Repository navigation
Future style changes #14331
Description
This is a collection to keep track of possibly relevant ideas, which we should consider if we ever do a new style (likely not before 4.0).
axhline,axvlineshould use the grid color by default axhline/axvline don't use prop_cycle #14222 (comment)- Color blind-friendly default color cycle Color blind-friendly default color cycle #9460
- How to handle changes in styles? Update seaborn style #13680 (Update seaborn style)
- Marker sizes could be perceptually uniform Sizes of different markers are not perceptually uniform #15703
- Make 2D and 3D plots look more similar Inconsistent background & mesh for 2D and 3D graphs #20352
- Consider deactivating using offsets for tick labels by default. This is regularly confusing users, e.g. [Bug]: The plot is incorrect for certain data #23053.
- Let
stem()take part in the colorcycle [ENH]: consistent behaviour for stem #23127.
Activity
WeatherGod commentedon May 26, 2019
aoz commentedon Aug 17, 2022
Not sure if this is the right place for this but had a comment about
vlines:I think
vlinesshould use the cycle for colors/linestyles, but it doesn't.vlinesplots a set of lines. If I'm doing multiple calls to it, I'm most likely plotting multiple sets of lines each of which represents a different "property", and thus each set should get its own visual representation. If I want a bunch of lines with the same visual representation, I can do a single call tovlines.I ended up here after spending a bunch of time trying to debug my code and eventually learning about a somewhat convoluted solution to my problem (which I'm still struggling to make work without diving too deeply into the internals of matplotlib).
jklymak commentedon Aug 17, 2022
Axvline discussion is better had at #14222. Thanks.
aoz commentedon Aug 18, 2022
If I understand correctly,
axvlineplots a single vertical line. I don't have a strong argument (or need) for cycling in that case.One closed issue discussing a similar topic was pointing to this issue as a place to collect feature requests on styles, but if there is a more appropriate issue, I'd be happy to copy the comment there.
timhoffm commentedon Aug 18, 2022
I agree,
vlinesshould take part in the color cycle. - It's currently fixed to "C0", so no change for the first call. It's also simple to fix the color to "C0" if needed on subsequent calls.babryce commentedon May 29, 2023
With 1 in 12 males colorblind, or about 4% of the user base impacted it is very important to do whatever is needed to move matplotlib to an accessible set of color cycles, styles, etc.
https://matplotlib.org/stable/users/prev_whats_new/whats_new_2.2.html#new-style-colorblind-friendly-color-cycle
For me the default color cycle for line plots is unusable.
It isn't really enough to have them available because a large fraction of the population just is unaware of the amount of content that is impossible to read/use. Obviously I switch to other styles always. Marginal aesthetics cannot justify excluding 4+% of users, so I think this should be an essential change even if it has to wait to the 4.0 release as noted.
The best research available should be used in picking the new defaults. https://colorbrewer2.org is a starting point. Of particular difficulty are color based contour/heat maps. It is extremely rare to find these rendered in a colorblind safe manner with the only parties doing so other colorblind persons in my 20+ years in science.
https://www.nature.com/articles/s41467-020-19160-7
story645 commentedon Dec 8, 2023
Came here 'cause I'd really like some keywords for passing color and alpha into stem