Re: plot data stream in realtime

[Matt Flax <flatmax flatmax org>]

OK . now I see this post... slipped by me at first.

I am open to the idea of removing the static type associated with the 
gtkdatabox vectors.

That is probably a good idea.

Currently we use the gtk_databox_values_to_pixels function to convert to 
pixel locations on the pixmaps ... I guess that would need addressing ?

Matt

On 11/12/2012 08:32 PM, Amish S. Dave wrote:
> Just some thoughts on this topic. Basically, it gets at the idea of 
> trying to handle different representations for the input data. This is 
> assumed to be representable by static arrays of gfloats. That's fine, 
> but it means that all applications that don't start with that 
> representation need to adapt their representation to what gtkdatabox 
> wants.
>
> It should be fairly straightforward to implement a slightly different 
> scheme that could allow the inputs arrays to be
>
> 1. the current linear arrays [x0 x1 x2 .. xn] [y0 y1 y2 .. yn]
> 2. represent a ring-buffer e.g. [xn-1 xn x0 x1 x2 ... xn-2]
> 3. dynamically growing arrays
> 4. interleaved X&Y values  [x0 y0 x1 y1 x2 y2 ... xn yn]
>
> I'm thinking of something like this:
>
> gtk_databox_graph_data_update(graph, int XYallocpts, void *X, void *Y, 
> int startpt, int numpts, int xstridebytes, int ystridebytes, GType 
> XYdatatype);
>
> If this function is never called, backwards compatibility is assured 
> by assuming that
> XYdatatype == G_TYPE_FLOAT;
> xstridebytes  = ystridebytes = sizeof(float);
> startpt = 0;
> numpts = XYallocpts
>
> But, if this function *is* called, the graph replaces the existing X 
> and Y with the new pointers, and it forces a redraw, and updates any 
> rulers as needed.
>
> For the example of a realtime plot of something where new data is 
> coming in and the graph is adding points to the rightward end, the 
> calling program could allocate an array at the beginning large enough 
> to hold the final size (XYallocpts), but by calling this function 
> periodically with numpts incremented to reflect number of points 
> currently available, it would have the effect of adding new data to 
> the right of the screen.   No new allocations, and no unnecessary 
> copying of arrays.
>
> If you wanted to scroll the data, you could keep numpts fixed once it 
> reached your number_of_points_to_graph, and then start incrementing 
> startpt such that the most recently acquired point was startpt + 
> numpts - 1...
>
> If you need to realloc() a larger array, then this function will note 
> the change in the XYallocpts value, or the X or Y pointers, and 
> realize that new arrays are being passed...
>
> If the arrays were actually ring buffers, then calling
> ...
> gtk_databox_graph_data_update(ptsgraph, 10, Xpts, Ypts, 0, 10, 
> sizeof(float), sizeof(float), G_TYPE_FLOAT);
> Ypts[9] = get_data_from_adc();
> gtk_databox_graph_data_update(ptsgraph, 10, Xpts, Ypts, 1, 10, 
> sizeof(float), sizeof(float), G_TYPE_FLOAT);
> Ypts[0] = get_data_from_adc();
> gtk_databox_graph_data_update(ptsgraph, 10, Xpts, Ypts, 2, 10, 
> sizeof(float), sizeof(float), G_TYPE_FLOAT);
> Ypts[1] = get_data_from_adc();
> gtk_databox_graph_data_update(ptsgraph, 10, Xpts, Ypts, 3, 10, 
> sizeof(float), sizeof(float), G_TYPE_FLOAT);
> Ypts[2] = get_data_from_adc();
> ...
> shows 10 points each time but the latest point is always on the right 
> of the graph.  The X Axis could be kept constant, or the application 
> could update it if so desired by modifying Xpts[x].
>
> This scheme is also an opportunity to add support for types other than 
> float...  I'd suspect most of the time, GtkDatabox is being grafted 
> onto another already developed application that has it's own data 
> representations.  It's annoying to have to convert 
> doubles/ints/whatever to floats for example.  Most of what needs to be 
> done to support non-float data is in one function: 
> gtk_databox_values_to_pixels... and the modification is not hard.
>
> What do people think?
>
> Sincerely,
> --
> Amish
>
> On 11/12/2012 02:30 PM, Johannes Deutsch wrote:
> > Hi all,
> >
> > I'm very thankful for all of your echos. I received interesting
> > suggestions and was particularly pleased that the mailing list is very
> > active!
> >
> > I already wrote a little demo that simulates plotting of data in
> > realtime. By now the code is messy but i will pretty it up
> > tomorrow and post i on this list.
> >
> > Maybe it is a valid example. If not i would be thankful for any
> > comments to improve it further...
> >
> > With best regards and thanks for this great project
> > _______________________________________________
> > gtkdatabox-list mailing list
> > gtkdatabox-list gnome org
> > https://mail.gnome.org/mailman/listinfo/gtkdatabox-list
>
> _______________________________________________
> gtkdatabox-list mailing list
> gtkdatabox-list gnome org
> https://mail.gnome.org/mailman/listinfo/gtkdatabox-list