qtdeclarative/src/qml/jsruntime/qv4qobjectwrapper.cpp

2320 lines
87 KiB
C++
Raw Normal View History

2011-05-11 07:20:40 +00:00
/****************************************************************************
**
** Copyright (C) 2016 The Qt Company Ltd.
** Contact: https://www.qt.io/licensing/
2011-05-11 07:20:40 +00:00
**
** This file is part of the QtQml module of the Qt Toolkit.
2011-05-11 07:20:40 +00:00
**
** $QT_BEGIN_LICENSE:LGPL$
** Commercial License Usage
** Licensees holding valid commercial Qt licenses may use this file in
** accordance with the commercial license agreement provided with the
** Software or, alternatively, in accordance with the terms contained in
** a written agreement between you and The Qt Company. For licensing terms
** and conditions see https://www.qt.io/terms-conditions. For further
** information use the contact form at https://www.qt.io/contact-us.
**
2011-05-11 07:20:40 +00:00
** GNU Lesser General Public License Usage
** Alternatively, this file may be used under the terms of the GNU Lesser
** General Public License version 3 as published by the Free Software
** Foundation and appearing in the file LICENSE.LGPL3 included in the
** packaging of this file. Please review the following information to
** ensure the GNU Lesser General Public License version 3 requirements
** will be met: https://www.gnu.org/licenses/lgpl-3.0.html.
**
** GNU General Public License Usage
** Alternatively, this file may be used under the terms of the GNU
** General Public License version 2.0 or (at your option) the GNU General
** Public license version 3 or any later version approved by the KDE Free
** Qt Foundation. The licenses are as published by the Free Software
** Foundation and appearing in the file LICENSE.GPL2 and LICENSE.GPL3
** included in the packaging of this file. Please review the following
** information to ensure the GNU General Public License requirements will
** be met: https://www.gnu.org/licenses/gpl-2.0.html and
** https://www.gnu.org/licenses/gpl-3.0.html.
2011-05-11 07:20:40 +00:00
**
** $QT_END_LICENSE$
**
****************************************************************************/
#include "qv4qobjectwrapper_p.h"
2011-05-11 07:20:40 +00:00
#include <private/qqmlpropertycache_p.h>
#include <private/qqmlengine_p.h>
#include <private/qqmlvmemetaobject_p.h>
#include <private/qqmlbinding_p.h>
#include <private/qjsvalue_p.h>
#include <private/qqmlexpression_p.h>
#include <private/qqmlglobal_p.h>
#include <private/qqmltypewrapper_p.h>
#include <private/qqmlvaluetypewrapper_p.h>
#include <private/qqmllistwrapper_p.h>
#include <private/qqmlbuiltinfunctions_p.h>
#include <private/qv8engine_p.h>
2011-05-11 07:20:40 +00:00
#include <private/qv4arraybuffer_p.h>
#include <private/qv4functionobject_p.h>
#include <private/qv4runtime_p.h>
#include <private/qv4variantobject_p.h>
#if QT_CONFIG(qml_sequence_object)
#include <private/qv4sequenceobject_p.h>
#endif
#include <private/qv4objectproto_p.h>
#include <private/qv4jsonobject_p.h>
#include <private/qv4regexpobject_p.h>
#include <private/qv4dateobject_p.h>
#include <private/qv4scopedvalue_p.h>
#include <private/qv4jscall_p.h>
#include <private/qv4mm_p.h>
#include <private/qqmlscriptstring_p.h>
#include <private/qv4compileddata_p.h>
#include <QtQml/qjsvalue.h>
#include <QtCore/qjsonarray.h>
#include <QtCore/qjsonobject.h>
#include <QtCore/qjsonvalue.h>
2011-05-11 07:20:40 +00:00
#include <QtCore/qvarlengtharray.h>
#include <QtCore/qtimer.h>
#include <QtCore/qatomic.h>
#include <QtCore/qmetaobject.h>
#include <QtCore/qabstractitemmodel.h>
#include <QtCore/qloggingcategory.h>
2011-05-11 07:20:40 +00:00
#include <vector>
QT_BEGIN_NAMESPACE
Q_LOGGING_CATEGORY(lcBindingRemoval, "qt.qml.binding.removal", QtWarningMsg)
2011-05-11 07:20:40 +00:00
// The code in this file does not violate strict aliasing, but GCC thinks it does
// so turn off the warnings for us to have a clean build
QT_WARNING_DISABLE_GCC("-Wstrict-aliasing")
using namespace QV4;
QPair<QObject *, int> QObjectMethod::extractQtMethod(const QV4::FunctionObject *function)
{
QV4::ExecutionEngine *v4 = function->engine();
if (v4) {
QV4::Scope scope(v4);
QV4::Scoped<QObjectMethod> method(scope, function->as<QObjectMethod>());
if (method)
return qMakePair(method->object(), method->methodIndex());
}
return qMakePair((QObject *)nullptr, -1);
}
static QPair<QObject *, int> extractQtSignal(const QV4::Value &value)
{
if (value.isObject()) {
QV4::ExecutionEngine *v4 = value.as<QV4::Object>()->engine();
QV4::Scope scope(v4);
QV4::ScopedFunctionObject function(scope, value);
if (function)
return QObjectMethod::extractQtMethod(function);
QV4::Scoped<QV4::QmlSignalHandler> handler(scope, value);
if (handler)
return qMakePair(handler->object(), handler->signalIndex());
}
return qMakePair((QObject *)nullptr, -1);
}
static QV4::ReturnedValue loadProperty(QV4::ExecutionEngine *v4, QObject *object,
const QQmlPropertyData &property)
{
Q_ASSERT(!property.isFunction());
QV4::Scope scope(v4);
if (property.isQObject()) {
QObject *rv = nullptr;
property.readProperty(object, &rv);
return QV4::QObjectWrapper::wrap(v4, rv);
} else if (property.isQList()) {
return QmlListWrapper::create(v4, object, property.coreIndex(), property.propType());
} else if (property.propType() == QMetaType::QReal) {
qreal v = 0;
property.readProperty(object, &v);
return QV4::Encode(v);
} else if (property.propType() == QMetaType::Int || property.isEnum()) {
int v = 0;
property.readProperty(object, &v);
return QV4::Encode(v);
} else if (property.propType() == QMetaType::Bool) {
bool v = false;
property.readProperty(object, &v);
return QV4::Encode(v);
} else if (property.propType() == QMetaType::QString) {
QString v;
property.readProperty(object, &v);
return v4->newString(v)->asReturnedValue();
} else if (property.propType() == QMetaType::UInt) {
uint v = 0;
property.readProperty(object, &v);
return QV4::Encode(v);
} else if (property.propType() == QMetaType::Float) {
float v = 0;
property.readProperty(object, &v);
return QV4::Encode(v);
} else if (property.propType() == QMetaType::Double) {
double v = 0;
property.readProperty(object, &v);
return QV4::Encode(v);
} else if (property.isV4Handle()) {
QQmlV4Handle handle;
property.readProperty(object, &handle);
return handle;
} else if (property.propType() == qMetaTypeId<QJSValue>()) {
QJSValue v;
property.readProperty(object, &v);
return QJSValuePrivate::convertedToValue(v4, v);
} else if (property.isQVariant()) {
QVariant v;
property.readProperty(object, &v);
if (QQmlValueTypeFactory::isValueType(v.userType())) {
if (const QMetaObject *valueTypeMetaObject = QQmlValueTypeFactory::metaObjectForMetaType(v.userType()))
return QV4::QQmlValueTypeWrapper::create(v4, object, property.coreIndex(), valueTypeMetaObject, v.userType()); // VariantReference value-type.
}
return scope.engine->fromVariant(v);
} else if (QQmlValueTypeFactory::isValueType(property.propType())) {
if (const QMetaObject *valueTypeMetaObject = QQmlValueTypeFactory::metaObjectForMetaType(property.propType()))
return QV4::QQmlValueTypeWrapper::create(v4, object, property.coreIndex(), valueTypeMetaObject, property.propType());
} else {
#if QT_CONFIG(qml_sequence_object)
// see if it's a sequence type
bool succeeded = false;
QV4::ScopedValue retn(scope, QV4::SequencePrototype::newSequence(v4, property.propType(), object, property.coreIndex(), !property.isWritable(), &succeeded));
if (succeeded)
return retn->asReturnedValue();
#endif
}
if (property.propType() == QMetaType::UnknownType) {
QMetaProperty p = object->metaObject()->property(property.coreIndex());
qWarning("QMetaProperty::read: Unable to handle unregistered datatype '%s' for property "
"'%s::%s'", p.typeName(), object->metaObject()->className(), p.name());
return QV4::Encode::undefined();
} else {
QVariant v(property.propType(), (void *)nullptr);
property.readProperty(object, v.data());
return scope.engine->fromVariant(v);
}
}
void QObjectWrapper::initializeBindings(ExecutionEngine *engine)
{
engine->functionPrototype()->defineDefaultProperty(QStringLiteral("connect"), method_connect);
engine->functionPrototype()->defineDefaultProperty(QStringLiteral("disconnect"), method_disconnect);
}
QQmlPropertyData *QObjectWrapper::findProperty(ExecutionEngine *engine, QQmlContextData *qmlContext, String *name, RevisionMode revisionMode, QQmlPropertyData *local) const
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
{
QObject *o = d()->object();
return findProperty(engine, o, qmlContext, name, revisionMode, local);
}
QQmlPropertyData *QObjectWrapper::findProperty(ExecutionEngine *engine, QObject *o, QQmlContextData *qmlContext, String *name, RevisionMode revisionMode, QQmlPropertyData *local)
{
Q_UNUSED(revisionMode);
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
QQmlData *ddata = QQmlData::get(o, false);
QQmlPropertyData *result = nullptr;
if (ddata && ddata->propertyCache)
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
result = ddata->propertyCache->property(name, o, qmlContext);
else
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
result = QQmlPropertyCache::property(engine->jsEngine(), o, name, qmlContext, *local);
return result;
}
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
ReturnedValue QObjectWrapper::getProperty(ExecutionEngine *engine, QObject *object, QQmlPropertyData *property, bool captureRequired)
{
QQmlData::flushPendingBinding(object, QQmlPropertyIndex(property->coreIndex()));
if (property->isFunction() && !property->isVarProperty()) {
if (property->isVMEFunction()) {
QQmlVMEMetaObject *vmemo = QQmlVMEMetaObject::get(object);
Q_ASSERT(vmemo);
return vmemo->vmeMethod(property->coreIndex());
} else if (property->isV4Function()) {
Scope scope(engine);
ScopedContext global(scope, engine->qmlContext());
if (!global)
global = engine->rootContext();
return QV4::QObjectMethod::create(global, object, property->coreIndex());
} else if (property->isSignalHandler()) {
QmlSignalHandler::initProto(engine);
return engine->memoryManager->allocate<QV4::QmlSignalHandler>(object, property->coreIndex())->asReturnedValue();
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
} else {
ExecutionContext *global = engine->rootContext();
return QV4::QObjectMethod::create(global, object, property->coreIndex());
}
}
QQmlEnginePrivate *ep = engine->qmlEngine() ? QQmlEnginePrivate::get(engine->qmlEngine()) : nullptr;
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
if (captureRequired && ep && ep->propertyCapture && !property->isConstant())
ep->propertyCapture->captureProperty(object, property->coreIndex(), property->notifyIndex());
if (property->isVarProperty()) {
QQmlVMEMetaObject *vmemo = QQmlVMEMetaObject::get(object);
Q_ASSERT(vmemo);
return vmemo->vmeProperty(property->coreIndex());
} else {
return loadProperty(engine, object, *property);
}
}
ReturnedValue QObjectWrapper::getQmlProperty(QQmlContextData *qmlContext, String *name, QObjectWrapper::RevisionMode revisionMode,
bool *hasProperty, bool includeImports) const
{
if (QQmlData::wasDeleted(d()->object())) {
if (hasProperty)
*hasProperty = false;
return QV4::Encode::undefined();
}
ExecutionEngine *v4 = engine();
if (name->equals(v4->id_destroy()) || name->equals(v4->id_toString())) {
int index = name->equals(v4->id_destroy()) ? QV4::QObjectMethod::DestroyMethod : QV4::QObjectMethod::ToStringMethod;
if (hasProperty)
*hasProperty = true;
ExecutionContext *global = v4->rootContext();
return QV4::QObjectMethod::create(global, d()->object(), index);
}
QQmlPropertyData local;
QQmlPropertyData *result = findProperty(v4, qmlContext, name, revisionMode, &local);
if (!result) {
if (includeImports && name->startsWithUpper()) {
// Check for attached properties
if (qmlContext && qmlContext->imports) {
QQmlTypeNameCache::Result r = qmlContext->imports->query(name);
if (hasProperty)
*hasProperty = true;
if (r.isValid()) {
if (r.scriptIndex != -1) {
return QV4::Encode::undefined();
} else if (r.type.isValid()) {
return QQmlTypeWrapper::create(v4, d()->object(),
r.type, Heap::QQmlTypeWrapper::ExcludeEnums);
} else if (r.importNamespace) {
return QQmlTypeWrapper::create(v4, d()->object(),
qmlContext->imports, r.importNamespace, Heap::QQmlTypeWrapper::ExcludeEnums);
}
Q_ASSERT(!"Unreachable");
}
}
}
return QV4::Object::virtualGet(this, name->propertyKey(), this, hasProperty);
}
QQmlData *ddata = QQmlData::get(d()->object(), false);
if (revisionMode == QV4::QObjectWrapper::CheckRevision && result->hasRevision()) {
if (ddata && ddata->propertyCache && !ddata->propertyCache->isAllowedInRevision(result)) {
if (hasProperty)
*hasProperty = false;
return QV4::Encode::undefined();
}
}
if (hasProperty)
*hasProperty = true;
return getProperty(v4, d()->object(), result);
}
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
ReturnedValue QObjectWrapper::getProperty(ExecutionEngine *engine, QObject *object, int propertyIndex, bool captureRequired)
{
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
if (QQmlData::wasDeleted(object))
return QV4::Encode::null();
QQmlData *ddata = QQmlData::get(object, /*create*/false);
if (!ddata)
return QV4::Encode::undefined();
if (Q_UNLIKELY(!ddata->propertyCache)) {
ddata->propertyCache = QQmlEnginePrivate::get(engine)->cache(object->metaObject());
ddata->propertyCache->addref();
}
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
QQmlPropertyCache *cache = ddata->propertyCache;
Q_ASSERT(cache);
QQmlPropertyData *property = cache->property(propertyIndex);
Q_ASSERT(property); // We resolved this property earlier, so it better exist!
