AuroraMiddleware/qt5base-lts
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

QtCore: fix the number precision in QJsonDocument.toJson() again

Browse Source 
Need to store 17 decimal digits for binary64, IEEE 754 double formats.

Autotest is included. Test cases from TC39 test suite for ECMAScript.

Task-number: QTBUG-31926
Change-Id: I546398f21ea7ff5e40e89fc9de8703f628f55df9
Reviewed-by: Thiago Macieira <thiago.macieira@intel.com>
Reviewed-by: Jędrzej Nowacki <jedrzej.nowacki@digia.com>
This commit is contained in:
Liang Qi 2013-07-02 13:16:48 +02:00 committed by The Qt Project
parent 6e8dddadd8
commit 94c8efff8a
2 changed files with 63 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/corelib/json/qjsonwriter.cpp
View File

@ -170,7 +170,7 @@ static void valueToJson(const QJsonPrivate::Base *b, const QJsonPrivate::Value &
json += v.toBoolean() ? "true" : "false";
break;
case QJsonValue::Double:
json += QByteArray::number(v.toDouble(b), 'g', 13);
json += QByteArray::number(v.toDouble(b), 'g', 17);
break;
case QJsonValue::String:
json += '"';

62
tests/auto/corelib/json/tst_qtjson.cpp
View File

@ -40,6 +40,7 @@
****************************************************************************/
#include <QtTest>
#include "qjsonarray.h"
#include "qjsonobject.h"
#include "qjsonvalue.h"
@ -62,6 +63,8 @@ private Q_SLOTS:
void testValueSimple();
void testNumbers();
void testNumbers_2();
void testNumbers_3();
void testObjectSimple();
void testObjectSmallKeys();
@ -329,6 +332,65 @@ void tst_QtJson::testNumbers()
}
void tst_QtJson::testNumbers_2()
{
// test cases from TC39 test suite for ECMAScript
// http://hg.ecmascript.org/tests/test262/file/d067d2f0ca30/test/suite/ch08/8.5/8.5.1.js
// Fill an array with 2 to the power of (0 ... -1075)
double value = 1;
double floatValues[1076], floatValues_1[1076];
QJsonObject jObject;
for (int power = 0; power <= 1075; power++) {
floatValues[power] = value;
jObject.insert(QString::number(power), QJsonValue(floatValues[power]));
// Use basic math operations for testing, which are required to support 'gradual underflow' rather
// than Math.pow etc..., which are defined as 'implementation dependent'.
value = value * 0.5;
}
QJsonDocument jDocument1(jObject);
QByteArray ba(jDocument1.toJson());
QJsonDocument jDocument2(QJsonDocument::fromJson(ba));
for (int power = 0; power <= 1075; power++) {
floatValues_1[power] = jDocument2.object().value(QString::number(power)).toDouble();
QVERIFY2(floatValues[power] == floatValues_1[power], QString("floatValues[%1] != floatValues_1[%1]").arg(power).toLatin1());
}
// The last value is below min denorm and should round to 0, everything else should contain a value
QVERIFY2(floatValues_1[1075] == 0, "Value after min denorm should round to 0");
// Validate the last actual value is min denorm
QVERIFY2(floatValues_1[1074] == 4.9406564584124654417656879286822e-324, QString("Min denorm value is incorrect: %1").arg(floatValues_1[1074]).toLatin1());
// Validate that every value is half the value before it up to 1
for (int index = 1074; index > 0; index--) {
QVERIFY2(floatValues_1[index] != 0, QString("2**- %1 should not be 0").arg(index).toLatin1());
QVERIFY2(floatValues_1[index - 1] == (floatValues_1[index] * 2), QString("Value should be double adjacent value at index %1").arg(index).toLatin1());
}
}
void tst_QtJson::testNumbers_3()
{
// test case from QTBUG-31926
double d1 = 1.123451234512345;
double d2 = 1.123451234512346;
QJsonObject jObject;
jObject.insert("d1", QJsonValue(d1));
jObject.insert("d2", QJsonValue(d2));
QJsonDocument jDocument1(jObject);
QByteArray ba(jDocument1.toJson());
QJsonDocument jDocument2(QJsonDocument::fromJson(ba));
double d1_1(jDocument2.object().value("d1").toDouble());
double d2_1(jDocument2.object().value("d2").toDouble());
QVERIFY(d1_1 != d2_1);
}
void tst_QtJson::testObjectSimple()
{
QJsonObject object;