Connecting QwtPlotCurve and QTableView (NOOB)

[yuvalal]

hello all,

What would be the best way to connect a QTableView and a QwtPlotCurve, in a way that I will need to keep the data only once?
I want to show the table and the curve, and connect them (data edited in the table will be reflected in the graph, selection in either will be reflected on the second).

I know it has something to do with subclassing QwtData, but as I'm new to all this, I didn't manage it on my own yet.

Any advice? (if anyone knows of an example it will be great!)

Thanks,
Yuvalal

[d_stranz]

I don't think you want to connect the QTableView to the plot; you want to connect the model to the plot. If you derive from QwtData and use that class to monitor your model's dataChanged() or modelReset() signals, then these will tell you when the plot needs to be updated. The data will still only live in one place (the model).

I don't have an example (yet), but I am planning to do exactly this at some point later when I revise my application to use Model/View instead of hard-coded data structures and table widgets. Once I do that, then QwtPlots will just become another view of the model. QwtData will be the "glue" that ties them together.

[Uwe]

I don't have an example (yet), but I am planning to do exactly this at some point later when I revise my application to use Model/View instead of hard-coded data structures and table widgets. Once I do that, then QwtPlots will just become another view of the model. QwtData will be the "glue" that ties them together.

Qt models might be a performance hog I wouldn't recommend to use for curves with many points.

QAbstractModel is made for almost any type of data structure - not only for simple arrays - with the price of several levels of indirections. Consider that these need to be resolved for every point ( in Qwt 5 for every coordinate ) for drawing a curve line + the same for the symbols !

That's the main reason why I don't use QAbstractModel in Qwt and why I believe, that other chart packages made some wrong decisions.

You can compare QAbstractModel with QwtData - both are bridges between the application and the real data structures. I recommend to implement both independently - not stacking them.

Uwe

[d_stranz]

Qt models might be a performance hog I wouldn't recommend to use for curves with many points.

Yes, I would agree that this could be the case. It might also happen that even with a proxy model, the proxy's filterAcceptsColumn() and filterAcceptsRow() methods get called for every row and column in the base model and this could cause some performance drain as well if the model has many rows (e.g. points).

My primary thoughts about implementing in this way was to give a user-friendly way to display the data in a table using QwtPlot. I have nearly finished porting a spreadsheet engine from C# into generic (non-Qt) C++ (google "CoreCalc"), and am now wrapping a QAbstractTableModel / QTableView around it so I can have an embeddable spreadsheet for my apps. Data reduction calculations done in the app will be filled into the spreadsheet as non-editable cells; users can then use spreadsheet functions to combine cells into new derived results. (For example, the user wants to scale a data reduction result column from linear to log).

Using a QSortFilterProxyModel between the base model and the plot also permits the user to sort results that may not be sorted in the base model to be able to view trends, etc.

My plots are not usually dynamic; once the spreadsheet is filled or recalculated, the plots will be automatically updated and then the user will play with the plots themselves. So in my case I don't think that the combination of QwtData and QAbstractTableModel will be too much of a performance drag.

I guess I will find out.

[yuvalal]

Thanks a lot for the replies and discussion.
I believe I'll try both options, and report back on what I find (or will ask more specific questions..)
Yuvalal