return getProperty(engine, object, property, captureRequired);
}
ReturnedValue QObjectWrapper::getQmlProperty(QV4::ExecutionEngine *engine, QQmlContextData *qmlContext, QObject *object, String *name, QObjectWrapper::RevisionMode revisionMode, bool *hasProperty)
{
if (QQmlData::wasDeleted(object)) {
if (hasProperty)
*hasProperty = false;
return QV4::Encode::null();
}
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
if (name->equals(engine->id_destroy()) || name->equals(engine->id_toString())) {
int index = name->equals(engine->id_destroy()) ? QV4::QObjectMethod::DestroyMethod : QV4::QObjectMethod::ToStringMethod;
if (hasProperty)
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
*hasProperty = true;
ExecutionContext *global = engine->rootContext();
return QV4::QObjectMethod::create(global, object, index);
}
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
QQmlData *ddata = QQmlData::get(object, false);
QQmlPropertyData local;
QQmlPropertyData *result = findProperty(engine, object, qmlContext, name, revisionMode, &local);
if (result) {
if (revisionMode == QV4::QObjectWrapper::CheckRevision && result->hasRevision()) {
if (ddata && ddata->propertyCache && !ddata->propertyCache->isAllowedInRevision(result)) {
if (hasProperty)
*hasProperty = false;
return QV4::Encode::undefined();
}
}
if (hasProperty)
*hasProperty = true;
return getProperty(engine, object, result);
} else {
// Check if this object is already wrapped.
if (!ddata || (ddata->jsWrapper.isUndefined() &&
(ddata->jsEngineId == 0 || // Nobody owns the QObject
!ddata->hasTaintedV4Object))) { // Someone else has used the QObject, but it isn't tainted
// Not wrapped. Last chance: try query QObjectWrapper's prototype.
// If it can't handle this, then there is no point
// to wrap the QObject just to look at an empty set of JS props.
QV4::Object *proto = QObjectWrapper::defaultPrototype(engine);
return proto->get(name, hasProperty);
}
}
// If we get here, we must already be wrapped (which implies a ddata).
// There's no point wrapping again, as there wouldn't be any new props.
Q_ASSERT(ddata);
QV4::Scope scope(engine);
QV4::Scoped<QObjectWrapper> wrapper(scope, wrap(engine, object));
if (!wrapper) {
if (hasProperty)
*hasProperty = false;
return QV4::Encode::null();
}
return wrapper->getQmlProperty(qmlContext, name, revisionMode, hasProperty);
}
QObjectWrapper: Try harder to avoid creating wrappers if we can When searching a QObject for properties, try to find QObject properties first, and then bail if we don't find a match and the object is not already wrapped. It makes no sense to wrap the object just to check the (empty) JS side props, which is often the case with QmlContextWrapper's scopeObject and contextObject. This can especially be seen when looking at QML's model activity, as they set a context object in order to hijack lookup of model data & index. This can be seen in a test like the following: Repeater { model: 10 delegate: Text { text: index } } Before this patch, the 'index' access would previously trigger wrapping of the QQmlDelegateModelItem. The same applies to the 'model' prop etc. Noteworthy improvements of a full qmlbench run that aren't part of the "usual" statistical noise are generally around the 5% mark. Outside performance, the advantages are also clear on memory. Examining for instance delegates_flipable.qml with QV4_MM_STATS=1 on my MBP: Before: Freed JS type: QmlContext (5000 instances) Freed JS type: QObjectWrapper (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) After: Freed JS type: QmlContext (5000 instances) Freed JS type: QmlContextWrapper (5000 instances) Freed JS type: QQuickItemWrapper (5000 instances) ... which is what we expect, no more QObjectWrappers here, a lot less garbage to clear up. Timewise, while the measured mark/sweep times are not stable, there is a proportional decrease of about 40% in time for both mark and sweep phases in the GC (slightly more on TX1 than desktop, curiously). Finally, GC's memory use is quite stable for each round of the test, and in this case, we see a decrease of 378kb used at the time of each GC. Before: Allocated 2515968 bytes in 39 chunks Used memory before GC: 2489824 Used memory after GC : 889856 Freed up bytes : 1599968 After: Allocated 2128896 bytes in 33 chunks Used memory before GC: 2098304 Used memory after GC : 818304 Freed up bytes : 1280000 As a final note, while this provides an improvement to many of qmlbench's benchmarks, as they make heavy use of Repeater, these benefits should map quite well to real world applications as well: anything that uses item views will likely see a significant reduction to the amount of GC'd objects. In the future, it would be nice to have tests for the amount of garbage created by various activities, but that is an item for another day. Change-Id: Idd14180e689235eea8883d68487570b0e14d47e6 Reviewed-by: Michael Brasser <michael.brasser@live.com> Reviewed-by: Simon Hausmann <simon.hausmann@qt.io>
2017-03-02 08:41:16 +00:00
bool QObjectWrapper::setQmlProperty(ExecutionEngine *engine, QQmlContextData *qmlContext, QObject *object, String *name,
QObjectWrapper::RevisionMode revisionMode, const Value &value)
{
if (QQmlData::wasDeleted(object))
return false;
QQmlPropertyData local;
QQmlPropertyData *result = QQmlPropertyCache::property(engine->jsEngine(), object, name, qmlContext, local);
if (!result)
return false;
if (revisionMode == QV4::QObjectWrapper::CheckRevision && result->hasRevision()) {
QQmlData *ddata = QQmlData::get(object);
if (ddata && ddata->propertyCache && !ddata->propertyCache->isAllowedInRevision(result))
return false;
}
setProperty(engine, object, result, value);
return true;
}
void QObjectWrapper::setProperty(ExecutionEngine *engine, QObject *object, QQmlPropertyData *property, const Value &value)
{
if (!property->isWritable() && !property->isQList()) {
QString error = QLatin1String("Cannot assign to read-only property \"") +
property->name(object) + QLatin1Char('\"');
engine->throwTypeError(error);
return;
}
QQmlBinding *newBinding = nullptr;
QV4::Scope scope(engine);
QV4::ScopedFunctionObject f(scope, value);
if (f) {
if (!f->isBinding()) {
if (!property->isVarProperty() && property->propType() != qMetaTypeId<QJSValue>()) {
// assigning a JS function to a non var or QJSValue property or is not allowed.
QString error = QLatin1String("Cannot assign JavaScript function to ");
if (!QMetaType::typeName(property->propType()))
error += QLatin1String("[unknown property type]");
else
error += QLatin1String(QMetaType::typeName(property->propType()));
scope.engine->throwError(error);
return;
}
} else {
// binding assignment.
