tlambert03 / cmap Goto Github PK

The catalog structure has already gotten a little hard to reason about (particularly with regards to aliases). It needs to be documented better, commented, and cleaned up

• work on namespacing and colormap string name parsing
• #19
• consider plugin system... or interface for registration
• refine catalog mechanism
• continue to improve attributions and references

• cmap version: 0.2.0
• Python version: 3.9.18
• Operating System: macOS 14.4.1

Description

I am using cmap to generate a qualitative colormap for integer valued images in a napari plugin. The integer values can be greater than the number of colors in the qualitative/categorical colormap, so I want to cycle through those colors.

I might be missing something in the API, but I found this to be a little awkward right now. I think I need to do something like the following:

from cmap import Colormap
index: int = 0 # but could take any non-negative integer value
colormap = Colormap("colorbrewer:set1_8")
color = colormap(index % len(colormap.color_stops))

The main reason I found this awkward is because it is was hard to find Colormap.color_stops. Alternatively, I also found len(tuple(Colormap.iter_colors())), but this felt wasteful and even more verbose.

For more context, see the equivalent lines at a commit in a recent PR.

Ideas

I'm not sure what to suggest here since I only just started using cmap and I don't want to pollute what is a very concise API.

Adding something like Colormap.num_colors is probably enough to make me happy here (where num_colors is equal to len(colormap.color_stops)).

I also thought of something like Colormap.cycle(values) as an alternative to Colormap.__call__ that would do this automatically, but not sure if I like that either.

Summary

This is a super nit-picky feature request, so you should feel free to close and move on. The only way I could justify putting more thought in here is if the cyclic mapping is one that is likely to used frequently.

If there is interest, I'm happy to think a bit more and have a go at an implementation.

Otherwise, I am loving cmap and its docsite - the catalog and perceptual plots are beautiful!

as mentioned by @Jhsmit in #48 (comment)

would be nice to support bad/missing/over/under values as matplotlib does

• cmap version: 0.0.2
• Python version: any
• Operating System: any

Description

The hue plots, e.g. at https://cmap-docs.readthedocs.io/en/latest/catalog/cyclic/cmocean%3Aphase/#hue-saturation, show a discontinuity when the hue jumps from 0 to 100 or the other way round, but that's an artefact as hues are defined modulo 100 (in the scale here). It would be nice if the vertical line connecting 0 to 100 was not drawn...

What I Did

N/A

• cmap version: 0.2.1
• Python version: 3.11.3
• Operating System: MacOS

Description

Hi!
I am trying to use cmap to plot a colorbar with Matplotlib. To do so, I need to create it manually using plt.cm.ScalarMappableas in the following MWE. It seems not possible to use plt.cm.ScalarMappable from matplotlib to scale directly the colorbars from cmap. Is there a work-around to this?

What I Did

Here the code crash at sm = plt.cm.ScalarMappable(cmap=colormap, norm=norm). If I replace colormap = Colormap('crameri:roma') by colormap = plt.get_cmap('cividis') (see commented line nb 12) it works as expected.

import numpy as np
import matplotlib.pyplot as plt
from cmap import Colormap

# Create a range of x values
x = np.linspace(0, 2 * np.pi, 100)

# Create a range of y values as sine and cosine curves
y = np.array([np.sin(x + phi) for phi in np.linspace(0, np.pi, 10)])

# Create a colormap
# colormap = plt.get_cmap('cividis')
colormap = Colormap('crameri:roma')

# Normalize the y values for the colormap
y_min = np.min(y)
y_max = np.max(y)
normalize_y = (y - y_min) / (y_max - y_min)

fig, ax = plt.subplots(1, 1)

# Plot each curve with a color from the colormap
for i in range(len(y)):
    ax.plot(x, y[i], color=colormap(normalize_y[i][0]))

# Add a colorbar to the plot
norm = plt.Normalize(vmin=y_min, vmax=y_max)
sm = plt.cm.ScalarMappable(cmap=colormap, norm=norm)

cbar = fig.colorbar(sm, ax=ax, shrink=0.95)
fig.show()

• cmap version: 0.0.2
• Python version: any
• Operating System: any

Description

The colorcet names colorcet:CET_<letter><number> are not particularly memorable. There are some aliases available, which are listed both on the colorcet python package (e.g. "bgy", "bgyw", etc.), and also, if you want a more "official" source for the names, in Peter Kovesi's colorcet.m (http://www.peterkovesi.com/matlabfns/Colourmaps/colorcet.m : "Some colour maps have alternate names for convenience and readability." -- also note that the matlab code effectively makes the colormap names case-insensitive, but I guess lowercase reads nicer). It would be nice to have these aliases available. Thanks!

