distinguishable matplotlib colours / symbols / line styles

[duncan smith]

On 16/12/2019 21:08, DL Neil wrote:
> On 17/12/19 5:19 am, Chris Angelico wrote:
>> On Tue, Dec 17, 2019 at 3:16 AM duncan smith <duncan at invalid.invalid>
>> wrote:
>>>
>>> Hello,
>>>        Not really specific to Python or matplotlib (but that's what I'm
>>> using). I'm looking for a good combination of colours and symbols for
>>> scatter plots, and combination of colours and line styles for line
>>> plots. Too often I find myself producing these when I don't know whether
>>> they will end up being printed in colour or greyscale. It would be a lot
>>> easier if I could produce decent plots that would work well in either
>>> case. I can find various stuff on colour palettes, but pretty much
>>> nothing on symbols or line styles. Is anybody aware of an approach that
>>> works well? I'm aware of issues with colour blindness and RGB versus
>>> CMYK. Ideally I'd have something that I could just use that would deal
>>> with all these issues. TIA.
>>>
>>
>> I'd recommend looking at the Web Content Accessibility Guidelines
>> published by the W3C; there are a number of tools out there that are
>> designed to help you pick colours for a web page, and the same sorts
>> of rules will apply here, I think.
>>
>> Also, thank you for even *thinking* about this. A lot of people don't. :)
> 
> +1
> 
> We spend a lot of time teaching this topic (non-Python courses). It
> receives a lot of often highly-polarised comments/discussion. Many folk
> have their 'eyes opened' to an issue which has not affected them
> personally. Some even have to be informed that it is a legal obligation
> in their jurisdiction. However, it also receives the highest number of
> 'why do I have to learn this stuff' complaints...
> 
> I learned (way back) that the incidence of "color blindness" is far
> higher than I had imagined. Secondly, that it affects males more than
> females. Thirdly, that calling it "blindness" is a bit of a misnomer,
> because whilst people often can't see red 'correctly' (most common
> symptom), they do see something (it varies). Which is why they are
> permitted to drive vehicles (traffic lights: red, amber/yellow, green -
> and arrows; plus stop/brake lights), but why many smart-phone apps/web
> pages which encode information-relevance (red is 'wrong' and green/blue
> is acceptable) can become almost unusable (without other cues).
> 
> Those key-words: "accessibility guidelines" will yield a swag of useful
> tools - ignore the ones which are basically 'help choose the color of my
> web page/color palette, because they are often aiming (only) for
> 'pretty'. The best tools enumerate the efficacy of fg/bg
> color-combinations, allowing one to experiment; and will enumerate
> grey-scale variation/comparisons.
> 
> 
> Hmmm, note to self (you've inspired me to specifically review/critique
> the printing-from-screen action): what happens when we take a
> color-checked screen display and print same but end-up viewing it as
> monochrome/grey-scale output? Probably not a main-stream demand, but
> worth tossing at the WCAG experts...

A paper I recently published contained a number of colour images (graphs
- of the nodes and edges variety - and line plots). But it turned out
that the funding we used to have for colour printing charges no longer
exists. So the print version is greyscale with links to online colour
images. Anyone accessing the paper online will be able to download a pdf
with colour images. I don't yet know whether there will be funding for
colour printing charges for the paper I'm currently working on, yet I'm
generating plots. So something that would work for all these possible
end points would be ideal. That brings up the issue of distinguishable
symbols / markers (for scatter plots) and distinguishable line styles
(for line plots). i.e. They might help for greyscale without being too
distracting for colour images. So a single image would do for all the
possible end points. I'm not 100% happy about readers of the print
journal (probably) having to go online to make sense of my recent paper,
but that's done now.

BTW, I've gone with the Seaborn qualitative 'colorblind' palette for
now. Still tinkering with different markers / marker sizes / line
weights etc. Cheers.

Duncan