QJsonDocument::fromJson fails on "foreign" characters

[Per Gunnar Holm]

@mrjj said:

msg.toUTF8().toStdString().c_str()

Thanks again.
I tested the last suggestion too, but the contents of 'ba' are always the same.
So, this (ba) is the content that is being sent to QJsonDocument::fromJson.
Also, you can note that msg (QString) is not in utf8.

		ba	"{"string": "æøå"}"	QByteArray
				'{' 	123    	0x7b	char
				'"' 	34    	0x22	char
				's' 	115    	0x73	char
				't' 	116    	0x74	char
				'r' 	114    	0x72	char
				'i' 	105    	0x69	char
				'n' 	110    	0x6e	char
				'g' 	103    	0x67	char
				'"' 	34    	0x22	char
				':' 	58    	0x3a	char
				' ' 	32    	0x20	char
				'"' 	34    	0x22	char
				'ￃ' 	-61/195	0xc3	char
				'ﾦ' 	-90/166	0xa6	char
				'ￃ' 	-61/195	0xc3	char
				'ﾸ' 	-72/184	0xb8	char
				'ￃ' 	-61/195	0xc3	char
				'ﾥ' 	-91/165	0xa5	char
				'"' 	34    	0x22	char
				'}' 	125    	0x7d	char
		msg	"{"string": "æøå"}"	QString &
			[0]	'{' 	123	0x007b	QChar
			[1]	'"' 	34	0x0022	QChar
			[2]	's' 	115	0x0073	QChar
			[3]	't' 	116	0x0074	QChar
			[4]	'r' 	114	0x0072	QChar
			[5]	'i' 	105	0x0069	QChar
			[6]	'n' 	110	0x006e	QChar
			[7]	'g' 	103	0x0067	QChar
			[8]	'"' 	34	0x0022	QChar
			[9]	':' 	58	0x003a	QChar
			[10]	' ' 	32	0x0020	QChar
			[11]	'"' 	34	0x0022	QChar
			[12]	'æ' 	230	0x00e6	QChar
			[13]	'ø' 	248	0x00f8	QChar
			[14]	'å' 	229	0x00e5	QChar
			[15]	'"' 	34	0x0022	QChar
			[16]	'}' 	125	0x007d	QChar

[VRonin]

can you replace jsontext = "{\"string\": \"æøå\"}"; with jsontext =QString::fromWCharArray(L"{\"string\": \"æøå\"}"); or try

jsontext =QString::fromWCharArray(
                L"{\"string\": \""
             L"\u00E6"
             L"\u00F8"
                L"\u00E5"
                L"\"}"
                );

What I mean is that the problem is not how you read the string but how you create it

[Per Gunnar Holm]

@VRonin
Thank you for looking into it!
Unfortunately the result is the same in both cases:

jsontext = "{\"string\": \"æøå\"}"
msg = {"string": "æøå"}
(Broken json), error code = invalid UTF8 string

This is a bit of a surprise to me, I thought the second option would force the characters down to single byte representation?

Here's the actual code I ran (first suggestion commented out):

int main()
{
    QString jsontext = "{\"string\": \"abc\"}";
    textMessageReceived(jsontext);
//    jsontext = QString::fromWCharArray(L"{\"string\": \"æøå\"}");
    jsontext = QString::fromWCharArray(L"{\"string\": \"" L"\u00E6" L"\u00F8" L"\u00E5" L"\"}");
    qDebug() << "jsontext =" << jsontext;
    textMessageReceived(jsontext);

    return 0;
}

[VRonin]

This works for me. Qt 5.5.1 on MSVC2013

#include <QtCore/QCoreApplication>
#include <QtCore/QJsonDocument>

#include <QDebug>

void textMessageReceived(const QString &msg)
{
    qDebug() << "msg = " << msg << '\n';
    QJsonParseError error;

    // Create a Json document from text. Fails for foreign characters!
    QJsonDocument doc = QJsonDocument::fromJson(msg.toUtf8(), &error);

    if(doc.isNull())
        qDebug() << "(Broken json), error code = "
                  << error.errorString() << '\n';
    else
        qDebug() << "(Valid json) : " << QString(doc.toJson()) << '\n';
}

int main()
{

    QString jsontext = "{\"string\": \"abc\"}";
    textMessageReceived(jsontext);

    jsontext = QString::fromWCharArray(
                L"{\"string\": \""
             L"\u00E6"
             L"\u00F8"
                L"\u00E5"
                L"\"}"
                );
    textMessageReceived(jsontext);

    return 0;

}

Output:

msg =  "{\"string\": \"abc\"}" 

(Valid json) :  "{\n    \"string\": \"abc\"\n}\n" 

msg =  "{\"string\": \"æøå\"}" 

(Valid json) :  "{\n    \"string\": \"æøå\"\n}\n" 

[Per Gunnar Holm]

Thanks again @VRonin !

That means this is platform dependent I suppose. As far as I can see there are no differences between your code and mine (except you use qDebug).
It would also explain why there hasn't been a torrent of complaints to Qt if this is a bug and not me making a mistake; probably not too many on my platform.
As stated initially I have Qt 5.6.0 and I'm on Linux (CentOS 6.8).

[VRonin]

Even on Qt5.7 works for me.
2 questions:

• did you try my code?
• What encoding do you use on your .cpp file? (in QtCreator you can see it in edit->select encoding

[Per Gunnar Holm]

@VRonin

Now I did :-)
Same result (unfortunately):

msg =  "{\"string\": \"abc\"}" 
(Valid json) :  "{\n    \"string\": \"abc\"\n}\n" 
msg =  "{\"string\": \"æøå\"}" 
(Broken json), error code =  "invalid UTF8 string" 

I checked the encoding as instructed, and the "Text Encoding" window comes up with "UTF-8" high-lighted.
I am assuming this is OK?
-- Gunnar

[mrjj]

hi
window 7, Qt 5.7 also shows it ok.

Linux Mint, Qt 5.5

Notice it dont render as æøå but still consider it valid.

[Per Gunnar Holm]

Thanks @mrjj
It looks increasingly as if this is a problem specific to my platform.
That's good news for the world, but bad news for me, of course, since it decreases the likelihood of getting it fixed :-(

[Per Gunnar Holm]

Today I have tested on virtual installations of Ubuntu 16.04 and CentOS-7, both using QT 5.6.1.
The test program passes without problems there, so this is definitely a CentOS-6 problem.

[mrjj]

@Per-Gunnar-Holm
well maybe Qt 5.6.1 has a bug on that distro.

I hope Is it an option to use CentOs-7 instead :)

[kshegunov]

@Per-Gunnar-Holm

QByteArray ba (msg.toStdString().c_str());

Bug or no, the above line doesn't seem correct. You should enforce the required encoding (as @VRonin has done) instead of relying on the internal representation of std::string and/or QString.
The above should be:

QByteArray ba = msg.toUtf8();

If you need to output QStrings to the standard streams, attach a QTextStream to them instead of converting the objects to std::string:

QTextStream cout(stdout);
QTextStream cerr(stderr);
QTextStream cin(stdin);

Kind regards.

[Per Gunnar Holm]

Thanks @kshegunov !

The c_str() was just something we tested along the way!
The original code was

    // Create a Json document from text. Fails for foreign characters!
    QJsonDocument doc = QJsonDocument::fromJson(msg.toUtf8(), &error);

However, all the variations we/I have tried display the same problem (on CentOS 6.5, as we have discovered).