Fix repeat ownership and close out capture evidence

Start each repeat after its own manifest entry, refuse packets from a restarted or stopped camera, keep a simulated repeat from reading as recorded evidence, and decide completeness from the writer report. Add the operator summary (four files plus an atomic pointer) and read-only video review of a hold window. Verified with the full Autocal suite (567 passed), the shared recorder/playback files (57 passed), the lint gate and the documentation gate; the two pre-existing panel_check layout failures also fail on a clean HEAD worktree.
2026-10-17 10:44:30 +08:00
parent d7540f52ec
commit 19426c3c4a
32 changed files with 5461 additions and 191 deletions
@@ -0,0 +1,265 @@
+"""One window table over both recording formats, with nothing invented."""
+
+from __future__ import annotations
+
+import pytest
+
+from evidence_helpers import (
+    BASE_NS,
+    FORCES_FIELDS,
+    FRAMES_FIELDS,
+    Hold,
+    HOLD_STRIDE_NS,
+    multipoint_session,
+    read_csv,
+    single_session,
+    write_csv,
+)
+from orisys_autocal.summary import (
+    HOLD_WINDOW_METHOD,
+    LEGACY_WINDOW_METHOD,
+    REASON_INSUFFICIENT_SAMPLES,
+    REASON_LEGACY,
+    REASON_NO_FRAME_SOURCE,
+    REASON_REMEASURED,
+    SCHEMA_VERSION,
+    build_summary,
+    legacy_window_rows,
+)
+
+
+def _levels() -> tuple[float, ...]:
+    return (5.0, 10.0)
+
+
+def _means(summary: dict) -> list[float | None]:
+    return [row["FN_mean"] for row in summary["windows"]]
+
+
+def test_both_formats_produce_the_same_identity_and_means(tmp_path):
+    multipoint = build_summary(multipoint_session(
+        tmp_path / "multi", sites=("P1",), repeats=1, levels=_levels()))
+    single = build_summary(single_session(
+        tmp_path / "single", sites=("site-001",), trials_per_site=1, levels=_levels()))
+
+    for summary in (multipoint, single):
+        assert summary["schema_version"] == SCHEMA_VERSION
+        assert summary["formal_conclusion"] is False
+        assert summary["units"]["raw"] == "raw"
+        assert summary["methods"]["fit"] == "raw_on_reference_ols_v1"
+        assert len(summary["windows"]) == 5
+        assert [row["branch"] for row in summary["windows"]] == [
+            "baseline", "load", "load", "unload", "unload"]
+        assert [row["hold_id"] for row in summary["windows"]] == [1, 2, 3, 4, 5]
+        assert all(row["session_id"] for row in summary["windows"])
+        assert all(row["trial_id"] for row in summary["windows"])
+        assert all(row["source_file"].endswith("reference.csv") for row in summary["windows"])
+        assert all(row["unit"] == "raw" for row in summary["windows"])
+        assert all(row["window_method"] == HOLD_WINDOW_METHOD for row in summary["windows"])
+        assert all(row["stability_verified"] is False for row in summary["windows"])
+
+    assert _means(multipoint) == pytest.approx([100.0, 300.0, 500.0, 300.0, 100.0])
+    assert _means(single) == pytest.approx(_means(multipoint))
+    assert [row["samples_valid"] for row in multipoint["windows"]] == [2, 2, 2, 2, 2]
+    assert all(row["capture_status"] == "unknown" for row in multipoint["windows"])
+
+
+def test_a_six_axis_site_keeps_raw_components_apart(tmp_path):
+    summary = build_summary(multipoint_session(
+        tmp_path / "rec", sites=("P1",), repeats=1, levels=_levels(),
+        raw_base=100.0, raw_per_n=40.0))
+
+    load = summary["windows"][2]
+    assert load["FN_mean"] == pytest.approx(500.0)
+    assert load["FX_mean"] == pytest.approx(0.0)
+    assert load["FY_mean"] == pytest.approx(0.0)
+    assert load["FN_std"] == pytest.approx(0.0)
+    # The newton columns never appear in this table.
