How to plot large time series with QCustomPlot efficiently?

Question

I am currently plotting a digital signal in Qt with QCustomPlot but it seems that when the number of samples is greater than 10000000 the operation becomes very slow. I have a time vector and a data vector and I'm setting the data like this:

QCustomPlot *plot;
QCPGraph *graph;
graph->setData(time, data); 

Any chance to make this more efficient?

Answer 1

Because QCustomPlot uses internally a QCPDataMap (which is a typedef of QMap<double, QCPData>) this means that it is using a map to store the actual data sorted by x coordinates (keys). Unfortunately the QCPGraph::setData(const QVector<double> &x, const QVector<double> &y) method doesn't take advantage of the fact that samples could be ordered and doesn't use the insertion hint, so this improved results significantly:

QCPDataMap *data = new QCPDataMap();
size_t len = x.size();
auto xp = std::begin(x);
auto yp = std::begin(y);
while (len--)
    data->insertMulti(data->constEnd(), *xp, QCPData(*xp++, *yp++)); 
graph->setData(data);

I don't think that std::maps or QMaps is the best structure to store samples on X,Y graphs because a new allocation and release is done for every entry in the map and we are talking about millions of them. QCustomPlot should implement a custom map class with a custom allocator to avoid these memory issues.