QQmlContextData *callingQmlContext = scope.engine->callingQmlContext();
QV4::Scoped<QQmlBindingFunction> bindingFunction(scope, (const Value &)f);
QV4::ScopedFunctionObject f(scope, bindingFunction->bindingFunction());
QV4::ScopedContext ctx(scope, bindingFunction->scope());
newBinding = QQmlBinding::create(property, f->function(), object, callingQmlContext, ctx);
newBinding->setSourceLocation(bindingFunction->currentLocation());
if (f->isBoundFunction())
newBinding->setBoundFunction(static_cast<QV4::BoundFunction *>(f.getPointer()));
newBinding->setTarget(object, *property, nullptr);
}
}
if (newBinding) {
QQmlPropertyPrivate::setBinding(newBinding);
} else {
if (Q_UNLIKELY(lcBindingRemoval().isInfoEnabled())) {
if (auto binding = QQmlPropertyPrivate::binding(object, QQmlPropertyIndex(property->coreIndex()))) {
Q_ASSERT(!binding->isValueTypeProxy());
const auto qmlBinding = static_cast<const QQmlBinding*>(binding);
const auto stackFrame = engine->currentStackFrame;
qCInfo(lcBindingRemoval,
"Overwriting binding on %s::%s at %s:%d that was initially bound at %s",
object->metaObject()->className(), qPrintable(property->name(object)),
qPrintable(stackFrame->source()), stackFrame->lineNumber(),
qPrintable(qmlBinding->expressionIdentifier()));
}
}
QQmlPropertyPrivate::removeBinding(object, QQmlPropertyIndex(property->coreIndex()));
}
if (!newBinding && property->isVarProperty()) {
// allow assignment of "special" values (null, undefined, function) to var properties
QQmlVMEMetaObject *vmemo = QQmlVMEMetaObject::get(object);
Q_ASSERT(vmemo);
vmemo->setVMEProperty(property->coreIndex(), value);
return;
}
#define PROPERTY_STORE(cpptype, value) \
cpptype o = value; \
int status = -1; \
int flags = 0; \
void *argv[] = { &o, 0, &status, &flags }; \
QMetaObject::metacall(object, QMetaObject::WriteProperty, property->coreIndex(), argv);
if (value.isNull() && property->isQObject()) {
PROPERTY_STORE(QObject*, nullptr);
} else if (value.isUndefined() && property->isResettable()) {
void *a[] = { nullptr };
QMetaObject::metacall(object, QMetaObject::ResetProperty, property->coreIndex(), a);
} else if (value.isUndefined() && property->propType() == qMetaTypeId<QVariant>()) {
PROPERTY_STORE(QVariant, QVariant());
} else if (value.isUndefined() && property->propType() == QMetaType::QJsonValue) {
PROPERTY_STORE(QJsonValue, QJsonValue(QJsonValue::Undefined));
} else if (!newBinding && property->propType() == qMetaTypeId<QJSValue>()) {
PROPERTY_STORE(QJSValue, QJSValue(scope.engine, value.asReturnedValue()));
} else if (value.isUndefined() && property->propType() != qMetaTypeId<QQmlScriptString>()) {
QString error = QLatin1String("Cannot assign [undefined] to ");
if (!QMetaType::typeName(property->propType()))
error += QLatin1String("[unknown property type]");
else
error += QLatin1String(QMetaType::typeName(property->propType()));
scope.engine->throwError(error);
return;
} else if (value.as<FunctionObject>()) {
// this is handled by the binding creation above
} else if (property->propType() == QMetaType::Int && value.isNumber()) {
PROPERTY_STORE(int, value.asDouble());
} else if (property->propType() == QMetaType::QReal && value.isNumber()) {
PROPERTY_STORE(qreal, qreal(value.asDouble()));
} else if (property->propType() == QMetaType::Float && value.isNumber()) {
PROPERTY_STORE(float, float(value.asDouble()));
} else if (property->propType() == QMetaType::Double && value.isNumber()) {
PROPERTY_STORE(double, double(value.asDouble()));
} else if (property->propType() == QMetaType::QString && value.isString()) {
PROPERTY_STORE(QString, value.toQStringNoThrow());
} else if (property->isVarProperty()) {
QQmlVMEMetaObject *vmemo = QQmlVMEMetaObject::get(object);
Q_ASSERT(vmemo);
vmemo->setVMEProperty(property->coreIndex(), value);
} else if (property->propType() == qMetaTypeId<QQmlScriptString>() && (value.isUndefined() || value.isPrimitive())) {
QQmlScriptString ss(value.toQStringNoThrow(), nullptr /* context */, object);
if (value.isNumber()) {
ss.d->numberValue = value.toNumber();
ss.d->isNumberLiteral = true;
} else if (value.isString()) {
ss.d->script = QV4::CompiledData::Binding::escapedString(ss.d->script);
ss.d->isStringLiteral = true;
}
PROPERTY_STORE(QQmlScriptString, ss);
} else {
QVariant v;
if (property->isQList())
v = scope.engine->toVariant(value, qMetaTypeId<QList<QObject *> >());
else
v = scope.engine->toVariant(value, property->propType());
QQmlContextData *callingQmlContext = scope.engine->callingQmlContext();
if (!QQmlPropertyPrivate::write(object, *property, v, callingQmlContext)) {
const char *valueType = nullptr;
if (v.userType() == QVariant::Invalid) valueType = "null";
else valueType = QMetaType::typeName(v.userType());
const char *targetTypeName = QMetaType::typeName(property->propType());
if (!targetTypeName)
targetTypeName = "an unregistered type";
QString error = QLatin1String("Cannot assign ") +
QLatin1String(valueType) +
QLatin1String(" to ") +
QLatin1String(targetTypeName);
scope.engine->throwError(error);
return;
}
}
}
ReturnedValue QObjectWrapper::wrap_slowPath(ExecutionEngine *engine, QObject *object)
{
Q_ASSERT(!QQmlData::wasDeleted(object));
QQmlData *ddata = QQmlData::get(object, true);
if (!ddata)
return QV4::Encode::undefined();
Scope scope(engine);
if (ddata->jsWrapper.isUndefined() &&
(ddata->jsEngineId == engine->m_engineId || // We own the QObject
ddata->jsEngineId == 0 || // No one owns the QObject
!ddata->hasTaintedV4Object)) { // Someone else has used the QObject, but it isn't tainted
QV4::ScopedValue rv(scope, create(engine, object));
ddata->jsWrapper.set(scope.engine, rv);
ddata->jsEngineId = engine->m_engineId;
return rv->asReturnedValue();
} else {
// If this object is tainted, we have to check to see if it is in our
// tainted object list
ScopedObject alternateWrapper(scope, (Object *)nullptr);
if (engine->m_multiplyWrappedQObjects && ddata->hasTaintedV4Object)
alternateWrapper = engine->m_multiplyWrappedQObjects->value(object);
// If our tainted handle doesn't exist or has been collected, and there isn't
// a handle in the ddata, we can assume ownership of the ddata->jsWrapper
if (ddata->jsWrapper.isUndefined() && !alternateWrapper) {
QV4::ScopedValue result(scope, create(engine, object));
ddata->jsWrapper.set(scope.engine, result);
ddata->jsEngineId = engine->m_engineId;
return result->asReturnedValue();
}
if (!alternateWrapper) {
alternateWrapper = create(engine, object);
if (!engine->m_multiplyWrappedQObjects)
engine->m_multiplyWrappedQObjects = new MultiplyWrappedQObjectMap;
engine->m_multiplyWrappedQObjects->insert(object, alternateWrapper->d());
ddata->hasTaintedV4Object = true;
}
return alternateWrapper.asReturnedValue();
}
}
void QObjectWrapper::markWrapper(QObject *object, MarkStack *markStack)
{
if (QQmlData::wasDeleted(object))
return;
QQmlData *ddata = QQmlData::get(object);
if (!ddata)
return;
if (ddata->jsEngineId == markStack->engine->m_engineId)
ddata->jsWrapper.markOnce(markStack);
else if (markStack->engine->m_multiplyWrappedQObjects && ddata->hasTaintedV4Object)
markStack->engine->m_multiplyWrappedQObjects->mark(object, markStack);
}
void QObjectWrapper::setProperty(ExecutionEngine *engine, int propertyIndex, const Value &value)
{
setProperty(engine, d()->object(), propertyIndex, value);
}
void QObjectWrapper::setProperty(ExecutionEngine *engine, QObject *object, int propertyIndex, const Value &value)
{
Q_ASSERT(propertyIndex < 0xffff);
Q_ASSERT(propertyIndex >= 0);
if (QQmlData::wasDeleted(object))
return;
QQmlData *ddata = QQmlData::get(object, /*create*/false);
if (!ddata)
return;
QQmlPropertyCache *cache = ddata->propertyCache;
Q_ASSERT(cache);
QQmlPropertyData *property = cache->property(propertyIndex);
Q_ASSERT(property); // We resolved this property earlier, so it better exist!
return setProperty(engine, object, property, value);
}
bool QObjectWrapper::virtualIsEqualTo(Managed *a, Managed *b)
{
Q_ASSERT(a->as<QV4::QObjectWrapper>());
QV4::QObjectWrapper *qobjectWrapper = static_cast<QV4::QObjectWrapper *>(a);
QV4::Object *o = b->as<Object>();
if (o) {
if (QV4::QQmlTypeWrapper *qmlTypeWrapper = o->as<QV4::QQmlTypeWrapper>())
return qmlTypeWrapper->toVariant().value<QObject*>() == qobjectWrapper->object();
}
return false;
}
ReturnedValue QObjectWrapper::create(ExecutionEngine *engine, QObject *object)
{
if (QJSEngine *jsEngine = engine->jsEngine()) {
if (QQmlPropertyCache *cache = QQmlData::ensurePropertyCache(jsEngine, object)) {
ReturnedValue result = QV4::Encode::null();
void *args[] = { &result, &engine };
if (cache->callJSFactoryMethod(object, args))
return result;
}
}
return (engine->memoryManager->allocate<QV4::QObjectWrapper>(object))->asReturnedValue();
}
QV4::ReturnedValue QObjectWrapper::virtualGet(const Managed *m, PropertyKey id, const Value *receiver, bool *hasProperty)
{
if (!id.isString())
return Object::virtualGet(m, id, receiver, hasProperty);
const QObjectWrapper *that = static_cast<const QObjectWrapper*>(m);
Scope scope(that);
ScopedString n(scope, id.asStringOrSymbol());
QQmlContextData *qmlContext = that->engine()->callingQmlContext();
return that->getQmlProperty(qmlContext, n, IgnoreRevision, hasProperty, /*includeImports*/ true);
}
bool QObjectWrapper::virtualPut(Managed *m, PropertyKey id, const Value &value, Value *receiver)
{
if (!id.isString())
return Object::virtualPut(m, id, value, receiver);
Scope scope(m);
QObjectWrapper *that = static_cast<QObjectWrapper*>(m);
ScopedString name(scope, id.asStringOrSymbol());
if (scope.engine->hasException || QQmlData::wasDeleted(that->d()->object()))
return false;
QQmlContextData *qmlContext = scope.engine->callingQmlContext();
if (!setQmlProperty(scope.engine, qmlContext, that->d()->object(), name, QV4::QObjectWrapper::IgnoreRevision, value)) {
QQmlData *ddata = QQmlData::get(that->d()->object());
// Types created by QML are not extensible at run-time, but for other QObjects we can store them
// as regular JavaScript properties, like on JavaScript objects.
if (ddata && ddata->context) {
QString error = QLatin1String("Cannot assign to non-existent property \"") +
name->toQString() + QLatin1Char('\"');
scope.engine->throwError(error);
return false;
} else {
return QV4::Object::virtualPut(m, id, value, receiver);
}
}
return true;
}
PropertyAttributes QObjectWrapper::virtualGetOwnProperty(Managed *m, PropertyKey id, Property *p)
{
if (id.isString()) {
QObjectWrapper *that = static_cast<QObjectWrapper*>(m);
const QObject *thatObject = that->d()->object();
if (!QQmlData::wasDeleted(thatObject)) {
Scope scope(m);
ScopedString n(scope, id.asStringOrSymbol());
QQmlContextData *qmlContext = scope.engine->callingQmlContext();
QQmlPropertyData local;
if (that->findProperty(scope.engine, qmlContext, n, IgnoreRevision, &local)
|| n->equals(scope.engine->id_destroy()) || n->equals(scope.engine->id_toString())) {
if (p) {
// ### probably not the fastest implementation
bool hasProperty;
p->value = that->getQmlProperty(qmlContext, n, IgnoreRevision, &hasProperty, /*includeImports*/ true);
}
return QV4::Attr_Data;
}
}
}
return QV4::Object::virtualGetOwnProperty(m, id, p);
}
void QObjectWrapper::virtualAdvanceIterator(Managed *m, ObjectIterator *it, Value *name, uint *index, Property *p, PropertyAttributes *attributes)
{
// Used to block access to QObject::destroyed() and QObject::deleteLater() from QML
static const int destroyedIdx1 = QObject::staticMetaObject.indexOfSignal("destroyed(QObject*)");
static const int destroyedIdx2 = QObject::staticMetaObject.indexOfSignal("destroyed()");
static const int deleteLaterIdx = QObject::staticMetaObject.indexOfSlot("deleteLater()");
name->setM(nullptr);
*index = UINT_MAX;
QObjectWrapper *that = static_cast<QObjectWrapper*>(m);
QObject *thatObject = that->d()->object();
if (thatObject && !QQmlData::wasDeleted(thatObject)) {
const QMetaObject *mo = thatObject->metaObject();
// These indices don't apply to gadgets, so don't block them.
