#1644 Wrong point types with binary matrix plots and the default colorsequence
Hi. I'm attaching two gnuplot scripts that are supposed to produce identical plots. One has inline ascii data, and the other has inline binary data. The resulting plots are also attached. The point positions are correct, but as you can see the binary plot is drawn with the wrong point types. For some reason these point types aren't even consistent within the same data set. Using the classic colorsequence makes the issue go away. I bisected the code, and this bug was there as soon as the colorsequence command was added.
There's another issue here. If I plot these with the x11 terminal, and interactively rotate the data then I see the binary matrix plot respond noticeably more slowly: something about that render path is much slower. Both the ascii plot and the binary-with-the-classic-colorsequence plots appear much quicker. I haven't made any measurements for this, so I can only talk about how these "feel". Hopefully addressing the first issue will take care of the second one too.
This bug turns out to have been introduce back in version 4.5 days. It has nothing to do with the color sequence per se. Rather the problem is that the 3D impulse and point code used an incorrect test for the presence of variable color in an extra column. This is triggered by some binary plots because the number of effective using specs is greater than the number of using specs that actually appear on the command line. The code for 3D lines doesn't suffer from this, so I have replaced the bad test with a shared copy of the good test.
Fixed in 5.0 and 5.1 cvs.
Last edit: Ethan Merritt 2015-07-03
Hi. I tested the patch, and it fixes both issues. Thank you very much for a lightning-fast fix.