Fix logc formatter error

src/plopp/graphics/colormapper.py

@@ -11,6 +11,7 @@
 import scipp as sc
 from matplotlib.colorbar import ColorbarBase
 from matplotlib.colors import Colormap, LinearSegmentedColormap, LogNorm, Normalize
+from matplotlib.ticker import FuncFormatter
 
 from ..backends.matplotlib.utils import fig_to_bytes
 from ..core.limits import find_limits, fix_empty_range
@@ -186,6 +187,7 @@ def __init__(
                 fig = plt.Figure(figsize=(height_inches * 0.2, height_inches))
                 self.cax = fig.add_axes([0.05, 0.02, 0.2, 0.98])
             self.colorbar = ColorbarBase(self.cax, cmap=self.cmap, norm=self.normalizer)
+            self._update_colorbar_formatter()
             self.cax.yaxis.set_label_coords(-0.9, 0.5)
             if self._clabel is not None:
                 self.cax.set_ylabel(self._clabel)
@@ -276,6 +278,9 @@ def apply_limits(self):
             self.normalizer.vmax = self._cmax["data"]
 
         if self.colorbar is not None:
+            # Re-apply the formatter after updating normalizer limits,
+            # as matplotlib may have reset it when the normalizer changed
+            self._update_colorbar_formatter()
             self._update_colorbar_widget()
         self.notify_artists()
 
@@ -378,6 +383,33 @@ def ylabel(self) -> str | None:
     def ylabel(self, lab: str):
         self.clabel = lab
 
+    def _update_colorbar_formatter(self):
+        """
+        Update the colorbar tick formatter based on the current normalization.
+        For log scale, use a plain text formatter to avoid mathtext parsing errors
+        during tight_layout() calculations.
+        """
+        if self.colorbar is not None and self.cax is not None:
+            if self._logc:
+                # Use a custom formatter that produces plain text labels,
+                # avoiding mathtext notation that can cause parsing errors
+                # during tight_layout() when matplotlib measures text extent
+                def log_formatter(x, pos):
+                    if x <= 0:
+                        return ''
+                    exponent = int(np.round(np.log10(x)))
+                    mantissa = x / (10**exponent)
+                    # For powers of 10, show as "10^N" without math mode
+                    if abs(mantissa - 1.0) < 0.05:
+                        return f'1e{exponent:+d}'
+                    # For other values, use exponential notation
+                    return f'{x:.1e}'
+
+                self.cax.yaxis.set_major_formatter(FuncFormatter(log_formatter))
+            else:
+                # Reset to default formatter for linear scale
+                self.cax.yaxis.set_major_formatter(mpl.ticker.ScalarFormatter())
+
     def toggle_norm(self):
         """
         Toggle the norm flag, between `linear` and `log`.
@@ -388,6 +420,7 @@ def toggle_norm(self):
         self._cmax["data"] = -np.inf
         if self.colorbar is not None:
             self.colorbar.mappable.norm = self.normalizer
+            self._update_colorbar_formatter()
         self.autoscale()
         if self._canvas is not None:
             self._canvas.draw()

tests/graphics/colormapper_test.py

@@ -398,3 +398,43 @@ def test_toolbar_log_norm_button_toggles_colormapper_norm(
         assert fig.view.colormapper.norm == 'linear'
         fig.toolbar['lognorm'].value = True
         assert fig.view.colormapper.norm == 'log'
+
+
+def test_logc_tight_layout_does_not_raise_parse_error():
+    r"""
+    Test that calling tight_layout() on a figure with logc colorbar does not raise
+    a ValueError from matplotlib's mathtext parser. This reproduces the issue
+    where LogNorm tick labels like "$\mathdefault{10^{-33}}$" cause parsing errors.
+    """
+    import warnings
+
+    # Create data - the specific range from the user's report
+    da = sc.DataArray(
+        data=sc.array(dims=['y', 'x'], values=np.arange(1, 801).reshape(20, 40)),
+    )
+    mapper = ColorMapper(cbar=True, logc=True)
+    artist = DummyChild(data=da, colormapper=mapper)
+    mapper.add_artist('data', artist)
+
+    mapper.autoscale()
+
+    # Call tight_layout() on the colorbar figure
+    # This is what causes the error in the user's report
+    with warnings.catch_warnings(record=True) as w:
+        warnings.simplefilter("always")
+        try:
+            # Draw the figure first to ensure rendering is done
+            mapper.cax.get_figure().canvas.draw()
+            # Now call tight_layout which can trigger the mathtext parsing issue
+            mapper.cax.get_figure().tight_layout()
+        except ValueError as e:
+            # Check if this is the mathtext parsing error we're trying to fix
+            error_str = str(e)
+            if 'ParseException' in error_str and (
+                '$' in error_str or 'mathdefault' in error_str
+            ):
+                pytest.fail(
+                    f"tight_layout() raised mathtext parsing error with logc=True: {e}"
+                )
+            # Re-raise if it's a different ValueError
+            raise