fix: EF5 lowercases gauge id in ts filename — read ts.<id.lower()>.<model>.csv so ungauged 'V…' points produce output (V22 hindcast + V23 nowcast were 0-row)

hf_data/pipeline.py

@@ -744,17 +744,13 @@ def _run_gauge_body(g, model, ef5_model, wb_model, t_start, t_end, use_mock,
             try:
                 with open(os.path.join(out_dir, "ef5_run.log"),
                           encoding="utf-8", errors="replace") as fh:
-                    log_txt = fh.read()
-                # gauge-placement lines are near the top; the failure tail is at
-                # the bottom — surface both, plus what EF5 actually wrote
-                head = log_txt[:1500]
-                tail = log_txt[-1500:]
+                    tail = fh.read()[-2000:]
+                # list what EF5 actually wrote — a name/case mismatch shows here
                 produced = sorted(os.path.basename(p) for p in
                                   glob.glob(os.path.join(out_dir, "ts.*.csv")))
-                yield ("status", f"⚠️ EF5 rc={ev['returncode']}, "
-                                 f"{len(new_rows)} new rows; expected {handle.ts_path}; "
-                                 f"wrote {produced or '[no ts.*.csv]'}\n"
-                                 f"--- log head ---\n{head}\n--- log tail ---\n{tail}")
+                yield ("status", f"⚠️ EF5 rc={ev['returncode']}, {len(new_rows)} new "
+                                 f"rows; expected {os.path.basename(handle.ts_path)}; "
+                                 f"wrote {produced or '[no ts.*.csv]'} — log tail:\n{tail}")
             except Exception:
                 pass
         yield (ev["kind"], ev)

hf_data/runner.py

@@ -140,7 +140,10 @@ def run_ef5(control_path: str, output_dir: str, gauge_id: str, model: str = "cre
     if os.name != "nt":            # line-buffer stdout so crashes keep the trail
         cmd = ["stdbuf", "-oL", "-eL"] + cmd
     proc = subprocess.Popen(cmd, stdout=log, stderr=subprocess.STDOUT, text=True)
-    ts = os.path.join(output_dir, f"ts.{gauge_id}.{model}.csv")
+    # EF5 lowercases the gauge id in the timeseries filename. All-digit USGS ids
+    # are unaffected, but a virtual point's "V…" id is written as "v…", so the
+    # expected path must lowercase too or the rows are never read (0-row runs).
+    ts = os.path.join(output_dir, f"ts.{gauge_id.lower()}.{model}.csv")
     return RunHandle(output_dir=output_dir, ts_path=ts, model=model, proc=proc)