[tlambert03]

fixes #153 by going with option 2 from the discussion there (thanks @matthiasschabel): scalar input with bytes=True now returns a tuple of four np.uint8, exactly like matplotlib does, rather than raising.

In [1]: from cmap import Colormap

In [2]: Colormap("viridis")(0.5, bytes=True)
Out[2]: (np.uint8(32), np.uint8(144), np.uint8(140), np.uint8(255))

In [3]: Colormap("viridis")(0.5)  # unchanged
Out[3]: #21918C

I went with the tuple (rather than a (4,) array) to match mpl's scalar behavior... it also means cm(0.5, bytes=True) == (32, 144, 140, 255) just works and the result is hashable. The only thing you lose vs an array is .dtype. (I didn't reuse RGBA8 here, since its alpha is a float in 0-1, while bytes mode gives alpha in 0-255.)

The overloads now split on bytes, so type checkers see Color for bytes=False and tuple[np.uint8, ...] for bytes=True.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.91%. Comparing base (0e5d91e) to head (1bf8413).

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #168   +/-   ##
=======================================
  Coverage   95.90%   95.91%           
=======================================
  Files         168      168           
  Lines        2248     2252    +4     
=======================================
+ Hits         2156     2160    +4     
  Misses         92       92           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.