[Qt-interest] QAbstractListModel Scrolling Performance Issue

[Scott Aron Bloom]

Back to the list :)

Have you tried the setUniformRowHeights trick?

Again, just trying to focus in on the problem, have you tried using a QTableView ?

Scott



> -----Original Message-----
> From: Stephan Rose [mailto:kermos at somrek.net]
> Sent: Wednesday, December 03, 2008 9:52 AM
> To: Scott Aron Bloom
> Subject: Re: [Qt-interest] QAbstractListModel Scrolling Performance Issue
> 
> >Try to simplfy the problem a bit..
> >
> >First... I would simply create a dummy model that had 40k rows, no SQL.
> >See if the problem still exists..
> 
> Well a dummy model essentially is all this model is. While the model is fed
> from a SQL query using a slot, the two are completely separate from each
> other. So when ::data is called, the data is retrieved from the model's
> internal record storage not from the SQL Query.
> 
> >
> >The main trick is going to be to represent the system your ::data() and
> >model queries in the same way.
> >
> >If that doesn't show the problem you can look into the SQL side of
> >things... Maybe there is a delay your not thinking of...
> 
> Can't be as I've made sure that the SQL query was fully finished and every
> single record submitted to the model's internal storage before attempting to
> scroll.
> 
> The whole point of my new implementation is to ensure that the model never has
> to wait on the SQL Server during a ::data() call. Even if the SQL Query was
> still busy retrieving additional data, the model woudln't even know about
> this. As far as the model is concerned, it always has all the data it's ever
> going to have until the addRecords slot is called and data is appended.
> 
> >
> >If the delay is still there, then focus on the model and the view side
> >of things.
> >
> >BTW What type of view are you using?
> 
> It's a QTreeView as I need a multi-column view and QListView doesn't seem to
> have that functionality as far as I can tell.
> 
> >
> >Have your tried using QSqlQueryModel (or one of its derivatives) as the
> >model?
> 
> Yep I have and I found it no better or worse than my original own
> implementation. Exact same results.
> 
> Thanks,
> 
> Stephan