How to get data from webkit requests

[rbp]

From the QNetworkReply docs:

QNetworkReply is a sequential-access QIODevice, which means that once data is read from the object, it no longer kept by the device. It is therefore the application's responsibility to keep this data if it needs to.

How can I save this data before it is read and lost?

I tried catching the QNetworkAccessManager::finished() signal but usually (not always) the data was already lost.

[wysota]

What do you mean it was "already lost"? Did you read it in some other place?

[rbp]

webkit would have read that data to help render the webpage

[rbp]

Perhaps more explanation is needed. I am using QWebView to render webpages and have subclassed QNetworkAccessManager to try and save the data from certain requests, however the data is usually already read by the time of the QNetworkAccessManager::finished() signal.

Would the solution be to override QNetworkAccessManager::createRequest() and create my own custom QNetworkReply objects that maintain the returned data?
If so do you know of any examples?

[wysota]

If you even manage to intercept the data then WebKit won't be able to access it. Is that really what you want? Isn't it better to pick it up directly from WebKit's webpage object after it is downloaded and used?

[rbp]

If you even manage to intercept the data then WebKit won't be able to access it. Is that really what you want?

No. For example if I use peek() rather than read() then data is still there. Problem is peeking before webkit reads. Or overriding QNetworkReply so this is not a problem.

Isn't it better to pick it up directly from WebKit's webpage object after it is downloaded and used?

Unfortunately not possible in my use case because data is not retained.

[rbp]

I struggled subclassing QNetworkReply - I don't sufficiently understand the internals.

I am using the Python bindings so does that open up any additional possibilities?

Or is there some way to just tell QIODevice to maintain the buffer?

[wysota]

Or is there some way to just tell QIODevice to maintain the buffer?

The device is sequential. Once you read it, the data is gone.

The only thing that I know will work which comes to my mind right now is to reimplement QNetworkAccessManager::createRequest() to return a custom QNetworkReply that will act as a proxy between a real network reply (which you have to create too) and the stuff that uses it (WebKit and your code). Then all calls to read() will be going through your network reply class so you can do anything you want with the data (i.e. copy it and provide elsewhere). I don't know if that's the best or the simplest approach but I know it will work (it might be quite a lot of work though).

[rbp]

yeah I was trying to take that proxy approach, but feel I am working in the dark because am not familiar with QNetworkReply internals.

I find that the html at the URL is downloaded correctly and the finished() signal is emitted, but the content of QWebView remains empty.
How does QWebView get the content from QNetworkReply? Neither peek(), read(), readLine(), readData(), or readAll() is ever called in this class, but it will render properly when I use a standard QNetworkReply.

Here is my code (adapted from this example):

class NetworkReply(QNetworkReply):
    def __init__(self, reply):
        self.reply = reply
        QNetworkReply.__init__(self)
 
        # handle these to forward
        reply.metaDataChanged.connect(self.applyMetaData)
        reply.readyRead.connect(self.readInternal)
        reply.error.connect(self.errorInternal)
        # forward signals
        reply.finished.connect(self.finished)
        reply.uploadProgress.connect(self.uploadProgress)
        reply.downloadProgress.connect(self.downloadProgress)
 
        self.setOpenMode(QNetworkReply.ReadOnly)
        self.data = self.buffer = ''
 
 
    def operation(self):
        return self.reply.operation()
 
 
    def request(self):
        return self.reply.request()
 
 
    def url(self):
        return self.reply.url()
 
 
    def abort(self):
        self.reply.abort()
 
    def close(self):
        self.reply.close()
 
 
    def isSequential(self):
        return self.reply.isSequential()
 
 
    def setReadBufferSize(self, size):
        QNetworkReply.setReadBufferSize(size)
        self.reply.setReadBufferSize(size)
 
 
    def applyMetaData(self):
        for header in self.reply.rawHeaderList():
            self.setRawHeader(header, self.reply.rawHeader(header))
 
