProperty;
uint64_t mValue;
};
/**
* Used with an array of PropertyValues to allow lookups that compare
* only on the FramePropertyDescriptor.
*/
class PropertyComparator {
public:
bool Equals(const PropertyValue& a, const PropertyValue& b) const {
return a.mProperty == b.mProperty;
}
bool Equals(UntypedDescriptor a, const PropertyValue& b) const {
return a == b.mProperty;
}
bool Equals(const PropertyValue& a, UntypedDescriptor b) const {
return a.mProperty == b;
}
};
nsTArray<PropertyValue> mProperties;
};
inline uint64_t FrameProperties::GetInternal(UntypedDescriptor aProperty,
bool* aFoundResult) const {
MOZ_ASSERT(aProperty, "Null property?");
Bug 1479996 - Combine nsTArray::IndexOf and element access into lambda-friendly functions - r=froydnj In many places, nsTArray::IndexOf is followed by accessing that element (hopefully with `Elements() + index`, which skips unnecessary bounds checks.) But this pattern introduces operations that could be avoided: - IndexOf converts the address of the found element into an index, - The caller must test for a special `NoIndex` value, - On success, accesses convert that index back into the original address. This patch introduces new 'ApplyIf...' functions that combine all these operations in a more efficient ensemble: If the sought element is found, it is passed by reference to a given callable object (usually a lambda); if not found, another callable is invoked. Depending on what they need, the first callable may take one of these parameter lists: (), (size_t), (maybe-const elem_type&), (size_t, maybe-const elem_type&). On top of removing the pointer->index->pointer operations in most cases, invoking callables directly from ApplyIf is safer, as the array is guaranteed to be untouched at this time. Also, being templates taking function objects, in most cases the compiler should be able to inline and optimize the search and its callables' code. This patch gives example uses in nsTArray::Contains, and in FrameProperties::GetInternal, SetInternal. And FrameProperties::Has now calls Contains, which is more efficient because the former code would compute an index (or NoIndex), and then convert that index to a bool; whereas the new code directly produces a bool from within the search algorithm. Differential Revision: https://phabricator.services.mozilla.com/D2758 --HG-- extra : moz-landing-system : lando
2018-08-28 22:04:09 +00:00
return mProperties.ApplyIf(
aProperty, 0, PropertyComparator(),
[&aFoundResult](const PropertyValue& aPV) -> uint64_t {
Bug 1479996 - Combine nsTArray::IndexOf and element access into lambda-friendly functions - r=froydnj In many places, nsTArray::IndexOf is followed by accessing that element (hopefully with `Elements() + index`, which skips unnecessary bounds checks.) But this pattern introduces operations that could be avoided: - IndexOf converts the address of the found element into an index, - The caller must test for a special `NoIndex` value, - On success, accesses convert that index back into the original address. This patch introduces new 'ApplyIf...' functions that combine all these operations in a more efficient ensemble: If the sought element is found, it is passed by reference to a given callable object (usually a lambda); if not found, another callable is invoked. Depending on what they need, the first callable may take one of these parameter lists: (), (size_t), (maybe-const elem_type&), (size_t, maybe-const elem_type&). On top of removing the pointer->index->pointer operations in most cases, invoking callables directly from ApplyIf is safer, as the array is guaranteed to be untouched at this time. Also, being templates taking function objects, in most cases the compiler should be able to inline and optimize the search and its callables' code. This patch gives example uses in nsTArray::Contains, and in FrameProperties::GetInternal, SetInternal. And FrameProperties::Has now calls Contains, which is more efficient because the former code would compute an index (or NoIndex), and then convert that index to a bool; whereas the new code directly produces a bool from within the search algorithm. Differential Revision: https://phabricator.services.mozilla.com/D2758 --HG-- extra : moz-landing-system : lando
2018-08-28 22:04:09 +00:00
if (aFoundResult) {
*aFoundResult = true;
}
return aPV.mValue;
},
[&aFoundResult]() -> uint64_t {
Bug 1479996 - Combine nsTArray::IndexOf and element access into lambda-friendly functions - r=froydnj In many places, nsTArray::IndexOf is followed by accessing that element (hopefully with `Elements() + index`, which skips unnecessary bounds checks.) But this pattern introduces operations that could be avoided: - IndexOf converts the address of the found element into an index, - The caller must test for a special `NoIndex` value, - On success, accesses convert that index back into the original address. This patch introduces new 'ApplyIf...' functions that combine all these operations in a more efficient ensemble: If the sought element is found, it is passed by reference to a given callable object (usually a lambda); if not found, another callable is invoked. Depending on what they need, the first callable may take one of these parameter lists: (), (size_t), (maybe-const elem_type&), (size_t, maybe-const elem_type&). On top of removing the pointer->index->pointer operations in most cases, invoking callables directly from ApplyIf is safer, as the array is guaranteed to be untouched at this time. Also, being templates taking function objects, in most cases the compiler should be able to inline and optimize the search and its callables' code. This patch gives example uses in nsTArray::Contains, and in FrameProperties::GetInternal, SetInternal. And FrameProperties::Has now calls Contains, which is more efficient because the former code would compute an index (or NoIndex), and then convert that index to a bool; whereas the new code directly produces a bool from within the search algorithm. Differential Revision: https://phabricator.services.mozilla.com/D2758 --HG-- extra : moz-landing-system : lando