+    assert "FN_mean" in load and "FN_N" not in load
+    sensitivity = {row["component"]: row for row in summary["sensitivity"]}
+    assert sensitivity["FN"]["slope_raw_per_N"] == pytest.approx(40.0)
+    assert sensitivity["FN"]["status"] == "available"
+
+
+def test_a_remeasured_hold_is_listed_with_its_reason(tmp_path):
+    replaced = Hold(
+        step_index=1, branch="load", target_load_n=5.0, reference_n=5.0,
+        start_ns=BASE_NS + 200 * HOLD_STRIDE_NS,
+        end_ns=BASE_NS + 200 * HOLD_STRIDE_NS + 4_000_000_000,
+        complete=0,
+    )
+    summary = build_summary(multipoint_session(
+        tmp_path / "rec", sites=("P1",), repeats=1, levels=_levels(),
+        extra_holds=[replaced]))
+
+    assert len(summary["windows"]) == 6
+    replaced_rows = [row for row in summary["windows"] if row["excluded_reason"] == REASON_REMEASURED]
+    assert len(replaced_rows) == 1
+    assert replaced_rows[0]["hold_id"] == 6
+    assert replaced_rows[0]["eligible"] is False
+    assert replaced_rows[0]["FN_mean"] is None
+    assert replaced_rows[0]["samples_total"] == 0
+
+
+def test_a_missing_frame_source_is_reported_not_filled_in(tmp_path):
+    summary = build_summary(multipoint_session(
+        tmp_path / "rec", sites=("P1",), repeats=1, levels=_levels(),
+        with_capture=False))
+
+    assert all(row["eligible"] is False for row in summary["windows"])
+    assert all(row["excluded_reason"] == REASON_NO_FRAME_SOURCE for row in summary["windows"])
+    assert all(row["FN_mean"] is None for row in summary["windows"])
+    assert all(row["capture_status"] == "missing" for row in summary["windows"])
+    assert REASON_NO_FRAME_SOURCE in summary["missing_reasons"]
+    assert summary["sensitivity"] == []
+
+
+def test_a_confirmed_window_uses_the_reviewed_interval(tmp_path):
+    root = multipoint_session(tmp_path / "rec", sites=("P1",), repeats=1, levels=_levels())
+    from orisys_autocal.session_review import review_session
+
+    first = review_session(root)
+    window = first["sites"][0]["windows"][2]
+    selections = {f"{window['trial_id']}:{window['step_index']}": {
+        "t_start_host_mono_ns": window["hold_start_host_mono_ns"] + 1,
+        "t_end_host_mono_ns": window["hold_end_host_mono_ns"],
+        "stability_verified": True,
+    }}
+
+    summary = build_summary(root, selections=selections)
+
+    confirmed = [row for row in summary["windows"] if row["stability_verified"]]
+    assert len(confirmed) == 1
+    assert confirmed[0]["window_method"] == "review_selection_v1"
+    assert confirmed[0]["review_status"] == "confirmed"
+    assert confirmed[0]["t_start_host_mono_ns"] == window["hold_start_host_mono_ns"] + 1
+    assert confirmed[0]["samples_valid"] == 3
+    assert summary["selections_sha256"]
+    assert {row["review_status"] for row in summary["windows"] if not row["stability_verified"]} == {"candidate"}
+
+
+def test_out_of_order_force_rows_do_not_change_the_means(tmp_path):
+    root = multipoint_session(tmp_path / "rec", sites=("P1",), repeats=1, levels=_levels())
+    before = build_summary(root)
+    capture = root / "sites" / "P1" / "repeat-001" / "capture"
+    rows = read_csv(capture / "forces.csv")
+    write_csv(capture / "forces.csv", FORCES_FIELDS, list(reversed(rows)))
+
+    after = build_summary(root)
+
+    assert _means(after) == pytest.approx(_means(before))
+
+
+def test_a_repeated_frame_identity_excludes_the_window_and_is_reported(tmp_path):
+    root = multipoint_session(tmp_path / "rec", sites=("P1",), repeats=1, levels=_levels())
+    capture = root / "sites" / "P1" / "repeat-001" / "capture"
+    frames = read_csv(capture / "frames.csv")
+    forces = read_csv(capture / "forces.csv")
+    last_frames = frames[-3:]
+    last_forces = forces[-3:]
+    for frame, force in zip(last_frames, last_forces):
+        copy = dict(frame)
+        copy["t_host_mono_ns"] = int(frame["t_host_mono_ns"]) + 1
+        frames.append(copy)
+        copied = dict(force)
+        copied["FN_raw"] = repr(float(force["FN_raw"]) + 1.0)
+        forces.append(copied)
+    write_csv(capture / "frames.csv", FRAMES_FIELDS, frames)
+    write_csv(capture / "forces.csv", FORCES_FIELDS, forces)
+
+    summary = build_summary(root)
+
+    assert any(issue["reason"] == "duplicate_frame_identity" for issue in summary["issues"])
+    last = summary["windows"][-1]
+    assert last["samples_total"] == 0
+    assert last["excluded_reason"] == REASON_INSUFFICIENT_SAMPLES
+    assert last["eligible"] is False
+    # The earlier windows still have their own, unambiguous samples.
