Custom colors for graphs?

I know that the version of relays graph https://metrics.torproject.org/network.html#versions has been modified to work for color blinds (red/green...).

Is it possible to use custom colors on this graph, at least?

On bad environmental conditions I can't tell them apart properly.

Considering all points (scalability, maintainability, code, other points I overlook) would it be OK to have a) a different color set (pre-defined); or b) custom colors

For b) I mean there's a legend. Fields for RGB values behind them? Sliders? Colormap? (something that doesn't get in my mind?)

(Hopefully no broken sentences due to editing, like in my last reply)

added component::metrics/website priority::medium resolution::implemented status::closed type::enhancement labels

Custom colors would take a lot of coding effort, so I'd prefer a different color set. Can you suggest a better palette for the versions graph that works for color blind people and that scales to new added versions? We're currently using YlOrBr as seen on http://www.datavis.ca/sasmac/brewerpal.html with 8 colors and 0.2.5 being the 9th. Or, as a compromise, we could use a color palette with only 5 different colors and the 6th being very similar to the 1st, the 7th being similar to the 2nd, and so on. The risk of confusing 0.1.2 relays with 0.2.4 relays is rather low.

Being colorblind appears to be more troublesome than ever expected. (Everything looks alike)

Would Accent work?

Deuteranope: http://vischeck.homeip.net/uploads/135677501723018/ Protanope: http://vischeck.homeip.net/uploads/135677506723155/ Tritanope: http://vischeck.homeip.net/uploads/135677512123292/

It has 8 colors that are totally different to non-colorblind. From the comparison above I feel that Accent "produces" the most distinctive colors. Every color is different from one to another and probably different enough from the whole set in the eyes of colorblind.

I'm open for other sets of course. Input of colorblind would be nice.

Want to try it out and maybe tweak colors until they're distinguishable for everyone? I just created a little tarball with the necessary R code to make a Tor versions graph: https://people.torproject.org/~karsten/volatile/task-7812-version-colors.tar.bz2. Thanks!

Alright, I plotted the graphs (R did), which worked like a charm, because it was semi-automated. Then I used ImageJ with the Vischeck plugin to create simulations for Deuteranope, Protanope, Tritanope of versions-all and versions-2012 for all color sets. This took longer because I did this manually as I found no way to automate this.

Now I got 12 graphs for each set, with the exception of "Greys" which are just 9. Well grey is good for anyone. I guess I can sort this one out. I tarballed the graphs (11 MB). Anyone interested in looking at them? (I'd like to not waste space here)

My plan was and still is to rate versions-all, versions-2012 and the simulations of them with 5 (best) to 1 (worst) in terms of how distinguishable they are. I have to see if I'm any good at this.

Pastel1 might not be appealing because of the colors them self, but in terms of telling them apart it is quite good.

Don't expect me to get wise soon. No time critical task and what's used is good enough to satisfy most people, I guess.

Trac:
comparison.ods

Please note that "Greys" contains ratings for graphs I did not created.

Please note that I rated the current color first (weeks ago) and rated the rest (bottom-up) today. My personal vision influenced the results to a high degree. It also appears that my rating could be skewed. Other people will have different opinions.

The current color is really good.

I can't upload the tarballed graphs here because around 11MB is too much. If someone is interested in them, please tell me where you would like to have them uploaded or send to.

Although "Accent" is good as well, there might be no need to change something; especially if it is not broken.

Even if the two processes (comparing four different colored images vs. creating graphs, modifying their colors, then comparing) have the same outcome I still learned something.

I'm fine with the status-quo, because a) The current color set is a really good choice, nobody complained. b) I can use any color I want if I so desire, it might be me anyway.

Trac:
YlOrBr_and_Accent.tar

I just uploaded the graphs with the current color (YlOrBr) and "Accent" for comparison.

I'm still storing all of them to show them on request.

I suggest to pick "WORKSFORME" in x weeks if nobody objects (leaving everything as it is).

"Accent" looks good and, IMHO, better than the current "YlOrBr". Switching over to "Accent". Thanks for putting all the work into this comparison! Closing.

Trac:
Resolution: N/A to implemented
Status: new to closed

closed