const bool preventDestruction = mo->superClass() || mo == &QObject::staticMetaObject;
const int propertyCount = mo->propertyCount();
if (it->arrayIndex < static_cast<uint>(propertyCount)) {
ExecutionEngine *thatEngine = that->engine();
Scope scope(thatEngine);
const QMetaProperty property = mo->property(it->arrayIndex);
ScopedString propName(scope, thatEngine->newString(QString::fromUtf8(property.name())));
Remove type punning from QV4::Value. The union in QV4::Value is used to do type punning. In C++, this is compiler-defined behavior. For example, Clang and GCC will try to detect it and try to do the proper thing. However, it can play havoc with Alias Analysis, and it is not guaranteed that some Undefined Behavior (or Compiler depenedent behavior) might occur. The really problematic part is the struct inside the union: depending on the calling convention and the register size, it results in some exciting code. For example, the AMD64 ABI specifies that a struct of two values of INTEGER class can be passed in separate registers when doing a function call. Now, if the AA in the compiler looses track of the fact that the tag overlaps with the double, you might get: ecx := someTag ... conditional jumps double_case: rdx := xorredDoubleValue callq someWhere If the someWhere function checks for the tag first, mayhem ensues: the double value in rdx does not overwrite the tag that is passed in ecx. Changing the code to do reinterpret_cast<>s might also give problems on 32bit architectures, because there is a double, whose size is not the same as the size of the tag, which could confuse AA. So, to fix this, the following is changed: - only have a quint64 field in the QV4::Value, which has the added benefit that it's very clear for the compiler that it's a POD - as memcpy is the only approved way to ensure bit-by-bit "conversion" between types (esp. FP<->non-FP types), change all conversions to use memcpy. Use bitops (shift/and/or) for anything else. - only use accessor functions for non-quint64 values As any modern compiler has memcpy as an intrinsic, the call will be replaced with one or a few move instructions. The accessor functions also get inlined, the bitops get optimized, so in all cases the compiler can generate the most compact code possible. This patch obsoletes f558bc48585c69de36151248c969a484a969ebb4 (which had the exact aliassing problem of the double and the tag as described above). Change-Id: I60a39d8564be5ce6106403a56a8de90943217006 Reviewed-by: Ulf Hermann <ulf.hermann@theqtcompany.com>
2015-07-08 08:52:59 +00:00
name->setM(propName->d());
++it->arrayIndex;
*attributes = QV4::Attr_Data;
QQmlPropertyData local;
local.load(property);
p->value = that->getProperty(thatEngine, thatObject, &local);
return;
}
const int methodCount = mo->methodCount();
while (it->arrayIndex < static_cast<uint>(propertyCount + methodCount)) {
const int index = it->arrayIndex - propertyCount;
const QMetaMethod method = mo->method(index);
++it->arrayIndex;
if (method.access() == QMetaMethod::Private || (preventDestruction && (index == deleteLaterIdx || index == destroyedIdx1 || index == destroyedIdx2)))
continue;
ExecutionEngine *thatEngine = that->engine();
Scope scope(thatEngine);
ScopedString methodName(scope, thatEngine->newString(QString::fromUtf8(method.name())));
Remove type punning from QV4::Value. The union in QV4::Value is used to do type punning. In C++, this is compiler-defined behavior. For example, Clang and GCC will try to detect it and try to do the proper thing. However, it can play havoc with Alias Analysis, and it is not guaranteed that some Undefined Behavior (or Compiler depenedent behavior) might occur. The really problematic part is the struct inside the union: depending on the calling convention and the register size, it results in some exciting code. For example, the AMD64 ABI specifies that a struct of two values of INTEGER class can be passed in separate registers when doing a function call. Now, if the AA in the compiler looses track of the fact that the tag overlaps with the double, you might get: ecx := someTag ... conditional jumps double_case: rdx := xorredDoubleValue callq someWhere If the someWhere function checks for the tag first, mayhem ensues: the double value in rdx does not overwrite the tag that is passed in ecx. Changing the code to do reinterpret_cast<>s might also give problems on 32bit architectures, because there is a double, whose size is not the same as the size of the tag, which could confuse AA. So, to fix this, the following is changed: - only have a quint64 field in the QV4::Value, which has the added benefit that it's very clear for the compiler that it's a POD - as memcpy is the only approved way to ensure bit-by-bit "conversion" between types (esp. FP<->non-FP types), change all conversions to use memcpy. Use bitops (shift/and/or) for anything else. - only use accessor functions for non-quint64 values As any modern compiler has memcpy as an intrinsic, the call will be replaced with one or a few move instructions. The accessor functions also get inlined, the bitops get optimized, so in all cases the compiler can generate the most compact code possible. This patch obsoletes f558bc48585c69de36151248c969a484a969ebb4 (which had the exact aliassing problem of the double and the tag as described above). Change-Id: I60a39d8564be5ce6106403a56a8de90943217006 Reviewed-by: Ulf Hermann <ulf.hermann@theqtcompany.com>
2015-07-08 08:52:59 +00:00
name->setM(methodName->d());
*attributes = QV4::Attr_Data;
QQmlPropertyData local;
local.load(method);
p->value = that->getProperty(thatEngine, thatObject, &local);
return;
}
}
QV4::Object::virtualAdvanceIterator(m, it, name, index, p, attributes);
}
struct QObjectWrapperOwnPropertyKeyIterator : ObjectOwnPropertyKeyIterator
{
int propertyIndex = 0;
~QObjectWrapperOwnPropertyKeyIterator() override = default;
PropertyKey next(const Object *o, Property *pd = nullptr, PropertyAttributes *attrs = nullptr) override;
};
PropertyKey QObjectWrapperOwnPropertyKeyIterator::next(const Object *o, Property *pd, PropertyAttributes *attrs)
{
// Used to block access to QObject::destroyed() and QObject::deleteLater() from QML
static const int destroyedIdx1 = QObject::staticMetaObject.indexOfSignal("destroyed(QObject*)");
static const int destroyedIdx2 = QObject::staticMetaObject.indexOfSignal("destroyed()");
static const int deleteLaterIdx = QObject::staticMetaObject.indexOfSlot("deleteLater()");
const QObjectWrapper *that = static_cast<const QObjectWrapper*>(o);
QObject *thatObject = that->d()->object();
if (thatObject && !QQmlData::wasDeleted(thatObject)) {
const QMetaObject *mo = thatObject->metaObject();
// These indices don't apply to gadgets, so don't block them.
const bool preventDestruction = mo->superClass() || mo == &QObject::staticMetaObject;
const int propertyCount = mo->propertyCount();
if (propertyIndex < propertyCount) {
ExecutionEngine *thatEngine = that->engine();
Scope scope(thatEngine);
const QMetaProperty property = mo->property(propertyIndex);
ScopedString propName(scope, thatEngine->newString(QString::fromUtf8(property.name())));
++propertyIndex;
if (attrs)
*attrs= QV4::Attr_Data;
if (pd) {
QQmlPropertyData local;
local.load(property);
pd->value = that->getProperty(thatEngine, thatObject, &local);
}
return propName->toPropertyKey();
}
const int methodCount = mo->methodCount();
while (propertyIndex < propertyCount + methodCount) {
Q_ASSERT(propertyIndex >= propertyCount);
int index = propertyIndex - propertyCount;
const QMetaMethod method = mo->method(index);
++propertyIndex;
if (method.access() == QMetaMethod::Private || (preventDestruction && (index == deleteLaterIdx || index == destroyedIdx1 || index == destroyedIdx2)))
continue;
ExecutionEngine *thatEngine = that->engine();
Scope scope(thatEngine);
ScopedString methodName(scope, thatEngine->newString(QString::fromUtf8(method.name())));
if (attrs)
*attrs = QV4::Attr_Data;
if (pd) {
QQmlPropertyData local;
local.load(method);
pd->value = that->getProperty(thatEngine, thatObject, &local);
}
return methodName->toPropertyKey();
}
}
return ObjectOwnPropertyKeyIterator::next(o, pd, attrs);
}
OwnPropertyKeyIterator *QObjectWrapper::virtualOwnPropertyKeys(const Object *)
{
return new QObjectWrapperOwnPropertyKeyIterator;
}
namespace QV4 {
struct QObjectSlotDispatcher : public QtPrivate::QSlotObjectBase
{
QV4::PersistentValue function;
QV4::PersistentValue thisObject;
int signalIndex;
QObjectSlotDispatcher()
: QtPrivate::QSlotObjectBase(&impl)
, signalIndex(-1)
{}
static void impl(int which, QSlotObjectBase *this_, QObject *r, void **metaArgs, bool *ret)
{
switch (which) {
case Destroy: {
delete static_cast<QObjectSlotDispatcher*>(this_);
}
break;
case Call: {
QObjectSlotDispatcher *This = static_cast<QObjectSlotDispatcher*>(this_);
QV4::ExecutionEngine *v4 = This->function.engine();
// Might be that we're still connected to a signal that's emitted long
// after the engine died. We don't track connections in a global list, so
// we need this safeguard.
if (!v4)
break;
QQmlMetaObject::ArgTypeStorage storage;
int *argsTypes = QQmlMetaObject(r).methodParameterTypes(This->signalIndex, &storage, nullptr);
int argCount = argsTypes ? argsTypes[0]:0;
QV4::Scope scope(v4);
QV4::ScopedFunctionObject f(scope, This->function.value());
QV4::JSCallData jsCallData(scope, argCount);
*jsCallData->thisObject = This->thisObject.isUndefined() ? v4->globalObject->asReturnedValue() : This->thisObject.value();
for (int ii = 0; ii < argCount; ++ii) {
int type = argsTypes[ii + 1];
if (type == qMetaTypeId<QVariant>()) {
jsCallData->args[ii] = v4->fromVariant(*((QVariant *)metaArgs[ii + 1]));
} else {
jsCallData->args[ii] = v4->fromVariant(QVariant(type, metaArgs[ii + 1]));
}
}
f->call(jsCallData);
if (scope.hasException()) {
QQmlError error = v4->catchExceptionAsQmlError();
if (error.description().isEmpty()) {
QV4::ScopedString name(scope, f->name());
error.setDescription(QStringLiteral("Unknown exception occurred during evaluation of connected function: %1").arg(name->toQString()));
}
if (QQmlEngine *qmlEngine = v4->qmlEngine()) {
QQmlEnginePrivate::get(qmlEngine)->warning(error);
} else {
QMessageLogger(error.url().toString().toLatin1().constData(),
error.line(), nullptr).warning().noquote()
<< error.toString();
}
}
}
break;
case Compare: {
QObjectSlotDispatcher *connection = static_cast<QObjectSlotDispatcher*>(this_);
if (connection->function.isUndefined()) {
*ret = false;
return;
}
// This is tricky. Normally the metaArgs[0] pointer is a pointer to the _function_
// for the new-style QObject::connect. Here we use the engine pointer as sentinel
// to distinguish those type of QSlotObjectBase connections from our QML connections.
QV4::ExecutionEngine *v4 = reinterpret_cast<QV4::ExecutionEngine*>(metaArgs[0]);
if (v4 != connection->function.engine()) {
*ret = false;
return;
}
QV4::Scope scope(v4);
QV4::ScopedValue function(scope, *reinterpret_cast<QV4::Value*>(metaArgs[1]));
QV4::ScopedValue thisObject(scope, *reinterpret_cast<QV4::Value*>(metaArgs[2]));
QObject *receiverToDisconnect = reinterpret_cast<QObject*>(metaArgs[3]);
int slotIndexToDisconnect = *reinterpret_cast<int*>(metaArgs[4]);
if (slotIndexToDisconnect != -1) {
// This is a QObject function wrapper
if (connection->thisObject.isUndefined() == thisObject->isUndefined() &&
(connection->thisObject.isUndefined() || RuntimeHelpers::strictEqual(*connection->thisObject.valueRef(), thisObject))) {
QV4::ScopedFunctionObject f(scope, connection->function.value());
QPair<QObject *, int> connectedFunctionData = QObjectMethod::extractQtMethod(f);
if (connectedFunctionData.first == receiverToDisconnect &&
connectedFunctionData.second == slotIndexToDisconnect) {
*ret = true;
return;
}
}
} else {
// This is a normal JS function
if (RuntimeHelpers::strictEqual(*connection->function.valueRef(), function) &&
connection->thisObject.isUndefined() == thisObject->isUndefined() &&
(connection->thisObject.isUndefined() || RuntimeHelpers::strictEqual(*connection->thisObject.valueRef(), thisObject))) {
*ret = true;
return;
}
}
*ret = false;
}
break;
case NumOperations:
break;
}
};
};
} // namespace QV4
ReturnedValue QObjectWrapper::method_connect(const FunctionObject *b, const Value *thisObject, const Value *argv, int argc)
{
QV4::Scope scope(b);
if (argc == 0)
THROW_GENERIC_ERROR("Function.prototype.connect: no arguments given");
QPair<QObject *, int> signalInfo = extractQtSignal(*thisObject);
QObject *signalObject = signalInfo.first;
int signalIndex = signalInfo.second; // in method range, not signal range!
