detection_space: one curve and legend entry per unique legend label

Sweeps over parameters the legend label does not encode (the HWO config
sweeps scene.planet.contrast, which the 5-sigma contrast curve does not
depend on) stacked N visually identical curves with N repeated legend
entries. Curves and legend entries now dedupe on the legend label, with
a logged notice; byte-exact curve comparison was not usable because the
per-cell HOSTS exozodi draw perturbs the curves invisibly.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: Ben Mazin

nullsim/outputs/plots/detection_space.py

@@ -224,12 +224,27 @@ def _style_axes_atmosphere(axes: plt.Axes) -> None:
 
 
 def _plot_detection_curves(axes: plt.Axes, runs: RunCollection) -> None:
-    """One paper-style detection-mode contrast curve per sweep cell."""
+    """One paper-style detection-mode contrast curve per sweep cell.
+
+    Cells whose legend label duplicates an already-drawn one are skipped
+    (with a logged notice): sweeps over parameters the label does not encode
+    (e.g. scene.planet.contrast) otherwise stack visually identical lines
+    and repeated legend entries. Only the first such cell is drawn, so
+    per-cell Monte Carlo scatter (e.g. the HOSTS exozodi draw) is not
+    averaged, just represented by one cell.
+    """
 
+    seen: set[str] = set()
+    n_duplicates = 0
     for record in _paper_curve_records(runs):
         curve = record.result.state.results.contrast_curve
         if curve is None or curve.size == 0:
             continue
+        key = _curve_legend_label(record, mode="det")
+        if key in seen:
+            n_duplicates += 1
+            continue
+        seen.add(key)
         separations_mas = np.asarray(curve["separation_mas"], dtype=np.float64)
         contrasts = np.asarray(curve["contrast"], dtype=np.float64)
         order = np.argsort(separations_mas)
@@ -272,6 +287,12 @@ def _plot_detection_curves(axes: plt.Axes, runs: RunCollection) -> None:
                 linewidth=0,
                 zorder=1,
             )
+    if n_duplicates:
+        logging.info(
+            "detection_space: skipped %d sweep cell(s) whose legend label "
+            "duplicates an already-plotted cell.",
+            n_duplicates,
+        )
 
 
 # ----------------------------------------------------------------------
@@ -508,12 +529,17 @@ def _bin_widths_um(wavelengths_um: NDArrayFloat) -> NDArrayFloat:
 def _plot_characterization_curves(axes: plt.Axes, runs: RunCollection) -> None:
     """Overlay char-mode 5σ contrast curves as dashed lines matched in colour."""
 
+    seen: set[str] = set()
     for record in _paper_curve_records(runs):
         payload = _characterization_curve(record)
         if payload is None:
             continue
         separations_mas = np.asarray(payload["separations_mas"], dtype=np.float64)
         contrasts = np.asarray(payload["contrast_5sigma"], dtype=np.float64)
+        key = _curve_legend_label(record, mode="char")
+        if key in seen:
+            continue
+        seen.add(key)
         mask = (
             np.isfinite(separations_mas)
             & np.isfinite(contrasts)
@@ -876,17 +902,31 @@ def _attach_legends(
 
     handles: list[Line2D] = []
     labels: list[str] = []
+    seen_det: set[str] = set()
     for record in _paper_curve_records(runs):
-        if record.result.state.results.contrast_curve is None:
+        curve = record.result.state.results.contrast_curve
+        if curve is None:
+            continue
+        # Mirror the label dedupe in _plot_detection_curves: one legend
+        # entry per drawn line.
+        key = _curve_legend_label(record, mode="det")
+        if key in seen_det:
             continue
+        seen_det.add(key)
         handles.append(
             Line2D([], [], color=_DETECTION_COLOR, **_line_style_for_record(record))
         )
         labels.append(_curve_legend_label(record, mode="det"))
 
+    seen_char: set[str] = set()
     for record in _paper_curve_records(runs):
-        if _characterization_curve(record) is None:
+        payload = _characterization_curve(record)
+        if payload is None:
+            continue
+        key = _curve_legend_label(record, mode="char")
+        if key in seen_char:
             continue
+        seen_char.add(key)
         handles.append(
             Line2D(
                 [],

tests/outputs/plots/test_detection_space_ideal_coronagraph.py

@@ -161,3 +161,32 @@ def test_inputs_assemble_perfect_optics_rate() -> None:
     # I=1.5 solar analog through a clear 8 m primary and QE 0.9 lands at
     # a few 1e10 cps; pin the order of magnitude, not the zero point.
     assert 1.0e9 < inputs["stellar_rate_hz"] < 1.0e12
+
+
+def test_degenerate_sweep_draws_one_curve_and_one_legend_entry() -> None:
+    """Cells identical in label and curve collapse to a single line/entry.
+
+    The HWO config sweeps scene.planet.contrast, which the 5-sigma contrast
+    curve does not depend on: pre-fix the figure stacked five identical
+    curves and legend entries.
+    """
+
+    from nullsim.outputs.plots.detection_space import (
+        _attach_legends,
+        _plot_detection_curves,
+    )
+
+    records = tuple(_record(orders=()) for _ in range(5))
+    collection = RunCollection(
+        config=records[0].config, records=records, output_dir=Path(".")
+    )
+    figure, axes = plt.subplots()
+    try:
+        _plot_detection_curves(axes, collection)
+        assert len(axes.lines) == 1
+        _attach_legends(axes, (), collection)
+        legend = axes.get_legend()
+        labels = [text.get_text() for text in legend.get_texts()]
+        assert len([lab for lab in labels if "HWO" in lab]) == 1
+    finally:
+        plt.close(figure)