bench-plot: per-file Vin-bin checkbox picker for combining runs

Combining multiple CSVs pooled everything blindly; when two runs cover
the same Vin bin the user now gets a checkbox grid (row per Vin bin,
column per file, counts per cell, overlapping bins highlighted) to
decide per bin which file contributes. Opens automatically on overlap;
--pick forces it (also useful to cut bad columns from a single file),
--no-pick suppresses it for scripted use.

Also: picker/file-dialog windows force themselves to the foreground
(spawned from the GUI button, Windows opened them hidden behind the
GUI); unreadable or still-being-written (empty) CSVs are skipped with a
note instead of killing the run; fatal errors show a messagebox since
stderr is invisible when launched from the GUI.

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

README.md

@@ -199,9 +199,18 @@ uv run bench-plot data_20260703_140000.csv
 uv run bench-plot run1_telem.csv run2_telem.csv --vin-bin 2 --save eff.png
 
 # options: --current iout|iin, --source auto|hioki|instr|stm (data logs),
-#          --min-pout W (default 5), --vin-bin V (default 1), --save PNG
+#          --min-pout W (default 5), --vin-bin V (default 1), --save PNG,
+#          --pick / --no-pick (per-file Vin-bin checkbox picker)
 ```
 
+Multiple files are pooled into one dataset. When two files cover the same
+Vin bin, a checkbox picker opens automatically -- one row per Vin bin, one
+column per file, cell numbers showing points per bin, overlapping bins
+highlighted -- so you decide per bin which file's data to use (e.g. keep the
+re-measured 72 V column from the new run, the rest from the old one).
+`--pick` forces the picker even without overlap (useful to cut bad columns
+from a single file); `--no-pick` skips it for scripted/batch use.
+
 Also reachable via the GUI's "Plot Eff..." button (Logging section, opens the
 same dialog preselecting the last log) and `plot_eff.bat` one level up
 (double-click for the dialog, or drag && drop CSV files onto it).

testbench/plot_eff.py

@@ -8,9 +8,15 @@ Auto-detects the three CSV formats produced by the tooling:
 Left panel: operating-point scatter (x = Vin, y = current, color = efficiency).
 Right panel: efficiency vs current, one curve per Vin bin.
 
+Multiple files are pooled into one dataset. When several files contain the
+same Vin bin, a checkbox picker opens (one row per bin, one column per file)
+so you can choose which file's data to use per bin; force it with --pick or
+suppress it with --no-pick.
+
 Usage:
     bench-plot data_20260703_120000.csv
     bench-plot run_telem.csv another_telem.csv --current iin --save eff.png
+    bench-plot sweep_a.csv sweep_b.csv --pick
 """
 
 from __future__ import annotations
@@ -18,7 +24,9 @@ from __future__ import annotations
 import argparse
 import csv
 import math
+import os
 import sys
+from collections import Counter
 
 import matplotlib.pyplot as plt
 import numpy as np
@@ -45,9 +53,29 @@ def _detect_format(header: list[str]) -> str:
         return "telem"
     if "voltage_set" in cols and "efficiency" in cols:
         return "sweep"
+    if not header:
+        raise ValueError("empty file (a log that is still being written?)")
     raise ValueError(f"unrecognized CSV header: {header[:6]}...")
 
 
+def _fatal(msg: str) -> None:
+    """Print the error AND show it in a messagebox: when spawned from the
+    GUI button nobody sees stderr, so exiting silently looks like a no-op."""
+    print(msg, file=sys.stderr)
+    if not os.environ.get("BENCH_PLOT_HEADLESS"):
+        try:
+            import tkinter as tk
+            from tkinter import messagebox
+            r = tk.Tk()
+            r.withdraw()
+            r.attributes("-topmost", True)
+            messagebox.showerror("bench-plot", msg, parent=r)
+            r.destroy()
+        except Exception:
+            pass
+    sys.exit(1)
+
+
 def _extract(path: str, source: str) -> tuple[dict, str]:
     """Read one CSV -> dict of float lists (vin_V, iin_A, iout_A, eff_pct,
     p_out_W) plus the efficiency-source label actually used."""
@@ -120,6 +148,110 @@ def _extract(path: str, source: str) -> tuple[dict, str]:
     return out, labels[source]
 
 
+class _BinPicker:
+    """Checkbox grid: one row per Vin bin, one column per file.
+
+    result is a list of selected-center sets (aligned with bin_table)
+    after OK, or None if the window was cancelled/closed.
+    """
+
+    ROW_H = 26
+
+    def __init__(self, bin_table: list[tuple[str, dict]], w: float):
+        import tkinter as tk