if (signalIndex < 0)
THROW_GENERIC_ERROR("Function.prototype.connect: this object is not a signal");
if (!signalObject)
THROW_GENERIC_ERROR("Function.prototype.connect: cannot connect to deleted QObject");
if (signalObject->metaObject()->method(signalIndex).methodType() != QMetaMethod::Signal)
THROW_GENERIC_ERROR("Function.prototype.connect: this object is not a signal");
QV4::ScopedFunctionObject f(scope);
QV4::ScopedValue object (scope, QV4::Encode::undefined());
if (argc == 1) {
f = argv[0];
} else if (argc >= 2) {
object = argv[0];
f = argv[1];
}
if (!f)
THROW_GENERIC_ERROR("Function.prototype.connect: target is not a function");
if (!object->isUndefined() && !object->isObject())
THROW_GENERIC_ERROR("Function.prototype.connect: target this is not an object");
QV4::QObjectSlotDispatcher *slot = new QV4::QObjectSlotDispatcher;
slot->signalIndex = signalIndex;
slot->thisObject.set(scope.engine, object);
slot->function.set(scope.engine, f);
if (QQmlData *ddata = QQmlData::get(signalObject)) {
if (QQmlPropertyCache *propertyCache = ddata->propertyCache) {
QQmlPropertyPrivate::flushSignal(signalObject, propertyCache->methodIndexToSignalIndex(signalIndex));
}
}
QObjectPrivate::connect(signalObject, signalIndex, slot, Qt::AutoConnection);
RETURN_UNDEFINED();
}
ReturnedValue QObjectWrapper::method_disconnect(const FunctionObject *b, const Value *thisObject, const Value *argv, int argc)
{
QV4::Scope scope(b);
if (argc == 0)
THROW_GENERIC_ERROR("Function.prototype.disconnect: no arguments given");
QPair<QObject *, int> signalInfo = extractQtSignal(*thisObject);
QObject *signalObject = signalInfo.first;
int signalIndex = signalInfo.second;
if (signalIndex == -1)
THROW_GENERIC_ERROR("Function.prototype.disconnect: this object is not a signal");
if (!signalObject)
THROW_GENERIC_ERROR("Function.prototype.disconnect: cannot disconnect from deleted QObject");
if (signalIndex < 0 || signalObject->metaObject()->method(signalIndex).methodType() != QMetaMethod::Signal)
THROW_GENERIC_ERROR("Function.prototype.disconnect: this object is not a signal");
QV4::ScopedFunctionObject functionValue(scope);
QV4::ScopedValue functionThisValue(scope, QV4::Encode::undefined());
if (argc == 1) {
functionValue = argv[0];
} else if (argc >= 2) {
functionThisValue = argv[0];
functionValue = argv[1];
}
if (!functionValue)
THROW_GENERIC_ERROR("Function.prototype.disconnect: target is not a function");
if (!functionThisValue->isUndefined() && !functionThisValue->isObject())
THROW_GENERIC_ERROR("Function.prototype.disconnect: target this is not an object");
QPair<QObject *, int> functionData = QObjectMethod::extractQtMethod(functionValue);
void *a[] = {
scope.engine,
functionValue.ptr,
functionThisValue.ptr,
functionData.first,
&functionData.second
};
QObjectPrivate::disconnect(signalObject, signalIndex, reinterpret_cast<void**>(&a));
RETURN_UNDEFINED();
}
static void markChildQObjectsRecursively(QObject *parent, QV4::MarkStack *markStack)
{
const QObjectList &children = parent->children();
for (int i = 0; i < children.count(); ++i) {
QObject *child = children.at(i);
if (!child)
continue;
QObjectWrapper::markWrapper(child, markStack);
markChildQObjectsRecursively(child, markStack);
}
}
void Heap::QObjectWrapper::markObjects(Heap::Base *that, QV4::MarkStack *markStack)
{
QObjectWrapper *This = static_cast<QObjectWrapper *>(that);
if (QObject *o = This->object()) {
QQmlVMEMetaObject *vme = QQmlVMEMetaObject::get(o);
if (vme)
vme->mark(markStack);
// Children usually don't need to be marked, the gc keeps them alive.
// But in the rare case of a "floating" QObject without a parent that
// _gets_ marked (we've been called here!) then we also need to
// propagate the marking down to the children recursively.
if (!o->parent())
markChildQObjectsRecursively(o, markStack);
}
Object::markObjects(that, markStack);
}
void QObjectWrapper::destroyObject(bool lastCall)
{
Heap::QObjectWrapper *h = d();
if (!h->internalClass)
return; // destroyObject already got called
if (h->object()) {
QQmlData *ddata = QQmlData::get(h->object(), false);
if (ddata) {
if (!h->object()->parent() && !ddata->indestructible) {
if (ddata && ddata->ownContext) {
Q_ASSERT(ddata->ownContext == ddata->context);
ddata->ownContext->emitDestruction();
ddata->ownContext = nullptr;
ddata->context = nullptr;
}
// This object is notionally destroyed now
ddata->isQueuedForDeletion = true;
if (lastCall)
delete h->object();
else
h->object()->deleteLater();
} else {
// If the object is C++-owned, we still have to release the weak reference we have
// to it.
ddata->jsWrapper.clear();
if (lastCall && ddata->propertyCache) {
ddata->propertyCache->release();
ddata->propertyCache = nullptr;
}
}
}
}
h->~Data();
}
DEFINE_OBJECT_VTABLE(QObjectWrapper);
2011-05-11 07:20:40 +00:00
namespace {
template<typename A, typename B, typename C, typename D, typename E,
typename F, typename G, typename H>
class MaxSizeOf8 {
template<typename Z, typename X>
struct SMax {
char dummy[sizeof(Z) > sizeof(X) ? sizeof(Z) : sizeof(X)];
};
public:
static const size_t Size = sizeof(SMax<A, SMax<B, SMax<C, SMax<D, SMax<E, SMax<F, SMax<G, H> > > > > > >);
};
struct CallArgument {
inline CallArgument();
inline ~CallArgument();
2011-05-11 07:20:40 +00:00
inline void *dataPtr();
inline void initAsType(int type);
inline void fromValue(int type, ExecutionEngine *, const QV4::Value &);
inline ReturnedValue toValue(ExecutionEngine *);
2011-05-11 07:20:40 +00:00
private:
CallArgument(const CallArgument &);
2011-05-11 07:20:40 +00:00
inline void cleanup();
template <class T, class M>
void fromContainerValue(const QV4::Object *object, int type, M CallArgument::*member, bool &queryEngine);
union {
float floatValue;
double doubleValue;
quint32 intValue;
bool boolValue;
QObject *qobjectPtr;
std::vector<int> *stdVectorIntPtr;
std::vector<qreal> *stdVectorRealPtr;
std::vector<bool> *stdVectorBoolPtr;
std::vector<QString> *stdVectorQStringPtr;
std::vector<QUrl> *stdVectorQUrlPtr;
std::vector<QModelIndex> *stdVectorQModelIndexPtr;
char allocData[MaxSizeOf8<QVariant,
QString,
QList<QObject *>,
QJSValue,
QQmlV4Handle,
QJsonArray,
QJsonObject,
QJsonValue>::Size];
qint64 q_for_alignment;
};
// Pointers to allocData
union {
QString *qstringPtr;
QByteArray *qbyteArrayPtr;
QVariant *qvariantPtr;
QList<QObject *> *qlistPtr;
QJSValue *qjsValuePtr;
QQmlV4Handle *handlePtr;
QJsonArray *jsonArrayPtr;
QJsonObject *jsonObjectPtr;
QJsonValue *jsonValuePtr;
};
2011-05-11 07:20:40 +00:00
int type;
};
}
static QV4::ReturnedValue CallMethod(const QQmlObjectOrGadget &object, int index, int returnType, int argCount,
int *argTypes, QV4::ExecutionEngine *engine, QV4::CallData *callArgs,
QMetaObject::Call callType = QMetaObject::InvokeMetaMethod)
2011-05-11 07:20:40 +00:00
{
if (argCount > 0) {
// Convert all arguments.
QVarLengthArray<CallArgument, 9> args(argCount + 1);
2011-05-11 07:20:40 +00:00
args[0].initAsType(returnType);
for (int ii = 0; ii < argCount; ++ii)
args[ii + 1].fromValue(argTypes[ii], engine, callArgs->args[ii]);
2011-05-11 07:20:40 +00:00
QVarLengthArray<void *, 9> argData(args.count());
for (int ii = 0; ii < args.count(); ++ii)
argData[ii] = args[ii].dataPtr();
object.metacall(callType, index, argData.data());
2011-05-11 07:20:40 +00:00
return args[0].toValue(engine);
2011-05-11 07:20:40 +00:00
} else if (returnType != QMetaType::Void) {
CallArgument arg;
2011-05-11 07:20:40 +00:00
arg.initAsType(returnType);
void *args[] = { arg.dataPtr() };
object.metacall(callType, index, args);
2011-05-11 07:20:40 +00:00
return arg.toValue(engine);
2011-05-11 07:20:40 +00:00
} else {
void *args[] = { nullptr };
object.metacall(callType, index, args);
return Encode::undefined();
2011-05-11 07:20:40 +00:00
}
}
/*
Returns the match score for converting \a actual to be of type \a conversionType. A
2011-05-11 07:20:40 +00:00
zero score means "perfect match" whereas a higher score is worse.
The conversion table is copied out of the \l QScript::callQtMethod() function.
2011-05-11 07:20:40 +00:00
*/
static int MatchScore(const QV4::Value &actual, int conversionType)
2011-05-11 07:20:40 +00:00
{
if (actual.isNumber()) {
2011-05-11 07:20:40 +00:00
switch (conversionType) {
case QMetaType::Double:
return 0;
case QMetaType::Float:
return 1;
case QMetaType::LongLong:
case QMetaType::ULongLong:
return 2;
case QMetaType::Long:
case QMetaType::ULong:
return 3;
case QMetaType::Int:
case QMetaType::UInt:
return 4;
case QMetaType::Short:
case QMetaType::UShort:
return 5;
break;
case QMetaType::Char:
case QMetaType::UChar:
return 6;
case QMetaType::QJsonValue:
return 5;
2011-05-11 07:20:40 +00:00
default:
return 10;
}
} else if (actual.isString()) {
2011-05-11 07:20:40 +00:00
switch (conversionType) {
case QMetaType::QString:
return 0;
case QMetaType::QJsonValue:
return 5;
2011-05-11 07:20:40 +00:00
default:
return 10;
}
} else if (actual.isBoolean()) {
2011-05-11 07:20:40 +00:00
switch (conversionType) {
case QMetaType::Bool:
return 0;
case QMetaType::QJsonValue:
return 5;
2011-05-11 07:20:40 +00:00
default:
return 10;
}
} else if (actual.as<DateObject>()) {
2011-05-11 07:20:40 +00:00
switch (conversionType) {
case QMetaType::QDateTime:
return 0;
case QMetaType::QDate:
return 1;
case QMetaType::QTime:
return 2;
default:
return 10;
}
} else if (actual.as<QV4::RegExpObject>()) {
2011-05-11 07:20:40 +00:00
switch (conversionType) {
case QMetaType::QRegExp:
return 0;
default:
return 10;
}
} else if (actual.as<ArrayBuffer>()) {
switch (conversionType) {
case QMetaType::QByteArray:
return 0;
default:
return 10;
}
} else if (actual.as<ArrayObject>()) {
2011-05-11 07:20:40 +00:00
switch (conversionType) {
case QMetaType::QJsonArray:
return 3;
2011-05-11 07:20:40 +00:00
case QMetaType::QStringList:
case QMetaType::QVariantList:
return 5;
case QMetaType::QVector4D:
case QMetaType::QMatrix4x4:
return 6;
case QMetaType::QVector3D:
return 7;
2011-05-11 07:20:40 +00:00
default:
return 10;
}
} else if (actual.isNull()) {
2011-05-11 07:20:40 +00:00
switch (conversionType) {
case QMetaType::Nullptr:
2011-05-11 07:20:40 +00:00
case QMetaType::VoidStar:
case QMetaType::QObjectStar:
case QMetaType::QJsonValue:
2011-05-11 07:20:40 +00:00
return 0;
default: {
const char *typeName = QMetaType::typeName(conversionType);
if (typeName && typeName[strlen(typeName) - 1] == '*')
2011-05-11 07:20:40 +00:00
return 0;
else
return 10;
}
2011-05-11 07:20:40 +00:00
}
} else if (const QV4::Object *obj = actual.as<QV4::Object>()) {
if (obj->as<QV4::VariantObject>()) {
if (conversionType == qMetaTypeId<QVariant>())
return 0;
if (obj->engine()->toVariant(actual, -1).userType() == conversionType)
return 0;
else
return 10;
}
if (obj->as<QObjectWrapper>()) {
2011-05-11 07:20:40 +00:00
switch (conversionType) {
case QMetaType::QObjectStar:
return 0;
default:
return 10;
}
}
if (obj->as<QV4::QQmlValueTypeWrapper>()) {
if (obj->engine()->toVariant(actual, -1).userType() == conversionType)
return 0;
return 10;
} else if (conversionType == QMetaType::QJsonObject) {
return 5;
} else if (conversionType == qMetaTypeId<QJSValue>()) {
return 0;
2011-05-11 07:20:40 +00:00
} else {
return 10;
}
} else {
return 10;
}
}
static inline int QMetaObject_methods(const QMetaObject *metaObject)
{
struct Private
{
int revision;
int className;
int classInfoCount, classInfoData;
int methodCount, methodData;
};
return reinterpret_cast<const Private *>(metaObject->d.data)->methodCount;
}
/*
2011-05-11 07:20:40 +00:00
Returns the next related method, if one, or 0.