2018-08-28 22:04:09 +00:00
if (aFoundResult) {
*aFoundResult = false;
}
return 0;
Bug 1479996 - Combine nsTArray::IndexOf and element access into lambda-friendly functions - r=froydnj In many places, nsTArray::IndexOf is followed by accessing that element (hopefully with `Elements() + index`, which skips unnecessary bounds checks.) But this pattern introduces operations that could be avoided: - IndexOf converts the address of the found element into an index, - The caller must test for a special `NoIndex` value, - On success, accesses convert that index back into the original address. This patch introduces new 'ApplyIf...' functions that combine all these operations in a more efficient ensemble: If the sought element is found, it is passed by reference to a given callable object (usually a lambda); if not found, another callable is invoked. Depending on what they need, the first callable may take one of these parameter lists: (), (size_t), (maybe-const elem_type&), (size_t, maybe-const elem_type&). On top of removing the pointer->index->pointer operations in most cases, invoking callables directly from ApplyIf is safer, as the array is guaranteed to be untouched at this time. Also, being templates taking function objects, in most cases the compiler should be able to inline and optimize the search and its callables' code. This patch gives example uses in nsTArray::Contains, and in FrameProperties::GetInternal, SetInternal. And FrameProperties::Has now calls Contains, which is more efficient because the former code would compute an index (or NoIndex), and then convert that index to a bool; whereas the new code directly produces a bool from within the search algorithm. Differential Revision: https://phabricator.services.mozilla.com/D2758 --HG-- extra : moz-landing-system : lando
2018-08-28 22:04:09 +00:00
});
}
inline void FrameProperties::SetInternal(UntypedDescriptor aProperty,
uint64_t aValue,
const nsIFrame* aFrame) {
MOZ_ASSERT(NS_IsMainThread());
MOZ_ASSERT(aProperty, "Null property?");
Bug 1479996 - Combine nsTArray::IndexOf and element access into lambda-friendly functions - r=froydnj In many places, nsTArray::IndexOf is followed by accessing that element (hopefully with `Elements() + index`, which skips unnecessary bounds checks.) But this pattern introduces operations that could be avoided: - IndexOf converts the address of the found element into an index, - The caller must test for a special `NoIndex` value, - On success, accesses convert that index back into the original address. This patch introduces new 'ApplyIf...' functions that combine all these operations in a more efficient ensemble: If the sought element is found, it is passed by reference to a given callable object (usually a lambda); if not found, another callable is invoked. Depending on what they need, the first callable may take one of these parameter lists: (), (size_t), (maybe-const elem_type&), (size_t, maybe-const elem_type&). On top of removing the pointer->index->pointer operations in most cases, invoking callables directly from ApplyIf is safer, as the array is guaranteed to be untouched at this time. Also, being templates taking function objects, in most cases the compiler should be able to inline and optimize the search and its callables' code. This patch gives example uses in nsTArray::Contains, and in FrameProperties::GetInternal, SetInternal. And FrameProperties::Has now calls Contains, which is more efficient because the former code would compute an index (or NoIndex), and then convert that index to a bool; whereas the new code directly produces a bool from within the search algorithm. Differential Revision: https://phabricator.services.mozilla.com/D2758 --HG-- extra : moz-landing-system : lando
2018-08-28 22:04:09 +00:00
mProperties.ApplyIf(
aProperty, 0, PropertyComparator(),
[&](PropertyValue& aPV) {
aPV.DestroyValueFor(aFrame);
aPV.mValue = aValue;
},
[&]() { mProperties.AppendElement(PropertyValue(aProperty, aValue)); });
}
inline void FrameProperties::AddInternal(UntypedDescriptor aProperty,
uint64_t aValue) {
MOZ_ASSERT(NS_IsMainThread());
MOZ_ASSERT(aProperty, "Null property?");
mProperties.AppendElement(PropertyValue(aProperty, aValue));
}
inline uint64_t FrameProperties::TakeInternal(UntypedDescriptor aProperty,
bool* aFoundResult) {
MOZ_ASSERT(NS_IsMainThread());
MOZ_ASSERT(aProperty, "Null property?");
auto index = mProperties.IndexOf(aProperty, 0, PropertyComparator());
if (index == nsTArray<PropertyValue>::NoIndex) {
if (aFoundResult) {
*aFoundResult = false;
}
return 0;
}
if (aFoundResult) {
*aFoundResult = true;
}
uint64_t result = mProperties.Elements()[index].mValue;
mProperties.RemoveElementAtUnsafe(index);
return result;
}
inline void FrameProperties::RemoveInternal(UntypedDescriptor aProperty,
const nsIFrame* aFrame) {
MOZ_ASSERT(NS_IsMainThread());
MOZ_ASSERT(aProperty, "Null property?");
auto index = mProperties.IndexOf(aProperty, 0, PropertyComparator());
if (index != nsTArray<PropertyValue>::NoIndex) {
mProperties.Elements()[index].DestroyValueFor(aFrame);
mProperties.RemoveElementAtUnsafe(index);
}
}
} // namespace mozilla
#endif /* FRAMEPROPERTIES_H_ */