Infinite loop and signal deluge
-
wrote on 2 May 2011, 13:41 last edited by
I have a problem with a loop in my code.
Basically I do this in a thread:
@while( !stop )
{
textResult = someLongProcessing(); // this takes ~50ms
emit displayText(textResult);
}@ The signal is connected to the setValue() slot of a QLabel, so nothing really time consuming.The problem is that sometimes someLongProcessing() returns immediately, causing the loop to emit a lot of signals.
My ram usage then increases a lot and the software crashes.I can't call processEvents() because I'm not in the main thread. Then what should I do?
I can not handle the issue slot-side, since it's a qt slot.
If I add a sleep(5) in the loop it's better, but how would I know that the computer it's running on will need 5 or 10ms or even more to handle the setValue()...?Thanks a lot for the help.
- EDIT: I just found about Qt::BlockingQueuedConnection, could this be a good idea?
-
wrote on 2 May 2011, 13:48 last edited by
I would simply make sure that you don't emit your signal if it should not emit. Best place to stop such an overflow it as the source...
Is there some way for your process to know if the result is new or updated or something like that? If not, you could use a technique like "this":http://developer.qt.nokia.com/wiki/Delay_action_to_wait_for_user_interaction to throttle your signal emissions, I guess.
-
wrote on 5 May 2011, 07:34 last edited by
[quote author="Andre" date="1304344127"]I would simply make sure that you don't emit your signal if it should not emit.[/quote]In fact I always want to emit this signal. I "rephrased" my question in the first post.
-
wrote on 5 May 2011, 07:56 last edited by
maybe this?
@while ( !stop ){
bool finished = false;
finished = someLongProcessing(*textResult); // if you finshed parsing your text return true
if(finished)
emit displayText(textResult);
}@ -
wrote on 5 May 2011, 09:12 last edited by
I don't get it. With your finished boolean you just turn someLongProcessing() into a blocking call. But it is already blocking.
-
wrote on 5 May 2011, 09:23 last edited by
Sorry forgot the stop = true; line.
@while ( !stop ){
bool finished = false;
finished = someLongProcessing(*textResult); // if you finshed parsing your text return true
if(finished){
emit displayText(textResult);
stop = true;
}
}@ -
wrote on 5 May 2011, 09:30 last edited by
This will break the loop. I don't want to break the loop when someLongProcessing() returns true. You obviously did not understood my problem.
-
wrote on 5 May 2011, 09:53 last edited by
It is generally only useful to update a text if it has actually changed, so assuming that some successive calls may produce the same text result:
@QString previousText;
while (true) {
QString text = someLongProcessing();
if (previousText != text) {
previousText = text;
emit displayText(text);
}
}@ -
wrote on 5 May 2011, 11:44 last edited by
I'm sorry but successive calls wont produce the same result.
There's nothing to do with the event queue? -
wrote on 5 May 2011, 11:52 last edited by
QApplication::processEvents() helps?
-
wrote on 5 May 2011, 11:59 last edited by
Hi neFast,
is this loop executed in a thread?
If yes, processEvents will never help. -
wrote on 5 May 2011, 12:04 last edited by
Yes it is in a thread. What do you suggest?
-
wrote on 5 May 2011, 12:16 last edited by
From that thread you have no influence on the main thread directly by calling processEvents etc. If you fill up the queue of one thread by another one, you have to think, whether it is the correct way for doing so.
Perhaps you should change the logic to something like this:
thread --> setText on intermediatObject --> emit signal
intermediatObject is a thread save object, which stores the last text and stores, whether the text was already read. If a text is set, it emits the signal, if after the last read there came no new text.
But that changes the logic completely...
-
wrote on 5 May 2011, 12:21 last edited by
Another consideration is that the text might change quickly enough for the user not even to notice it. It would probably make sense to not let the longProcessThingy() control the text the user sees unless there's an error of sorts. The current approach seems indeed too naive for successful operation.
-
wrote on 26 May 2011, 14:09 last edited by
Sorry to bump this thread, but are there any best practice that I could find to solve my problem?
- EDIT: I just found about Qt::BlockingQueuedConnection, could this be a good idea?
-
wrote on 26 May 2011, 14:35 last edited by
If your thread can wait, sure, try that.
-
wrote on 26 May 2011, 14:37 last edited by
I'm affraid it will affect performances. I'll try that and report here.
-
wrote on 26 May 2011, 14:40 last edited by
It will slow down your thread somewhat, yes. If the thread cannot wait, you should probably go for Gerolf's solution or something like it.
-
wrote on 26 May 2011, 14:44 last edited by
[quote author="Franzk" date="1306420801"]Gerolf's solution or something like it.[/quote] Any details on this?
Gerolf method could work with text (because it does not always change), but I also emit pointer to images. And after each loop the image changes, so I need to either:- wait for the event to be processed (in a performance friendly fashion)
- trash events if too many of them are fired
I cannot use this "emit only if required" idea, because I can not decide if it's required based on the content of the image.
-
wrote on 26 May 2011, 14:51 last edited by
Hmno, that was Andre's suggestion. Read "Gerolf's suggestion":https://developer.qt.nokia.com/forums/viewreply/33804/ again.