QString::localeAwareCompare() incorrect for å, ä and ö in Swedish

AndyBrice

If I run this code:

#include <QCoreApplication>
#include <QLocale>

int main( int argc, char* argv[] )
{
    QCoreApplication app( argc, argv );

    QLocale::setDefault( QLocale( QLocale::Swedish ) );

    QString s1 = "a";
    QString s2 = "b";
    QString s3 = "o";
    QString s4 = "z";
    QString s5 = "å";
    QString s6 = "ä";
    QString s7 = "ö";

    qDebug() << "QString::localeAwareCompare( s1, s2 )=" << QString::localeAwareCompare( s1, s2 );
    qDebug() << "QString::localeAwareCompare( s2, s3 )=" << QString::localeAwareCompare( s2, s3 );
    qDebug() << "QString::localeAwareCompare( s3, s4 )=" << QString::localeAwareCompare( s3, s4 );
    qDebug() << "QString::localeAwareCompare( s4, s5 )=" << QString::localeAwareCompare( s4, s5 );
    qDebug() << "QString::localeAwareCompare( s5, s6 )=" << QString::localeAwareCompare( s5, s6 );
    qDebug() << "QString::localeAwareCompare( s6, s7 )=" << QString::localeAwareCompare( s6, s7 );

    return app.exec();
}

It returns:

QString::localeAwareCompare( s1, s2 )= -1
QString::localeAwareCompare( s2, s3 )= -1
QString::localeAwareCompare( s3, s4 )= -1
QString::localeAwareCompare( s4, s5 )= 1
QString::localeAwareCompare( s5, s6 )= 1
QString::localeAwareCompare( s6, s7 )= -1

The locale aware sort order in Swedish is: a-z, å, ä, ö. So the results should all be -1.

I am on Windows 11 and I get the same result to Qt 5.15.19 and Qt 6.7.3.

Is this a bug, or am I doing something wrong?

It seems the same issue was also in Qt 4, but was fixed: https://stackoverflow.com/questions/3773914/why-isnt-qstringlocaleawarecompare-working-correctly

SGaist

Hi,

Can you check with a more recent version of Qt ? I remember work being done on that field recently.

aha_1980

@AndyBrice I'd print the contents of s1 to s7 just to make sure they really contain what you expect them to contain...

AndyBrice

Did it again in Qt 6.10.0 on Windows 11:

#include <QCoreApplication>
#include <QLocale>

int main( int argc, char* argv[] )
{
    QCoreApplication app( argc, argv );

    QLocale::setDefault( QLocale( QLocale::Swedish ) );

    QString s1 = "a";
    QString s2 = "b";
    QString s3 = "o";
    QString s4 = "z";
    QString s5 = "å";
    QString s6 = "ä";
    QString s7 = "ö";

    qDebug() << "s1=" << s1;
    qDebug() << "s2=" << s3;
    qDebug() << "s3=" << s3;
    qDebug() << "s4=" << s4;
    qDebug() << "s5=" << s5;
    qDebug() << "s6=" << s6;
    qDebug() << "s7=" << s7;

    qDebug() << "QString::localeAwareCompare( s1, s2 )=" << QString::localeAwareCompare( s1, s2 );
    qDebug() << "QString::localeAwareCompare( s2, s3 )=" << QString::localeAwareCompare( s2, s3 );
    qDebug() << "QString::localeAwareCompare( s3, s4 )=" << QString::localeAwareCompare( s3, s4 );
    qDebug() << "QString::localeAwareCompare( s4, s5 )=" << QString::localeAwareCompare( s4, s5 );
    qDebug() << "QString::localeAwareCompare( s5, s6 )=" << QString::localeAwareCompare( s5, s6 );
    qDebug() << "QString::localeAwareCompare( s6, s7 )=" << QString::localeAwareCompare( s6, s7 );

    return app.exec();
}

Still wrong:

s1= "a"
s2= "o"
s3= "o"
s4= "z"
s5= "å"
s6= "ä"
s7= "ö"
QString::localeAwareCompare( s1, s2 )= -1
QString::localeAwareCompare( s2, s3 )= -1
QString::localeAwareCompare( s3, s4 )= -1
QString::localeAwareCompare( s4, s5 )= 1
QString::localeAwareCompare( s5, s6 )= 1
QString::localeAwareCompare( s6, s7 )= -1

AndyBrice

Note that QCollator returns the correct result. This code:

    QCollator collator(QLocale::Swedish);
    QStringList list = {s7, s1, s2, s3, s4, s5, s6 };
    std::sort(list.begin(), list.end(), collator);
    qDebug() << "list=" << list;

Returns:

list= QList("a", "b", "o", "z", "å", "ä", "ö")

In Qt 5.15.19, Qt 6.7.3 and Qt 6.10.0.

AndyBrice

I have added a bug report: https://bugreports.qt.io/browse/QTBUG-141463

ChrisW67

On Linux, online installer Qt 6.10.0 or 6.8.1, with bundled libicui18n.so.73 the result is all -1 as expected.
With Qt 5.15.2 and libicui18n.so.56 on the same system I have this alternate version:

s1= "a"
s2= "o"
s3= "o"
s4= "z"
s5= "å"
s6= "ä"
s7= "ö"
QString::localeAwareCompare( s1, s2 )= -1
QString::localeAwareCompare( s2, s3 )= -1
QString::localeAwareCompare( s3, s4 )= -1
QString::localeAwareCompare( s4, s5 )= 1
QString::localeAwareCompare( s5, s6 )= -1
QString::localeAwareCompare( s6, s7 )= -1

So even older ICU versions display variation.

Edit: Given the comments on the bug, I don't think this will actually help unless you use an ICU-enabled build of Qt.