*/
static const QQmlPropertyData * RelatedMethod(const QQmlObjectOrGadget &object,
const QQmlPropertyData *current,
QQmlPropertyData &dummy,
const QQmlPropertyCache *propertyCache)
2011-05-11 07:20:40 +00:00
{
if (!current->isOverload())
return nullptr;
2011-05-11 07:20:40 +00:00
Q_ASSERT(!current->overrideIndexIsProperty());
if (propertyCache) {
return propertyCache->method(current->overrideIndex());
2011-05-11 07:20:40 +00:00
} else {
const QMetaObject *mo = object.metaObject();
2011-05-11 07:20:40 +00:00
int methodOffset = mo->methodCount() - QMetaObject_methods(mo);
while (methodOffset > current->overrideIndex()) {
2011-05-11 07:20:40 +00:00
mo = mo->superClass();
methodOffset -= QMetaObject_methods(mo);
}
// If we've been called before with the same override index, then
// we can't go any further...
if (&dummy == current && dummy.coreIndex() == current->overrideIndex())
return nullptr;
QMetaMethod method = mo->method(current->overrideIndex());
2011-05-11 07:20:40 +00:00
dummy.load(method);
2011-05-11 07:20:40 +00:00
// Look for overloaded methods
QByteArray methodName = method.name();
for (int ii = current->overrideIndex() - 1; ii >= methodOffset; --ii) {
if (methodName == mo->method(ii).name()) {
dummy.setOverload(true);
dummy.setOverrideIndexIsProperty(0);
dummy.setOverrideIndex(ii);
2011-05-11 07:20:40 +00:00
return &dummy;
}
}
return &dummy;
}
}
static QV4::ReturnedValue CallPrecise(const QQmlObjectOrGadget &object, const QQmlPropertyData &data,
QV4::ExecutionEngine *engine, QV4::CallData *callArgs,
QMetaObject::Call callType = QMetaObject::InvokeMetaMethod)
2011-05-11 07:20:40 +00:00
{
QByteArray unknownTypeError;
int returnType = object.methodReturnType(data, &unknownTypeError);
if (returnType == QMetaType::UnknownType) {
return engine->throwError(QLatin1String("Unknown method return type: ")
+ QLatin1String(unknownTypeError));
}
if (data.hasArguments()) {
2011-05-11 07:20:40 +00:00
int *args = nullptr;
QQmlMetaObject::ArgTypeStorage storage;
if (data.isConstructor())
args = static_cast<const QQmlStaticMetaObject&>(object).constructorParameterTypes(
data.coreIndex(), &storage, &unknownTypeError);
else
args = object.methodParameterTypes(data.coreIndex(), &storage, &unknownTypeError);
if (!args) {
return engine->throwError(QLatin1String("Unknown method parameter type: ")
+ QLatin1String(unknownTypeError));
2011-05-11 07:20:40 +00:00
}
if (args[0] > callArgs->argc()) {
2011-05-11 07:20:40 +00:00
QString error = QLatin1String("Insufficient arguments");
return engine->throwError(error);
2011-05-11 07:20:40 +00:00
}
return CallMethod(object, data.coreIndex(), returnType, args[0], args + 1, engine, callArgs, callType);
2011-05-11 07:20:40 +00:00
} else {
return CallMethod(object, data.coreIndex(), returnType, 0, nullptr, engine, callArgs, callType);
2011-05-11 07:20:40 +00:00
}
}
/*
2011-05-11 07:20:40 +00:00
Resolve the overloaded method to call. The algorithm works conceptually like this:
1. Resolve the set of overloads it is *possible* to call.
Impossible overloads include those that have too many parameters or have parameters
of unknown type.
2. Filter the set of overloads to only contain those with the closest number of
2011-05-11 07:20:40 +00:00
parameters.
For example, if we are called with 3 parameters and there are 2 overloads that
take 2 parameters and one that takes 3, eliminate the 2 parameter overloads.
3. Find the best remaining overload based on its match score.
2011-05-11 07:20:40 +00:00
If two or more overloads have the same match score, call the last one. The match
score is constructed by adding the matchScore() result for each of the parameters.
*/
static QV4::ReturnedValue CallOverloaded(const QQmlObjectOrGadget &object, const QQmlPropertyData &data,
QV4::ExecutionEngine *engine, QV4::CallData *callArgs, const QQmlPropertyCache *propertyCache,
QMetaObject::Call callType = QMetaObject::InvokeMetaMethod)
2011-05-11 07:20:40 +00:00
{
int argumentCount = callArgs->argc();
2011-05-11 07:20:40 +00:00
QQmlPropertyData best;
2011-05-11 07:20:40 +00:00
int bestParameterScore = INT_MAX;
int bestMatchScore = INT_MAX;
QQmlPropertyData dummy;
const QQmlPropertyData *attempt = &data;
2011-05-11 07:20:40 +00:00
QV4::Scope scope(engine);
QV4::ScopedValue v(scope);
2011-05-11 07:20:40 +00:00
do {
QQmlMetaObject::ArgTypeStorage storage;
int methodArgumentCount = 0;
int *methodArgTypes = nullptr;
if (attempt->hasArguments()) {
int *args = object.methodParameterTypes(attempt->coreIndex(), &storage, nullptr);
if (!args) // Must be an unknown argument
continue;
2011-05-11 07:20:40 +00:00
methodArgumentCount = args[0];
methodArgTypes = args + 1;
}
2011-05-11 07:20:40 +00:00
if (methodArgumentCount > argumentCount)
continue; // We don't have sufficient arguments to call this method
int methodParameterScore = argumentCount - methodArgumentCount;
if (methodParameterScore > bestParameterScore)
continue; // We already have a better option
int methodMatchScore = 0;
for (int ii = 0; ii < methodArgumentCount; ++ii)
methodMatchScore += MatchScore((v = callArgs->args[ii]), methodArgTypes[ii]);
2011-05-11 07:20:40 +00:00
if (bestParameterScore > methodParameterScore || bestMatchScore > methodMatchScore) {
best = *attempt;
2011-05-11 07:20:40 +00:00
bestParameterScore = methodParameterScore;
bestMatchScore = methodMatchScore;
}
if (bestParameterScore == 0 && bestMatchScore == 0)
break; // We can't get better than that
} while ((attempt = RelatedMethod(object, attempt, dummy, propertyCache)) != nullptr);
2011-05-11 07:20:40 +00:00
if (best.isValid()) {
return CallPrecise(object, best, engine, callArgs, callType);
2011-05-11 07:20:40 +00:00
} else {
QString error = QLatin1String("Unable to determine callable overload. Candidates are:");
const QQmlPropertyData *candidate = &data;
2011-05-11 07:20:40 +00:00
while (candidate) {
error += QLatin1String("\n ") +
QString::fromUtf8(object.metaObject()->method(candidate->coreIndex())
.methodSignature());
candidate = RelatedMethod(object, candidate, dummy, propertyCache);
2011-05-11 07:20:40 +00:00
}
return engine->throwError(error);
2011-05-11 07:20:40 +00:00
}
}
CallArgument::CallArgument()
: type(QVariant::Invalid)
2011-05-11 07:20:40 +00:00
{
}
CallArgument::~CallArgument()
2011-05-11 07:20:40 +00:00
{
cleanup();
}
void CallArgument::cleanup()
2011-05-11 07:20:40 +00:00
{
if (type == QMetaType::QString) {
qstringPtr->~QString();
} else if (type == QMetaType::QByteArray) {
qbyteArrayPtr->~QByteArray();
} else if (type == -1 || type == QMetaType::QVariant) {
qvariantPtr->~QVariant();
} else if (type == qMetaTypeId<QJSValue>()) {
qjsValuePtr->~QJSValue();
2011-05-11 07:20:40 +00:00
} else if (type == qMetaTypeId<QList<QObject *> >()) {
qlistPtr->~QList<QObject *>();
} else if (type == QMetaType::QJsonArray) {
jsonArrayPtr->~QJsonArray();
} else if (type == QMetaType::QJsonObject) {
jsonObjectPtr->~QJsonObject();
} else if (type == QMetaType::QJsonValue) {
jsonValuePtr->~QJsonValue();
}
2011-05-11 07:20:40 +00:00
}
void *CallArgument::dataPtr()
2011-05-11 07:20:40 +00:00
{
if (type == -1)
return qvariantPtr->data();
else if (type == qMetaTypeId<std::vector<int>>())
return stdVectorIntPtr;
else if (type == qMetaTypeId<std::vector<qreal>>())
return stdVectorRealPtr;
else if (type == qMetaTypeId<std::vector<bool>>())
return stdVectorBoolPtr;
else if (type == qMetaTypeId<std::vector<QString>>())
return stdVectorQStringPtr;
else if (type == qMetaTypeId<std::vector<QUrl>>())
return stdVectorQUrlPtr;
else if (type == qMetaTypeId<std::vector<QModelIndex>>())
return stdVectorQModelIndexPtr;
else if (type != 0)
return (void *)&allocData;
return nullptr;
2011-05-11 07:20:40 +00:00
}
void CallArgument::initAsType(int callType)
2011-05-11 07:20:40 +00:00
{
if (type != 0) { cleanup(); type = 0; }
if (callType == QMetaType::UnknownType || callType == QMetaType::Void) return;
2011-05-11 07:20:40 +00:00
if (callType == qMetaTypeId<QJSValue>()) {
qjsValuePtr = new (&allocData) QJSValue();
2011-05-11 07:20:40 +00:00
type = callType;
} else if (callType == QMetaType::Int ||
callType == QMetaType::UInt ||
callType == QMetaType::Bool ||
callType == QMetaType::Double ||
callType == QMetaType::Float) {
type = callType;
} else if (callType == QMetaType::QObjectStar) {
qobjectPtr = nullptr;
2011-05-11 07:20:40 +00:00
type = callType;
} else if (callType == QMetaType::QString) {
qstringPtr = new (&allocData) QString();
2011-05-11 07:20:40 +00:00
type = callType;