What I Did

N/A

• cmap version: 0.0.3
• Python version: any
• Operating System: any

Description

colorcet (the Python package) exposes a few variants on the glasbey color series (not really colormaps, but rather lists of maximally different colors), per holoviz/colorcet#11 (comment).

Admittedly these are constructs from the colorcet maintainers and not so much from the original glasbey paper, but it would be nice to include them, as I don't think there's any other color lists (designed for that use case) in cmap.

(Also, as a side point, glasbey should probably be categorized as "qualitative", not as "misc.")

What I Did

N/A

Hi, thanks for developing this very nice library!
I keep installing matplotlib/colorcet etc etc dependencies to have the full selection of colormaps I like to use :)

One set of color maps I like a lot is from Paul Tol: https://personal.sron.nl/~pault/
Are you willing to add / accept a PR for adding these colormaps?

There is python code provided here: https://personal.sron.nl/~pault/data/tol_colors.py
The code is licensed under BSD but I think the color data themselves should be no problem to add if we include the license?

Is there a preferred format for the colormaps or can be anything accepted by cmap.Colormap constructor?

import cmap as cm
cm_cmap = cm.Colormap(['#364B9A', '#4A7BB7', '#6EA6CD', '#98CAE1', '#C2E4EF',
                '#EAECCC', '#FEDA8B', '#FDB366', '#F67E4B', '#DD3D2D',
                '#A50026'], interpolation=False)

matplotlib result (expected):

cm_cmap.to_matplotlib()

However,

cmap = cm.Colormap('colorbrewer:Accent_5')
cmap

does produce a discrete colors png

• cmap version: 0.0.3
• Python version: any
• Operating System: any

Description

For cyclic colormaps, a useful operation can be to shift the position of the zero. For example, CET_C3 maps 0 to white, 0.5 to black, and 1 to white again; one could want to do the opposite (0 and 1 mapped to black, 0.5 mapped to white).
An API such as cmap.shifted(0.5) could be useful there (it's up to you how you want to handle the name kwarg, but note that unlike reverse()ing, there can be multiple meaningful shifts).

What I Did

N/A

• cmap version: 0.0.3
• Python version: any
• Operating System: any

Description

bwr is classified as diverging at https://matplotlib.org/stable/tutorials/colors/colormaps.html#diverging (and clearly looks diverging anyways...); cmap currently documents it as "misc."

What I Did

N/A

• cmap version: 0.0.1

Description

One nice feature of matplotlib colormaps is that when they are displayed in a notebook or other context where html makes sense they render themselves for the user. This is acheived through the __repr__html__ function. it would be great if cmap also had this functionality.

matplotlib implements this here:

https://github.com/matplotlib/matplotlib/blob/89452ceb1dad4a00d43c1ab571c5dad076977e52/lib/matplotlib/colors.py#L915

• cmap version: N/A
• Python version: N/A
• Operating System: N/A

Description

Clicking any of the chrisluts on the catalog page:
https://cmap-docs.readthedocs.io/en/latest/catalog/
returns 404 error (stable and latest Build #20340125).

However, using the burger menu does allow access, so the files are there, e.g.:
https://cmap-docs.readthedocs.io/en/latest/catalog/sequential/chrisluts%3AI_Blue/

It appears that the links in the catalog page use lower case letters, while the actual file names have caps, so I suspect this could affect other cmaps—if not now, then in the future.

Also, for chrisluts:I_Red the link from the catalog uses chrisluts:ied

Edit: Here's two of the problematic div (the nav ones are fine):

Ah, so the I_Red issue is this matching gone awry:

Edit2:
the cmap-name constructed here is correct:

What I Did

Was clicking the different color maps to see the cool visuals.