Honour colorbar=False for colour-only corrplot glyphs

[fesanghary]

Found while writing the beginner's-guide corrplot figure in #30, where I had to work around it.

The bug

corrplot(..., colorbar=False) is silently ignored whenever the glyph method encodes magnitude in the fill — method="color" or "shade", or either one as an upper/lower half. Those panels keep their colorbar while every other method honours the flag.

causalts/plotting/corrplot.py:

if not colorbar and not has_color_only:
    cl_pos = "n"

has_color_only is True for those methods, so the condition never fires. The intent was presumably that a colour-only plot is hard to read without a scale — defensible as a default, but it overrode an explicit argument with no warning, and left cl_pos="n" as the only way to get what you asked for.

Reproduction (axes added to a fresh figure; 1 = colorbar drawn):

method=circle   colorbar=False -> 0
method=color    colorbar=False -> 1   <-- ignored
method=shade    colorbar=False -> 1   <-- ignored
method=number   colorbar=False -> 0

The fix

colorbar=False now suppresses the colorbar for every method. The default is unchanged — colorbar defaults to True, so cl_pos still resolves to "r" and every existing call that doesn't pass the flag renders identically. has_color_only existed only for that guard and is removed.

Regression tests

tests/test_plotting.py had no corrplot coverage at all. Added four cases: a smoke test, colorbar=False across all seven glyph methods, the default still drawing a colorbar, and a mixed upper/lower with one colour-only half.

Verified red-then-green — reverting only the corrplot change fails two of them:

FAILED test_corrplot_colorbar_false_suppresses_for_every_method
FAILED test_corrplot_colorbar_false_with_color_only_half
2 failed, 2 passed

13 pass with the fix (9 before this PR).

One notebook figure changes

examples/plotting.ipynb cell 50 renders all seven glyphs with colorbar=False, so its committed figure showed the bug directly: five panels without a colorbar, two with. Re-executed and spliced that one cell — the other 63 keep their committed bytes. The panels are now uniformly sized, since color and shade are no longer squeezed by a colorbar the others don't have.

Pre-existing bug found, deliberately not fixed here

examples/plotting.ipynb does not execute top to bottom — on main, before this change. Cell 0 is an orphaned code cell sitting above the title and above the import cell, referencing plt, tp, gt1 and names1 before any of them exist:

cell 0: NameError: name 'plt' is not defined

So "Run All" fails on the very first cell for any reader. Confirmed on a pristine checkout, so it is not from this PR. Regenerating cell 50 needed allow_errors=True as a result; cell 50 itself runs clean because the imports land in cell 2.

The cell looks misplaced rather than wrong — it belongs somewhere in the target-node section. Left for a separate PR since fixing it properly means deciding where it goes, and moving it would re-render unrelated figures. Happy to do it next if you want.

🤖 Generated with Claude Code

[github-actions]

⚠️ Deprecation Warning: The deny-licenses option is deprecated for possible removal in the next major release. For more information, see issue 997.

Dependency Review

✅ No vulnerabilities or license issues or OpenSSF Scorecard issues found.

Scanned Files

None

[fesanghary]

Superseded by #34, which carries this colorbar fix along with two other corrplot rendering fixes (a clipped grid border, and a new diag='glyph' mode for directed matrices). The tests added here were adopted verbatim in #34 — they loop over all seven glyph methods, which is strictly stronger coverage than what I'd written. Closing in favour of that PR.