Doing multiple QNetworkRequest on the same QNetworkAccessManager

[Dimbreath]

I have a QStringList with some urls to text files (Changelogs) on the internet. (Github RAWs from gist, if interesed). Each version has it's struct (Inside that struct i have QString to place the changelog).
I have no problem requesting the changelogs, but i don't know how to assign them to their specific Version struct.

@ struct VersionInfo {
QString versionID;
QString updateVersionTo;
QString updateDate;
QString operatingSystems;
QString changelog;
QString downloadLink;
};
@

That's the struct of each VersionInfo. And i use this code to request:

@ m_manager = new QNetworkAccessManager(this);

QStringList m_urllist;
m_urllist << "Link 1" << "Link 2";
for(int counter = 0; counter < m_urllist.length(); counter++) {
    QNetworkRequest m_request;
    m_request.setRawHeader("User-Agent", "Autoupdater");
    m_request.setUrl(QUrl(m_urllist[counter]));
    m_manager->get(m_request);
}@

[Max13]

Hello,
QNetworkAccessManager is an asynchronous API, which means if you want to do multiple calls, you'll have to connect them to multiple slots. Or calling 1 by 1 the requests, and connect them to the same slot where you would manage them.

But I don't know if calling 2 times the same slot will call 2 difference slot and not the same one.

Moreover, I noticed that you're using non-pointer QNetworkRequest, which means it will disappear and be removed from the known variables, so I think the SLOT wouldn't be called if maybe the SLOT corresponding to "counter==0" is called DURING "counter==1" for example...

Sorry if it's not clear, if you need precision, I will ;)

[Dimbreath]

[quote author="Max13" date="1356829581"]Hello,
QNetworkAccessManager is an asynchronous API, which means if you want to do multiple calls, you'll have to connect them to multiple slots. Or calling 1 by 1 the requests, and connect them to the same slot where you would manage them.

But I don't know if calling 2 times the same slot will call 2 difference slot and not the same one.

Moreover, I noticed that you're using non-pointer QNetworkRequest, which means it will disappear and be removed from the known variables, so I think the SLOT wouldn't be called if maybe the SLOT corresponding to "counter==0" is called DURING "counter==1" for example...

Sorry if it's not clear, if you need precision, I will ;)[/quote]

Now that i think about it, i have the same question. o.O

My problem is that i can't identify each QNetworkReply with each VersionInfo.

[JKSH]

One solution is to create one QObject to handle each VersionInfo + QNetworkReply pair, e.g.

@
for(int counter = 0; counter < m_urllist.length(); counter++) {
QNetworkRequest m_request;
m_request.setRawHeader("User-Agent", "Autoupdater");
m_request.setUrl(QUrl(m_urllist[counter]));
QNetworkReply *reply = m_manager->get(m_request);

    // Store the reply, and connect the QNetworkReply::finished() signal
    // an internal slot that creates a VersionInfo based on the URL and the reply
    ChangelogProcessor *processor = new ChangelogProcessor(m_urllist[counter], reply);
}

@

(Just be careful of memory leaks: You need to delete the ChangelogProcessor AND the QNetworkReply when you're done!)

[Dimbreath]

@ void Updater::requestChangelogDownload()
{
QNetworkRequest m_request;
m_request.setRawHeader("User-Agent", "Autoupdater");
for(int counter = 0; counter < m_versionList.size(); counter++) {
m_request.setUrl(QUrl(m_versionList[counter].changelog));
m_changelogRequest.insert(m_versionList[counter].versionID, m_manager->get(m_request));
qDebug() << "A";
connect(m_changelogRequest.value(m_versionList[counter].versionID), SIGNAL(finished()), this, SLOT(changelogDownloaded()));
}
}

void Updater::changelogDownloaded()
{
qDebug() << sender();
QNetworkReply a = (QNetworkReply) sender();
qDebug() << a->isFinished();
qDebug() << a->readAll();
}
@

I can say, not a pretty code. FOr some reason it keeps printing "A" even though m_versionList.size() is 2.
Still it doesn't work, a->readAll() returns nothing. But i think i'm doing it poorly.

[JKSH]

[quote author="Blastcore" date="1356846308"]
@
QNetworkReply a = (QNetworkReply) sender();
@
[/quote]

Try a qobject_cast<> instead of a C-style cast

[Dimbreath]

Nope, it didn't work. :/ I wonder what's going on o.O

sender() is a QNetworkReplyImpl. Both are different.

[Dimbreath]

Using readyRead() signal from QNetworkReply fixed the issue.

[lgeyer]

Be aware that readyRead() only indicates that data is available, not necessarily that the reply has finished.

If you want to pass information, for example a reference to the <code>VersionInfo</code> which was used to initiate the request you can use an "attribute":http://qt-project.org/doc/qt-5.0/qtnetwork/qnetworkrequest.html#setAttribute.

QNetworkAccessManager provides a "finished":http://qt-project.org/doc/qt-5.0/qtnetwork/qnetworkaccessmanager.html#finished signal which passes the QNetworkReply.
@
struct VersionInfo
{
...
};

Q_DECLARE_METATYPE(VersionInfo*)

static const QNetworkRequest::Attribute VersionInfoAttribute = QNetworkRequest::User;

...

connect(m_manager, &QNetworkAccessManager::finished, [](QNetworkReply reply)
{
VersionInfo versionInfo = reply->request().attribute(VersionInfoAttribute)
.value<VersionInfo*>();
if (versionInfo != nullptr)
{
versionInfo->...
}

reply->deleteLater();

});

...

for (int counter = 0; counter < m_versionList.size(); counter++)
{
QNetworkRequest request;

request.setAttribute(VersionInfoAttribute,
                     QVariant::fromValue<VersionInfo*>(&m_versionList[counter]));
request.setRawHeader("User-Agent", "Autoupdater");
request.setUrl(m_versionList[counter].changelog);

m_manager->get(request);

}
@
If you don't want to connect to the finished(QNetworkReply*) signal (for instance because you are reusing the QNetworkAccessManager object and don't want to receive finished signals from other requests) you can use sender() (QNetworkReplyImpl is-a QNetworkReply).