Page 1 of 1

TCharts (fastline) decimation process ...

Posted: Fri May 28, 2004 1:52 am
by 8123616
Hello,

I have a request to make for tChart's method of data decimation. It seems as if tChart's current process for decimating data (i.e. this.fastLine1.DrawAllPoints = false;) is to take the first-of-N, as opposed to taking the max-of-N (or peak).

I am currently working with VERY large datasets (500K+ points) and must implement decimation for charting. I would prefer to use tChart's inherent capabilities, but need the max-of-N decimation (peak). Would it be possible for Steema to implement this feature, and if so how soon before you could release it?

Thank you for your consideration of this feature.

Posted: Wed Jun 02, 2004 3:09 pm
by Chris
Hi John,
I am currently working with VERY large datasets (500K+ points) and must implement decimation for charting. I would prefer to use tChart's inherent capabilities, but need the max-of-N decimation (peak). Would it be possible for Steema to implement this feature, and if so how soon before you could release it?
Steema crew member Marjan tells me that this can be done, basically by grouping the points into groups with the same pixel coordinate and then replacing the points with the average, max, min, or two points (min and max).

I'll try to implement his fix into the .NET version this week, meaning that this feature may be implemented in the upcoming .NET maintenance release.

Posted: Wed Jun 09, 2004 2:13 pm
by 9333098
Could this also be included in the VCL version (FastLine and ColorGrid) ?

From my comments in the VCL forum April 29 "TCursorTool & Chart redraw" :
>
When all points are drawn, the last point plotted at a given X pixel is connected by a "sloped" line to the first point drawn at the next X pixel. So that should also be taken into account instead of simply drawing vertical lines at each X pixel. These last point and first points should be found and determined whether they lie within the vertical line segments at each X pixel.

This can be checked by drawing one series with all points and another on top of it with the one point reduction method and observing how different they appear. For thousands of points, there's often a large difference. I've used a candle series to draw vertica line results from the local min/max approach and it appears much closer to the all-points series. It does take more time to draw the chart this way.
>