+        from tkinter import ttk
+        self._tk = tk
+        self.root = tk.Tk()
+        self.root.title(f"Select Vin bins per file ({w:g} V bins)")
+        # Spawned from the GUI's Plot Eff button, Windows denies foreground
+        # to the new process (the user is mid-click in the GUI) and the
+        # picker opens hidden BEHIND it -- keep this short-lived dialog
+        # on top and grab focus.
+        self.root.lift()
+        self.root.attributes("-topmost", True)
+        self.root.focus_force()
+        self._n_files = len(bin_table)
+        self._vars: dict[tuple[int, float], object] = {}
+        self.result: list[set] | None = None
+
+        centers = sorted({c for _, bins in bin_table for c in bins})
+        shared = {c for c in centers
+                  if sum(c in bins for _, bins in bin_table) > 1}
+
+        top = ttk.Frame(self.root, padding=6)
+        top.pack(fill="both", expand=True)
+        canvas = tk.Canvas(
+            top, highlightthickness=0,
+            height=min(self.ROW_H * (len(centers) + 2) + 8, 560))
+        vsb = ttk.Scrollbar(top, orient="vertical", command=canvas.yview)
+        grid = ttk.Frame(canvas)
+        grid.bind("<Configure>",
+                  lambda e: canvas.configure(scrollregion=canvas.bbox("all"),
+                                             width=grid.winfo_reqwidth()))
+        canvas.create_window((0, 0), window=grid, anchor="nw")
+        canvas.configure(yscrollcommand=vsb.set)
+        canvas.pack(side="left", fill="both", expand=True)
+        vsb.pack(side="right", fill="y")
+        canvas.bind_all("<MouseWheel>", lambda e: canvas.yview_scroll(
+            -1 if e.delta > 0 else 1, "units"))
+
+        ttk.Label(grid, text="Vin bin").grid(row=0, column=0, padx=6, sticky="w")
+        for fi, (name, _bins) in enumerate(bin_table):
+            short = name if len(name) <= 26 else "..." + name[-23:]
+            ttk.Label(grid, text=short).grid(row=0, column=1 + fi, padx=8)
+            ttk.Button(grid, text="all/none", width=8,
+                       command=lambda fi=fi: self._toggle_col(fi)
+                       ).grid(row=1, column=1 + fi, padx=8)
+        for ri, c in enumerate(centers):
+            lbl = ttk.Label(grid, text=f"{c:g} V")
+            if c in shared:
+                lbl.configure(foreground="#cc6600")
+            lbl.grid(row=2 + ri, column=0, padx=6, sticky="w")
+            for fi, (_name, bins) in enumerate(bin_table):
+                if c not in bins:
+                    continue
+                var = tk.BooleanVar(self.root, value=True)
+                self._vars[(fi, c)] = var
+                ttk.Checkbutton(grid, text=str(bins[c]), variable=var
+                                ).grid(row=2 + ri, column=1 + fi,
+                                       padx=8, sticky="w")
+
+        bar = ttk.Frame(self.root, padding=(6, 0, 6, 6))
+        bar.pack(fill="x")
+        ttk.Label(bar, text="numbers = points per bin, "
+                            "orange = bin present in several files"
+                  ).pack(side="left")
+        ttk.Button(bar, text="Cancel",
+                   command=self.root.destroy).pack(side="right", padx=4)
+        ttk.Button(bar, text="Plot", command=self._ok).pack(side="right")
+
+    def _toggle_col(self, fi: int) -> None:
+        cells = [v for (f, _c), v in self._vars.items() if f == fi]
+        state = not all(v.get() for v in cells)
+        for v in cells:
+            v.set(state)
+
+    def _ok(self) -> None:
+        self.result = [
+            {c for (f, c), v in self._vars.items() if f == fi and v.get()}
+            for fi in range(self._n_files)]
+        self.root.destroy()
+
+    def run(self) -> list[set] | None:
+        self.root.mainloop()
+        return self.result
+
+
+def _pick_bins(bin_table: list[tuple[str, dict]], w: float) -> list[set] | None:
+    """Show the picker; on a headless/Tk failure fall back to everything."""
+    try:
+        picker = _BinPicker(bin_table, w)
+    except Exception as e:
+        print(f"bin picker unavailable ({e}) - using all bins")
+        return [set(bins) for _, bins in bin_table]
+    return picker.run()
+
+
 def main() -> None:
     ap = argparse.ArgumentParser(
         description="Plot efficiency vs input voltage vs current from bench CSVs.")
@@ -137,6 +269,12 @@ def main() -> None:
                     help="drop points below this output power (default 5.0)")
     ap.add_argument("--save", metavar="PNG", help="write the figure instead of showing it")