} else if (callType == QMetaType::QVariant) {
2011-05-11 07:20:40 +00:00
type = callType;
qvariantPtr = new (&allocData) QVariant();
2011-05-11 07:20:40 +00:00
} else if (callType == qMetaTypeId<QList<QObject *> >()) {
type = callType;
qlistPtr = new (&allocData) QList<QObject *>();
} else if (callType == qMetaTypeId<QQmlV4Handle>()) {
2011-05-11 07:20:40 +00:00
type = callType;
handlePtr = new (&allocData) QQmlV4Handle;
} else if (callType == QMetaType::QJsonArray) {
type = callType;
jsonArrayPtr = new (&allocData) QJsonArray();
} else if (callType == QMetaType::QJsonObject) {
type = callType;
jsonObjectPtr = new (&allocData) QJsonObject();
} else if (callType == QMetaType::QJsonValue) {
type = callType;
jsonValuePtr = new (&allocData) QJsonValue();
2011-05-11 07:20:40 +00:00
} else {
type = -1;
qvariantPtr = new (&allocData) QVariant(callType, (void *)nullptr);
2011-05-11 07:20:40 +00:00
}
}
#if QT_CONFIG(qml_sequence_object)
template <class T, class M>
void CallArgument::fromContainerValue(const QV4::Object *object, int callType, M CallArgument::*member, bool &queryEngine)
{
if (object && object->isListType()) {
T* ptr = static_cast<T*>(QV4::SequencePrototype::getRawContainerPtr(object, callType));
if (ptr) {
(this->*member) = ptr;
type = callType;
queryEngine = false;
}
}
}
#endif
void CallArgument::fromValue(int callType, QV4::ExecutionEngine *engine, const QV4::Value &value)
2011-05-11 07:20:40 +00:00
{
if (type != 0) {
cleanup();
type = 0;
}
QV4::Scope scope(engine);
2011-05-11 07:20:40 +00:00
bool queryEngine = false;
if (callType == qMetaTypeId<QJSValue>()) {
qjsValuePtr = new (&allocData) QJSValue(scope.engine, value.asReturnedValue());
type = qMetaTypeId<QJSValue>();
2011-05-11 07:20:40 +00:00
} else if (callType == QMetaType::Int) {
intValue = quint32(value.toInt32());
2011-05-11 07:20:40 +00:00
type = callType;
} else if (callType == QMetaType::UInt) {
intValue = quint32(value.toUInt32());
2011-05-11 07:20:40 +00:00
type = callType;
} else if (callType == QMetaType::Bool) {
boolValue = value.toBoolean();
2011-05-11 07:20:40 +00:00
type = callType;
} else if (callType == QMetaType::Double) {
doubleValue = double(value.toNumber());
2011-05-11 07:20:40 +00:00
type = callType;
} else if (callType == QMetaType::Float) {
floatValue = float(value.toNumber());
2011-05-11 07:20:40 +00:00
type = callType;
} else if (callType == QMetaType::QString) {
if (value.isNull() || value.isUndefined())
qstringPtr = new (&allocData) QString();
2011-05-11 07:20:40 +00:00
else
qstringPtr = new (&allocData) QString(value.toQStringNoThrow());
2011-05-11 07:20:40 +00:00
type = callType;
} else if (callType == QMetaType::QObjectStar) {
qobjectPtr = nullptr;
if (const QV4::QObjectWrapper *qobjectWrapper = value.as<QV4::QObjectWrapper>())
qobjectPtr = qobjectWrapper->object();
else if (const QV4::QQmlTypeWrapper *qmlTypeWrapper = value.as<QV4::QQmlTypeWrapper>())
queryEngine = qmlTypeWrapper->isSingleton();
2011-05-11 07:20:40 +00:00
type = callType;
} else if (callType == qMetaTypeId<QVariant>()) {
qvariantPtr = new (&allocData) QVariant(scope.engine->toVariant(value, -1));
2011-05-11 07:20:40 +00:00
type = callType;
} else if (callType == qMetaTypeId<QList<QObject*> >()) {
qlistPtr = new (&allocData) QList<QObject *>();
QV4::ScopedArrayObject array(scope, value);
if (array) {
Scoped<QV4::QObjectWrapper> qobjectWrapper(scope);
uint length = array->getLength();
for (uint ii = 0; ii < length; ++ii) {
QObject *o = nullptr;
qobjectWrapper = array->get(ii);
if (!!qobjectWrapper)
o = qobjectWrapper->object();
qlistPtr->append(o);
}
2011-05-11 07:20:40 +00:00
} else {
QObject *o = nullptr;
if (const QV4::QObjectWrapper *qobjectWrapper = value.as<QV4::QObjectWrapper>())
o = qobjectWrapper->object();
qlistPtr->append(o);
2011-05-11 07:20:40 +00:00
}
type = callType;
} else if (callType == qMetaTypeId<QQmlV4Handle>()) {
handlePtr = new (&allocData) QQmlV4Handle(value.asReturnedValue());
2011-05-11 07:20:40 +00:00
type = callType;
} else if (callType == QMetaType::QJsonArray) {
QV4::ScopedArrayObject a(scope, value);
jsonArrayPtr = new (&allocData) QJsonArray(QV4::JsonObject::toJsonArray(a));
type = callType;
} else if (callType == QMetaType::QJsonObject) {
QV4::ScopedObject o(scope, value);
jsonObjectPtr = new (&allocData) QJsonObject(QV4::JsonObject::toJsonObject(o));
type = callType;
} else if (callType == QMetaType::QJsonValue) {
jsonValuePtr = new (&allocData) QJsonValue(QV4::JsonObject::toJsonValue(value));
type = callType;
} else if (callType == QMetaType::Void) {
*qvariantPtr = QVariant();
#if QT_CONFIG(qml_sequence_object)
} else if (callType == qMetaTypeId<std::vector<int>>()
|| callType == qMetaTypeId<std::vector<qreal>>()
|| callType == qMetaTypeId<std::vector<bool>>()
|| callType == qMetaTypeId<std::vector<QString>>()
|| callType == qMetaTypeId<std::vector<QUrl>>()
|| callType == qMetaTypeId<std::vector<QModelIndex>>()) {
queryEngine = true;
const QV4::Object* object = value.as<QV4::Object>();
if (callType == qMetaTypeId<std::vector<int>>()) {
stdVectorIntPtr = nullptr;
fromContainerValue<std::vector<int>>(object, callType, &CallArgument::stdVectorIntPtr, queryEngine);
} else if (callType == qMetaTypeId<std::vector<qreal>>()) {
stdVectorRealPtr = nullptr;
fromContainerValue<std::vector<qreal>>(object, callType, &CallArgument::stdVectorRealPtr, queryEngine);
} else if (callType == qMetaTypeId<std::vector<bool>>()) {
stdVectorBoolPtr = nullptr;
fromContainerValue<std::vector<bool>>(object, callType, &CallArgument::stdVectorBoolPtr, queryEngine);
} else if (callType == qMetaTypeId<std::vector<QString>>()) {
stdVectorQStringPtr = nullptr;
fromContainerValue<std::vector<QString>>(object, callType, &CallArgument::stdVectorQStringPtr, queryEngine);
} else if (callType == qMetaTypeId<std::vector<QUrl>>()) {
stdVectorQUrlPtr = nullptr;
fromContainerValue<std::vector<QUrl>>(object, callType, &CallArgument::stdVectorQUrlPtr, queryEngine);
} else if (callType == qMetaTypeId<std::vector<QModelIndex>>()) {
stdVectorQModelIndexPtr = nullptr;
fromContainerValue<std::vector<QModelIndex>>(object, callType, &CallArgument::stdVectorQModelIndexPtr, queryEngine);
}
#endif
2011-05-11 07:20:40 +00:00
} else {
queryEngine = true;
}
if (queryEngine) {
qvariantPtr = new (&allocData) QVariant();
2011-05-11 07:20:40 +00:00
type = -1;
QQmlEnginePrivate *ep = engine->qmlEngine() ? QQmlEnginePrivate::get(engine->qmlEngine()) : nullptr;
QVariant v = scope.engine->toVariant(value, callType);
2011-05-11 07:20:40 +00:00
if (v.userType() == callType) {
*qvariantPtr = v;
} else if (v.canConvert(callType)) {
*qvariantPtr = v;
qvariantPtr->convert(callType);
2011-05-11 07:20:40 +00:00
} else {
QQmlMetaObject mo = ep ? ep->rawMetaObjectForType(callType) : QQmlMetaObject();
if (!mo.isNull()) {
QObject *obj = ep->toQObject(v);
if (obj != nullptr && !QQmlMetaObject::canConvert(obj, mo))
obj = nullptr;
*qvariantPtr = QVariant(callType, &obj);
} else {
*qvariantPtr = QVariant(callType, (void *)nullptr);
}
2011-05-11 07:20:40 +00:00
}
}
}
QV4::ReturnedValue CallArgument::toValue(QV4::ExecutionEngine *engine)
2011-05-11 07:20:40 +00:00
{
QV4::Scope scope(engine);
if (type == qMetaTypeId<QJSValue>()) {
return QJSValuePrivate::convertedToValue(scope.engine, *qjsValuePtr);
2011-05-11 07:20:40 +00:00
} else if (type == QMetaType::Int) {
return QV4::Encode(int(intValue));
2011-05-11 07:20:40 +00:00
} else if (type == QMetaType::UInt) {
return QV4::Encode((uint)intValue);
2011-05-11 07:20:40 +00:00
} else if (type == QMetaType::Bool) {
return QV4::Encode(boolValue);
2011-05-11 07:20:40 +00:00
} else if (type == QMetaType::Double) {
return QV4::Encode(doubleValue);
2011-05-11 07:20:40 +00:00
} else if (type == QMetaType::Float) {
return QV4::Encode(floatValue);
2011-05-11 07:20:40 +00:00
} else if (type == QMetaType::QString) {
return QV4::Encode(engine->newString(*qstringPtr));
} else if (type == QMetaType::QByteArray) {
return QV4::Encode(engine->newArrayBuffer(*qbyteArrayPtr));
2011-05-11 07:20:40 +00:00
} else if (type == QMetaType::QObjectStar) {
QObject *object = qobjectPtr;
2011-05-11 07:20:40 +00:00
if (object)
QQmlData::get(object, true)->setImplicitDestructible();
return QV4::QObjectWrapper::wrap(scope.engine, object);
2011-05-11 07:20:40 +00:00
} else if (type == qMetaTypeId<QList<QObject *> >()) {
2011-06-09 04:40:44 +00:00
// XXX Can this be made more by using Array as a prototype and implementing
// directly against QList<QObject*>?