        self.setHeader(QNetworkRequest.ContentTypeHeader, self.reply.header(QNetworkRequest.ContentTypeHeader))
        self.setHeader(QNetworkRequest.ContentLengthHeader, self.reply.header(QNetworkRequest.ContentLengthHeader))
        self.setHeader(QNetworkRequest.LocationHeader, self.reply.header(QNetworkRequest.LocationHeader))
        self.setHeader(QNetworkRequest.LastModifiedHeader, self.reply.header(QNetworkRequest.LastModifiedHeader))
        self.setHeader(QNetworkRequest.SetCookieHeader, self.reply.header(QNetworkRequest.SetCookieHeader))
 
        self.setAttribute(QNetworkRequest.HttpStatusCodeAttribute, self.reply.attribute(QNetworkRequest.HttpStatusCodeAttribute))
        self.setAttribute(QNetworkRequest.HttpReasonPhraseAttribute, self.reply.attribute(QNetworkRequest.HttpReasonPhraseAttribute))
        self.setAttribute(QNetworkRequest.RedirectionTargetAttribute, self.reply.attribute(QNetworkRequest.RedirectionTargetAttribute))
        self.setAttribute(QNetworkRequest.ConnectionEncryptedAttribute, self.reply.attribute(QNetworkRequest.ConnectionEncryptedAttribute))
        self.setAttribute(QNetworkRequest.CacheLoadControlAttribute, self.reply.attribute(QNetworkRequest.CacheLoadControlAttribute))
        self.setAttribute(QNetworkRequest.CacheSaveControlAttribute, self.reply.attribute(QNetworkRequest.CacheSaveControlAttribute))
        self.setAttribute(QNetworkRequest.SourceIsFromCacheAttribute, self.reply.attribute(QNetworkRequest.SourceIsFromCacheAttribute))
        # attribute does not exist
        #self.setAttribute(QNetworkRequest.DoNotBufferUploadDataAttribute, self.reply.attribute(QNetworkRequest.DoNotBufferUploadDataAttribute))
        self.metaDataChanged.emit()
 
 
    def errorInternal(self, e):
        self.error.emit(e)
        self.setError(e, str(e))
 
 
    def readInternal(self):
        # this is called 
        s = self.reply.readAll()
        self.data += s
        self.buffer += s
        self.readyRead.emit()
 
 
    def bytesAvailable(self):
        return len(self.buffer) + self.reply.bytesAvailable()
 
 
    def readAll(self):
        # this is never called
        return self.data
 
 
    def readData(self, data, size):
        # this is never called
        size = min(size, len(self.buffer))
        data, self.buffer = self.buffer[:size], self.buffer[size:]
        return size

class NetworkReply(QNetworkReply):
    def __init__(self, reply):
        self.reply = reply
        QNetworkReply.__init__(self)

        # handle these to forward
        reply.metaDataChanged.connect(self.applyMetaData)
        reply.readyRead.connect(self.readInternal)
        reply.error.connect(self.errorInternal)
        # forward signals
        reply.finished.connect(self.finished)
        reply.uploadProgress.connect(self.uploadProgress)
        reply.downloadProgress.connect(self.downloadProgress)

        self.setOpenMode(QNetworkReply.ReadOnly)
        self.data = self.buffer = ''


    def operation(self):
        return self.reply.operation()


    def request(self):
        return self.reply.request()


    def url(self):
        return self.reply.url()


    def abort(self):
        self.reply.abort()

    def close(self):
        self.reply.close()


    def isSequential(self):
        return self.reply.isSequential()


    def setReadBufferSize(self, size):
        QNetworkReply.setReadBufferSize(size)
        self.reply.setReadBufferSize(size)


    def applyMetaData(self):
        for header in self.reply.rawHeaderList():
            self.setRawHeader(header, self.reply.rawHeader(header))

        self.setHeader(QNetworkRequest.ContentTypeHeader, self.reply.header(QNetworkRequest.ContentTypeHeader))
        self.setHeader(QNetworkRequest.ContentLengthHeader, self.reply.header(QNetworkRequest.ContentLengthHeader))
        self.setHeader(QNetworkRequest.LocationHeader, self.reply.header(QNetworkRequest.LocationHeader))
        self.setHeader(QNetworkRequest.LastModifiedHeader, self.reply.header(QNetworkRequest.LastModifiedHeader))
        self.setHeader(QNetworkRequest.SetCookieHeader, self.reply.header(QNetworkRequest.SetCookieHeader))

