[Qt-interest] Subclass QNetworkRequest

[Markus Goetz]

I don't have the time right now to look at all of it, but emitting 
readyRead() from readData() sounds wrong. The QNAM connects to 
readyRead() and then would call read() and this calls the readData(). It 
would recurse :)

What I would do is to have a QTimer that sets the "quota" for the next 
time internal back to some value and emits the readyRead(). The 
readData() should then return only as many bytes as the quota allows.

Markus

ext Jason Wood wrote:
> Thanks for the help, I've pasted my QIODevice below, hopefully not too 
> large!
>
> m_Bps - This is the throttle rate which can be changed at any time
> m_Stream - This is the underlying data which is a QBuffer
>
> Here's is my QIODevice:
>
> ThrottledDevice::ThrottledDevice(QByteArray *data)
>
> : m_Stream(data),
>
> m_Bps(0)
>
> {
>
> m_ThrottleTimer.start();
>
> m_iQuota = 1024;
>
> emit readyRead();
>
> }
>
>
> qint64 ThrottledDevice::readData(char * data, qint64 maxSize)
>
> {
>
> if( !m_Bps )
>
> {
>
> emit readyRead();
>
> return m_Stream.read(data, maxSize);
>
> }
>
> QTimer::singleShot(50, this, SLOT("throttleData()"));
>
> qint64 bytesToRead = m_iQuota;
>
> m_iQuota = 0;
>
> return m_Stream.read(data, bytesToRead);
>
> }
>
>
> void ThrottledDevice::throttleData()
>
> {
>
> int msecs = 1000;
>
>         if( !m_ThrottleTimer.isNull() )
>
>             msecs = qMin<int>(1000, qMax<int>(1, 
> m_ThrottleTimer.elapsed()));
>
> m_ThrottleTimer.restart();
>
> m_iQuota = qMax<int>(1, m_Bps * msecs / 1000);
>
> emit readyRead();
>
> }
>
>
> qint64 ThrottledBuffer::writeData(const char * data, qint64 maxSize)
>
> {
>
> return m_Stream.write(data, maxSize);
>
> }
>
>
> bool ThrottledDevice::atEnd() const
>
> {
>
> return m_Stream.atEnd();
>
> }
>
>
> qint64 ThrottledDevice::bytesAvailable() const
>
> {
>
> return m_Stream.bytesAvailable();
>
> }
>
>
> qint64 ThrottledDevice::bytesToWrite() const
>
> {
>
> return m_Stream.bytesToWrite();
>
> }
>
>
> bool ThrottledDevice::canReadLine() const
>
> {
>
> return m_Stream.canReadLine();
>
> }
>
>
> void ThrottledDevice::close()
>
> {
>
> m_Stream.close();
>
> }
>
>
> bool ThrottledDevice::isSequential() const
>
> {
>
> return false;
>
> }
>
>
> bool ThrottledDevice::open(OpenMode mode)
>
> {
>
> if( !m_Stream.open(mode) )
>
> return false;
>
>
> QIODevice::open(mode);
>
> return true;
>
> }
>
>
> qint64 ThrottledDevice::pos() const
>
> {
>
> return m_Stream.pos();
>
> }
>
>
> bool ThrottledDevice::reset()
>
> {
>
> return m_Stream.reset();
>
> }
>
>
> bool ThrottledDevice::seek(qint64 pos)
>
> {
>
> return m_Stream.seek(pos);
>
> }
>
>
> qint64 ThrottledDevice::size() const
>
> {
>
> return m_Stream.size();
>
> }
>
>
> bool ThrottledDevice::waitForBytesWritten(int msecs)
>
> {
>
> return m_Stream.waitForBytesWritten(msecs);
>
> }
>
>
> bool ThrottledDevice::waitForReadyRead(int msecs)
>
> {
>
> return m_Stream.waitForReadyRead(msecs);
>
> }
>
>
> void ThrottledDevice::setMaxSpeed(int bps)
>
> {
>
> m_Bps = bps;
>
> m_ThrottleTimer.restart();
>
> QTimer::singleShot(50, this, SLOT("throttleData()"));
>
> }
>
>
>
> On Tue, Feb 23, 2010 at 7:57 AM, Markus Goetz <Markus.Goetz at nokia.com 
> <mailto:Markus.Goetz at nokia.com>> wrote:
>
>     ext Thiago Macieira wrote:
>     > Em Terça-feira 23. Fevereiro 2010, às 08.39.38, Markus Goetz
>     escreveu:
>     >
>     >> ext Thiago Macieira wrote:
>     >>
>     >>>> It
>     >>>> doesn't seem to implement QIODevice so I'm not exactly sure
>     how I could
>     >>>> do it, does anyone have any ideas? I was originally using a
>     throttled
>     >>>> QIODevice as the data in the QNetworkAccessManager::post()
>     method which
>     >>>> is OK in Windows but OSX seems to have a much larger buffer
>     size and so
>     >>>> buffering to low speed (~5kB/s) causes the connection to timeout
>     >>>> because the data is not sent until the internal buffer is
>     full. Please
>     >>>> correct me if I'm wrong but that does seem to be the case from my
>     >>>> testing.
>     >>>>
>     >> Should not happen. The OS should send as soon as it has ~MTU bytes.
>     >>
>     >
>     > Didn't we turn off the Nagle's algorithm in 4.6?
>     >
>     >
>     True, we set TCP_NODELAY for the QNAM upload.
>     Especially then a upload should not timeout. Maybe there is a flaw in
>     his QIODevice.
>
>     Jason, we can try to help if you show us the QIODevice.
>
>     Markus
>     _______________________________________________
>     Qt-interest mailing list
>     Qt-interest at trolltech.com <mailto:Qt-interest at trolltech.com>
>     http://lists.trolltech.com/mailman/listinfo/qt-interest
>
>