     ap.add_argument("--title", default=None, help="figure title override")
+    g = ap.add_mutually_exclusive_group()
+    g.add_argument("--pick", action="store_true",
+                   help="always show the per-file Vin-bin checkbox picker")
+    g.add_argument("--no-pick", action="store_true",
+                   help="never show the picker (default: it opens when "
+                        "several files contain the same Vin bin)")
     args = ap.parse_args()
 
     if not args.csv:
@@ -144,33 +282,65 @@ def main() -> None:
         from tkinter import filedialog
         root = tk.Tk()
         root.withdraw()
+        root.attributes("-topmost", True)
         args.csv = list(filedialog.askopenfilenames(
+            parent=root,
             title="Select logged CSV(s) to plot",
             filetypes=[("CSV files", "*.csv"), ("All files", "*.*")]))
         root.destroy()
         if not args.csv:
             sys.exit("no file selected")
 
-    data: dict[str, list] = {k: [] for k in ("vin_V", "iin_A", "iout_A", "eff_pct", "p_out_W")}
-    labels = set()
+    # Load each file separately (kept apart for the per-file bin picker)
+    w = args.vin_bin
+    parts, labels, n_total, bad = [], set(), 0, []
     for path in args.csv:
-        part, label = _extract(path, args.source)
-        for k in data:
-            data[k].extend(part[k])
+        try:
+            part, label = _extract(path, args.source)
+        except Exception as e:
+            bad.append(f"{os.path.basename(path)}: {e}")
+            print(f"  skipping {os.path.basename(path)}: {e}")
+            continue
         labels.add(label)
+        fvin = np.asarray(part["vin_V"])
+        fcur = np.asarray(part["iin_A" if args.current == "iin" else "iout_A"])
+        feff = np.asarray(part["eff_pct"])
+        fpout = np.asarray(part["p_out_W"])
+        n_total += len(fvin)
+        keep = (np.isfinite(fvin) & np.isfinite(fcur) & np.isfinite(feff)
+                & (feff > 0.0) & (feff <= 105.0) & (fpout >= args.min_pout))
+        p = {"name": os.path.basename(path), "vin": fvin[keep],
+             "cur": fcur[keep], "eff": feff[keep]}
+        p["bins"] = np.round(p["vin"] / w) * w
+        parts.append(p)
+        if len(args.csv) > 1:
+            rng = (f"Vin bins {p['bins'].min():g}..{p['bins'].max():g} V"
+                   if len(p["vin"]) else "no usable points")
+            print(f"  {p['name']}: {len(p['vin'])} pts, {rng}")
 
-    vin = np.asarray(data["vin_V"])
-    cur = np.asarray(data["iin_A" if args.current == "iin" else "iout_A"])
-    eff = np.asarray(data["eff_pct"])
-    pout = np.asarray(data["p_out_W"])
+    if not parts:
+        _fatal("none of the selected files were readable:\n" + "\n".join(bad))
 
-    keep = (np.isfinite(vin) & np.isfinite(cur) & np.isfinite(eff)
-            & (eff > 0.0) & (eff <= 105.0) & (pout >= args.min_pout))
-    n_total = len(vin)
-    vin, cur, eff = vin[keep], cur[keep], eff[keep]
+    # Per-file Vin-bin selection: opens automatically when files overlap
+    bin_table = [(p["name"], Counter(p["bins"].tolist())) for p in parts]
+    files_per_bin = Counter(c for _, bins in bin_table for c in bins)
+    overlap = any(n > 1 for n in files_per_bin.values())
+    if not args.no_pick and (args.pick or (overlap and len(parts) > 1)):
+        sels = _pick_bins(bin_table, w)
+        if sels is None:
+            sys.exit("cancelled")
+        for p, sel in zip(parts, sels):
+            m = (np.isin(p["bins"], sorted(sel)) if sel
+                 else np.zeros(len(p["bins"]), dtype=bool))
+            for k in ("vin", "cur", "eff", "bins"):
+                p[k] = p[k][m]
+
+    vin = np.concatenate([p["vin"] for p in parts])
+    cur = np.concatenate([p["cur"] for p in parts])
+    eff = np.concatenate([p["eff"] for p in parts])
     if len(vin) == 0:
-        sys.exit(f"no usable points ({n_total} rows read; all filtered — "
-                 f"check --min-pout / --source)")
+        _fatal(f"no usable points ({n_total} rows read; all filtered or "
+               f"deselected — check --min-pout / --source / bin picker)")
 
     ipk = int(np.argmax(eff))
     print(f"{len(vin)} points ({n_total - len(vin)} filtered) | "
@@ -195,7 +365,6 @@ def main() -> None:
     ax1.grid(alpha=0.3)
 
     # Right: efficiency vs current, one mean curve per Vin bin
-    w = args.vin_bin
     centers = np.unique(np.round(vin / w) * w)
     cmap = plt.cm.plasma(np.linspace(0.0, 0.9, len(centers)))
     for color, c0 in zip(cmap, centers):