        self.setAttribute(QNetworkRequest.HttpStatusCodeAttribute, self.reply.attribute(QNetworkRequest.HttpStatusCodeAttribute))
        self.setAttribute(QNetworkRequest.HttpReasonPhraseAttribute, self.reply.attribute(QNetworkRequest.HttpReasonPhraseAttribute))
        self.setAttribute(QNetworkRequest.RedirectionTargetAttribute, self.reply.attribute(QNetworkRequest.RedirectionTargetAttribute))
        self.setAttribute(QNetworkRequest.ConnectionEncryptedAttribute, self.reply.attribute(QNetworkRequest.ConnectionEncryptedAttribute))
        self.setAttribute(QNetworkRequest.CacheLoadControlAttribute, self.reply.attribute(QNetworkRequest.CacheLoadControlAttribute))
        self.setAttribute(QNetworkRequest.CacheSaveControlAttribute, self.reply.attribute(QNetworkRequest.CacheSaveControlAttribute))
        self.setAttribute(QNetworkRequest.SourceIsFromCacheAttribute, self.reply.attribute(QNetworkRequest.SourceIsFromCacheAttribute))
        # attribute does not exist
        #self.setAttribute(QNetworkRequest.DoNotBufferUploadDataAttribute, self.reply.attribute(QNetworkRequest.DoNotBufferUploadDataAttribute))
        self.metaDataChanged.emit()


    def errorInternal(self, e):
        self.error.emit(e)
        self.setError(e, str(e))


    def readInternal(self):
        # this is called 
        s = self.reply.readAll()
        self.data += s
        self.buffer += s
        self.readyRead.emit()


    def bytesAvailable(self):
        return len(self.buffer) + self.reply.bytesAvailable()


    def readAll(self):
        # this is never called
        return self.data


    def readData(self, data, size):
        # this is never called
        size = min(size, len(self.buffer))
        data, self.buffer = self.buffer[:size], self.buffer[size:]
        return size

To copy to clipboard, switch view to plain text mode 

[rbp]

figured it out - the docs are wrong. readData() only takes an int and returns a string:

def readData(self, size):
    size = min(size, len(self.buffer))
    data, self.buffer = self.buffer[:size], self.buffer[size:]
    return str(data)

def readData(self, size):
    size = min(size, len(self.buffer))
    data, self.buffer = self.buffer[:size], self.buffer[size:]
    return str(data)

To copy to clipboard, switch view to plain text mode 

[wysota]

figured it out - the docs are wrong. readData() only takes an int and returns a string:

Hmm? Where the docs are wrong?

[rbp]

See Python QIODevice documentation

[wysota]

Didn't you mix readData() with read()? The latter does what you say.

[rbp]

Mix them how? My example wouldn't work until I defined readData() as above, which means that documentation is wrong.

[wysota]

No the docs are not wrong because Qt's QIODevice::readData() really takes two parameters. If something is wrong then it's the PyQt wrapping of that method.

[rbp]

Yes you are right that Qt's readData() takes two parameters - I am not disputing that.
However the Python bindings use a modified API to better suit the language, which is why I was able to guess the true syntax. The documentation is wrong.

[wysota]

In that case there should be no readData() method at all as it completely duplicates what read() does In C++ it makes sense to have two methods, in Python, it seems it doesn't.

[piotr.dobrogost]

Once I wanted to create a proxy for QNetworkReply too. That's when I started the thread Subclassing QNetworkReply.

I'm curious; how did you find this example?

[rbp]

someone passed me that link on the Qt webkit mailing list.

[KBHomes]

Wow, this is *exactly* what I needed! I was in the same position, couldn't figure out how to correctly subclass QNetworkReply, but this did just the trick!

Thanks a ton.