QList<QObject *> &list = *qlistPtr;
QV4::ScopedArrayObject array(scope, scope.engine->newArrayObject());
array->arrayReserve(list.count());
QV4::ScopedValue v(scope);
for (int ii = 0; ii < list.count(); ++ii)
array->arrayPut(ii, (v = QV4::QObjectWrapper::wrap(scope.engine, list.at(ii))));
array->setArrayLengthUnchecked(list.count());
return array.asReturnedValue();
} else if (type == qMetaTypeId<QQmlV4Handle>()) {
return *handlePtr;
} else if (type == QMetaType::QJsonArray) {
return QV4::JsonObject::fromJsonArray(scope.engine, *jsonArrayPtr);
} else if (type == QMetaType::QJsonObject) {
return QV4::JsonObject::fromJsonObject(scope.engine, *jsonObjectPtr);
} else if (type == QMetaType::QJsonValue) {
return QV4::JsonObject::fromJsonValue(scope.engine, *jsonValuePtr);
2011-05-11 07:20:40 +00:00
} else if (type == -1 || type == qMetaTypeId<QVariant>()) {
QVariant value = *qvariantPtr;
QV4::ScopedValue rv(scope, scope.engine->fromVariant(value));
QV4::Scoped<QV4::QObjectWrapper> qobjectWrapper(scope, rv);
if (!!qobjectWrapper) {
if (QObject *object = qobjectWrapper->object())
QQmlData::get(object, true)->setImplicitDestructible();
}
return rv->asReturnedValue();
2011-05-11 07:20:40 +00:00
} else {
return QV4::Encode::undefined();
2011-05-11 07:20:40 +00:00
}
}
ReturnedValue QObjectMethod::create(ExecutionContext *scope, QObject *object, int index)
{
Scope valueScope(scope);
Scoped<QObjectMethod> method(valueScope, valueScope.engine->memoryManager->allocate<QObjectMethod>(scope));
method->d()->setObject(object);
if (QQmlData *ddata = QQmlData::get(object))
method->d()->setPropertyCache(ddata->propertyCache);
method->d()->index = index;
return method.asReturnedValue();
}
ReturnedValue QObjectMethod::create(ExecutionContext *scope, const QQmlValueTypeWrapper *valueType, int index)
{
Scope valueScope(scope);
Scoped<QObjectMethod> method(valueScope, valueScope.engine->memoryManager->allocate<QObjectMethod>(scope));
method->d()->setPropertyCache(valueType->d()->propertyCache());
method->d()->index = index;
method->d()->valueTypeWrapper.set(valueScope.engine, valueType->d());
return method.asReturnedValue();
}
void Heap::QObjectMethod::init(QV4::ExecutionContext *scope)
{
Heap::FunctionObject::init(scope);
}
const QMetaObject *Heap::QObjectMethod::metaObject()
{
if (propertyCache())
return propertyCache()->createMetaObject();
return object()->metaObject();
}
QV4::ReturnedValue QObjectMethod::method_toString(QV4::ExecutionEngine *engine) const
{
QString result;
if (const QMetaObject *metaObject = d()->metaObject()) {
result += QString::fromUtf8(metaObject->className()) +
QLatin1String("(0x") + QString::number((quintptr)d()->object(),16);
if (d()->object()) {
QString objectName = d()->object()->objectName();
if (!objectName.isEmpty())
result += QLatin1String(", \"") + objectName + QLatin1Char('\"');
}
result += QLatin1Char(')');
} else {
result = QLatin1String("null");
}
return engine->newString(result)->asReturnedValue();
}
QV4::ReturnedValue QObjectMethod::method_destroy(QV4::ExecutionEngine *engine, const Value *args, int argc) const
{
if (!d()->object())
return Encode::undefined();
if (QQmlData::keepAliveDuringGarbageCollection(d()->object()))
return engine->throwError(QStringLiteral("Invalid attempt to destroy() an indestructible object"));
int delay = 0;
if (argc > 0)
delay = args[0].toUInt32();
if (delay > 0)
QTimer::singleShot(delay, d()->object(), SLOT(deleteLater()));
else
d()->object()->deleteLater();
return Encode::undefined();
}
ReturnedValue QObjectMethod::virtualCall(const FunctionObject *m, const Value *thisObject, const Value *argv, int argc)
{
const QObjectMethod *This = static_cast<const QObjectMethod*>(m);
return This->callInternal(thisObject, argv, argc);
}
ReturnedValue QObjectMethod::callInternal(const Value *thisObject, const Value *argv, int argc) const
{
ExecutionEngine *v4 = engine();
if (d()->index == DestroyMethod)
return method_destroy(v4, argv, argc);
else if (d()->index == ToStringMethod)
return method_toString(v4);
QQmlObjectOrGadget object(d()->object());
if (!d()->object()) {
if (!d()->valueTypeWrapper)
return Encode::undefined();
object = QQmlObjectOrGadget(d()->propertyCache(), d()->valueTypeWrapper->gadgetPtr);
}
QQmlPropertyData method;
if (d()->propertyCache()) {
QQmlPropertyData *data = d()->propertyCache()->method(d()->index);
if (!data)
return QV4::Encode::undefined();
method = *data;
} else {
const QMetaObject *mo = d()->object()->metaObject();
const QMetaMethod moMethod = mo->method(d()->index);
method.load(moMethod);
if (method.coreIndex() == -1)
return QV4::Encode::undefined();
// Look for overloaded methods
QByteArray methodName = moMethod.name();
const int methodOffset = mo->methodOffset();
for (int ii = d()->index - 1; ii >= methodOffset; --ii) {
if (methodName == mo->method(ii).name()) {
method.setOverload(true);
method.setOverrideIndexIsProperty(0);
method.setOverrideIndex(ii);
break;
}
}
}
Scope scope(v4);
JSCallData cData(scope, argc, argv, thisObject);
CallData *callData = cData.callData();
if (method.isV4Function()) {
QV4::ScopedValue rv(scope, QV4::Primitive::undefinedValue());
QQmlV4Function func(callData, rv, v4);
QQmlV4Function *funcptr = &func;
void *args[] = { nullptr, &funcptr };
object.metacall(QMetaObject::InvokeMetaMethod, method.coreIndex(), args);
return rv->asReturnedValue();
}
if (!method.isOverload()) {
return CallPrecise(object, method, v4, callData);
} else {
return CallOverloaded(object, method, v4, callData, d()->propertyCache());
}
}
DEFINE_OBJECT_VTABLE(QObjectMethod);
void Heap::QMetaObjectWrapper::init(const QMetaObject *metaObject)
{
FunctionObject::init();
this->metaObject = metaObject;
constructors = nullptr;
constructorCount = 0;
}
void Heap::QMetaObjectWrapper::destroy()
{
delete[] constructors;
}
void Heap::QMetaObjectWrapper::ensureConstructorsCache() {
const int count = metaObject->constructorCount();
if (constructorCount != count) {
delete[] constructors;
constructorCount = count;
if (count == 0) {
constructors = nullptr;
return;
}
constructors = new QQmlPropertyData[count];
for (int i = 0; i < count; ++i) {
QMetaMethod method = metaObject->constructor(i);
QQmlPropertyData &d = constructors[i];
d.load(method);
d.setCoreIndex(i);
}
}
}
ReturnedValue QMetaObjectWrapper::create(ExecutionEngine *engine, const QMetaObject* metaObject) {
QV4::Scope scope(engine);
Scoped<QMetaObjectWrapper> mo(scope, engine->memoryManager->allocate<QV4::QMetaObjectWrapper>(metaObject)->asReturnedValue());
mo->init(engine);
return mo->asReturnedValue();
}
void QMetaObjectWrapper::init(ExecutionEngine *) {
const QMetaObject & mo = *d()->metaObject;
for (int i = 0; i < mo.enumeratorCount(); i++) {
QMetaEnum Enum = mo.enumerator(i);
for (int k = 0; k < Enum.keyCount(); k++) {
const char* key = Enum.key(k);
const int value = Enum.value(k);
defineReadonlyProperty(QLatin1String(key), Primitive::fromInt32(value));
}
}
}
ReturnedValue QMetaObjectWrapper::virtualCallAsConstructor(const FunctionObject *f, const Value *argv, int argc, const Value *)
{
const QMetaObjectWrapper *This = static_cast<const QMetaObjectWrapper*>(f);
return This->constructInternal(argv, argc);
}
ReturnedValue QMetaObjectWrapper::constructInternal(const Value *argv, int argc) const
{
d()->ensureConstructorsCache();
ExecutionEngine *v4 = engine();
const QMetaObject* mo = d()->metaObject;
if (d()->constructorCount == 0) {
return v4->throwTypeError(QLatin1String(mo->className())
+ QLatin1String(" has no invokable constructor"));
}
Scope scope(v4);
Scoped<QObjectWrapper> object(scope);
JSCallData cData(scope, argc, argv);
CallData *callData = cData.callData();
if (d()->constructorCount == 1) {
object = callConstructor(d()->constructors[0], v4, callData);
}
else {
object = callOverloadedConstructor(v4, callData);
}
Scoped<QMetaObjectWrapper> metaObject(scope, this);
object->defineDefaultProperty(v4->id_constructor(), metaObject);
object->setPrototypeOf(const_cast<QMetaObjectWrapper*>(this));
return object.asReturnedValue();
}
ReturnedValue QMetaObjectWrapper::callConstructor(const QQmlPropertyData &data, QV4::ExecutionEngine *engine, QV4::CallData *callArgs) const {
const QMetaObject* mo = d()->metaObject;
const QQmlStaticMetaObject object(mo);
return CallPrecise(object, data, engine, callArgs, QMetaObject::CreateInstance);
}
ReturnedValue QMetaObjectWrapper::callOverloadedConstructor(QV4::ExecutionEngine *engine, QV4::CallData *callArgs) const {
const int numberOfConstructors = d()->constructorCount;
const int argumentCount = callArgs->argc();
const QQmlStaticMetaObject object(d()->metaObject);
QQmlPropertyData best;
int bestParameterScore = INT_MAX;
int bestMatchScore = INT_MAX;
QV4::Scope scope(engine);
QV4::ScopedValue v(scope);
for (int i = 0; i < numberOfConstructors; i++) {
const QQmlPropertyData & attempt = d()->constructors[i];
QQmlMetaObject::ArgTypeStorage storage;
int methodArgumentCount = 0;
int *methodArgTypes = nullptr;
if (attempt.hasArguments()) {
int *args = object.constructorParameterTypes(attempt.coreIndex(), &storage, nullptr);
if (!args) // Must be an unknown argument
continue;
methodArgumentCount = args[0];
methodArgTypes = args + 1;
}
if (methodArgumentCount > argumentCount)
continue; // We don't have sufficient arguments to call this method
int methodParameterScore = argumentCount - methodArgumentCount;
if (methodParameterScore > bestParameterScore)
continue; // We already have a better option
int methodMatchScore = 0;
for (int ii = 0; ii < methodArgumentCount; ++ii)
methodMatchScore += MatchScore((v = callArgs->args[ii]), methodArgTypes[ii]);
if (bestParameterScore > methodParameterScore || bestMatchScore > methodMatchScore) {
best = attempt;
bestParameterScore = methodParameterScore;
bestMatchScore = methodMatchScore;
}
if (bestParameterScore == 0 && bestMatchScore == 0)
break; // We can't get better than that
};
if (best.isValid()) {
return CallPrecise(object, best, engine, callArgs, QMetaObject::CreateInstance);
} else {
QString error = QLatin1String("Unable to determine callable overload. Candidates are:");
for (int i = 0; i < numberOfConstructors; i++) {
const QQmlPropertyData & candidate = d()->constructors[i];
error += QLatin1String("\n ") +
QString::fromUtf8(d()->metaObject->constructor(candidate.coreIndex())
.methodSignature());
}
return engine->throwError(error);
}
}
bool QMetaObjectWrapper::virtualIsEqualTo(Managed *a, Managed *b)
{
Q_ASSERT(a->as<QMetaObjectWrapper>());
QMetaObjectWrapper *aMetaObject = a->as<QMetaObjectWrapper>();
QMetaObjectWrapper *bMetaObject = b->as<QMetaObjectWrapper>();
if (!bMetaObject)
return true;
return aMetaObject->metaObject() == bMetaObject->metaObject();
}
DEFINE_OBJECT_VTABLE(QMetaObjectWrapper);
void Heap::QmlSignalHandler::init(QObject *object, int signalIndex)
{
Object::init();
this->signalIndex = signalIndex;
setObject(object);
}
DEFINE_OBJECT_VTABLE(QmlSignalHandler);
void QmlSignalHandler::initProto(ExecutionEngine *engine)
{
if (engine->signalHandlerPrototype()->d_unchecked())
return;
Scope scope(engine);
ScopedObject o(scope, engine->newObject());
QV4::ScopedString connect(scope, engine->newIdentifier(QStringLiteral("connect")));
QV4::ScopedString disconnect(scope, engine->newIdentifier(QStringLiteral("disconnect")));
o->put(connect, QV4::ScopedValue(scope, engine->functionPrototype()->get(connect)));
o->put(disconnect, QV4::ScopedValue(scope, engine->functionPrototype()->get(disconnect)));
engine->jsObjects[QV4::ExecutionEngine::SignalHandlerProto] = o->d();
}
void MultiplyWrappedQObjectMap::insert(QObject *key, Heap::Object *value)
{
QV4::WeakValue v;
v.set(value->internalClass->engine, value);
QHash<QObject*, QV4::WeakValue>::insert(key, v);
connect(key, SIGNAL(destroyed(QObject*)), this, SLOT(removeDestroyedObject(QObject*)));
}
MultiplyWrappedQObjectMap::Iterator MultiplyWrappedQObjectMap::erase(MultiplyWrappedQObjectMap::Iterator it)
{
disconnect(it.key(), SIGNAL(destroyed(QObject*)), this, SLOT(removeDestroyedObject(QObject*)));
return QHash<QObject*, QV4::WeakValue>::erase(it);
}
void MultiplyWrappedQObjectMap::remove(QObject *key)
{
Iterator it = find(key);
if (it == end())
return;
erase(it);
}
void MultiplyWrappedQObjectMap::mark(QObject *key, MarkStack *markStack)
{
Iterator it = find(key);
if (it == end())
return;
it->markOnce(markStack);
}
void MultiplyWrappedQObjectMap::removeDestroyedObject(QObject *object)
{
QHash<QObject*, QV4::WeakValue>::remove(object);
}
QT_END_NAMESPACE
#include "moc_qv4qobjectwrapper_p.cpp"