Hi Folks -- The Colored XY plot feature has the potential to be extremely useful. I can give you a long list of important use-cases if you want (see e.g. below).
Right now it is marginally useful. The main drawback is lack of control of the Z axis, i.e. the color. Different Z values result in different colors, but the mapping from Z value to color is undocumented and inscrutable. The documentation for Colored XY plots, in its entirety, says: > 10.3.4. Colored XY Plots > > Work in Progress Reference: https://help.gnome.org/users/gnumeric/stable/sect-graphs-overview-types.html.en#sect-graphs-overview-types-coloredxy Suggestion: It would be great if the Z axis would accept *strings* like this: "#00ff00" simplest case: solid green "#00ff0040" green at 25% opacity "#ff0000,#0000ff" red outline, blue fill Users who want to plot numerical data can construct their own colormap consisting of an array of strings, then apply the offset() function. Once the dust settles, it would be nice to document the Colored XY plot feature. ============================ Here is one use-case among many: Suppose we have a QAM64 modem. There are 64 possible codes that could be sent, each with an X,Y value. At the receiver we would like to plot the data. The X,Y position is given by the voltages as received (including noise), and the color is determined by what the symbol was /supposed/ to be. That way we can see where the category boundaries are. We can see to what extent a given symbol is encroaching on another category. It would be unsatisfactory to implement this as 64 separate XY plots for multiple reasons. First, that would be grossly laborious. What's worse, if the red plot is in front, that would put *all* the red points in front, which is not what we want. Instead, we want the Z-order to be determined by the order in which the symbols were received. =============================== Additional remarks: *) Do not let the perfect be the enemy of the good. I can happily live without the following creeping features. *) Creeping feature: "#ff0000,#0000ff,3,5.5" red outline, blue fill, shape 3, size 5.5 This would allow the marker shape (circle, square, diamond, ...) to be determined on the fly, and also the marker size. This would allow a Colored XY plot to do everything a Bubble plot does, and more. *) I'm not sure, but I wonder whether it might be better to put the information in 4 columns rather than in a single Z column. That is, separate columns for outline color, fill color, marker shape, and marker size. Hmmmm. *) Creeping feature: As for the lines, as opposed to the symbols: Let the line segment from point 1 to point 2 be a linear gradient, sweeping from the point-1 color to the point-2 color. Anybody who wants uniform-color segments can plot two points at each X,Y location (ending-point and re-starting-point). *) Creeping feature: One could imagine other strings, perhaps X11 color names such as DarkGoldenrod4, or perhaps SVG color names (which are inconsistent with the X11 names) ... but for my purposes names would be less useful than hex strings. I can compute the hex strings as a function of the data using dec2hex(), but I can't compute the names. I suppose the names are harmless, and they will appeal to casual users (as opposed to power users). However, they are neither necessary nor sufficient. They are not a viable substitute for the hex strings. ========================================= So, how hard would it be to implement color-strings? Does anybody have any better ideas along these lines? _______________________________________________ gnumeric-list mailing list gnumeric-list@gnome.org https://mail.gnome.org/mailman/listinfo/gnumeric-list