+    assert all(row["samples_valid"] == 2 for row in summary["windows"][:-1])
+
+
+def test_two_sessions_are_never_merged(tmp_path):
+    first = build_summary(multipoint_session(
+        tmp_path / "a", sites=("P1",), repeats=1, levels=_levels()))
+    second = build_summary(multipoint_session(
+        tmp_path / "b", sites=("P2",), repeats=2, levels=(5.0,)))
+
+    assert {row["session_id"] for row in first["windows"]} == {first["session_id"]}
+    assert {row["session_id"] for row in second["windows"]} == {second["session_id"]}
+    assert first["session_id"] != second["session_id"]
+    assert len(second["windows"]) == 6  # baseline, load, unload per repeat
+
+
+def test_an_old_stable_means_table_is_kept_but_never_fitted(tmp_path):
+    path = tmp_path / "stable_means.csv"
+    write_csv(path, [
+        "site_id", "U", "V", "负载类型", "repeat_index", "step_index", "branch",
+        "target_load_n", "samples", "FN_mean", "FX_mean", "FY_mean", "unit",
+    ], [{
+        "site_id": "20261016point0", "U": "-145", "V": "604", "负载类型": "砝码",
+        "repeat_index": 1, "step_index": 1, "branch": "load", "target_load_n": 1.0,
+        "samples": 181, "FN_mean": 10013.7, "FX_mean": -6484.37, "FY_mean": 707.088,
+        "unit": "raw",
+    }])
+
+    rows = legacy_window_rows(path, summary_id="legacy-1")
+
+    assert len(rows) == 1
+    row = rows[0]
+    assert row["window_method"] == LEGACY_WINDOW_METHOD
+    assert row["eligible"] is False
+    assert row["excluded_reason"] == REASON_LEGACY
+    assert row["FN_mean"] == pytest.approx(10013.7)
+    assert row["t_start_host_mono_ns"] is None and row["FN_std"] is None
+    assert row["unit"] == "raw"
+
+
+def test_the_table_states_what_it_did_not_conclude(tmp_path):
+    summary = build_summary(multipoint_session(
+        tmp_path / "rec", sites=("P1",), repeats=1, levels=_levels()))
+
+    assert summary["formal_conclusion"] is False
+    assert summary["methods"]["window_unverified"] == HOLD_WINDOW_METHOD
+    assert summary["input_sha256"]
+    assert summary["software"]["numpy"]
+    assert summary["generated_at"].startswith("20")
+    assert summary["summary_id"]
+    assert all(row["summary_id"] == summary["summary_id"] for row in summary["windows"])
+    assert summary["sensitivity_summary"]
+    assert summary["sensitivity_summary"][0]["trials"] == 1
+    assert summary["sensitivity_summary"][0]["slope_std_raw_per_N"] is None
+
+
+def test_a_sensitivity_row_exists_per_trial_and_review_set(tmp_path):
+    summary = build_summary(multipoint_session(
+        tmp_path / "rec", sites=("P1",), repeats=2, levels=_levels()))
+
+    fn_rows = [row for row in summary["sensitivity"] if row["component"] == "FN"]
+    assert len(fn_rows) == 2
+    assert {row["review_status"] for row in fn_rows} == {"candidate"}
+    assert [row["window_count"] for row in fn_rows] == [3, 3]
+    assert all(row["status"] == "available" for row in fn_rows)
+    assert all(row["slope_raw_per_N"] == pytest.approx(40.0) for row in fn_rows)
+    assert all(row["endpoint_raw_per_N"] == pytest.approx(40.0) for row in fn_rows)
+    assert all(row["initial_step_raw"] == pytest.approx(200.0) for row in fn_rows)
+
+
+def test_an_axis_without_samples_is_unavailable_without_blocking_the_others(tmp_path):
+    root = multipoint_session(tmp_path / "rec", sites=("P1",), repeats=1, levels=_levels())
+    capture = root / "sites" / "P1" / "repeat-001" / "capture"
+    forces = read_csv(capture / "forces.csv")
+    for row in forces:
+        row["FX_raw"] = ""
+    write_csv(capture / "forces.csv", FORCES_FIELDS, forces)
+
+    summary = build_summary(root)
+
+    by_axis = {row["component"]: row for row in summary["sensitivity"]}
+    assert by_axis["FN"]["status"] == "available"
+    assert by_axis["FX"]["status"] == "no_data"
+    # A constant component has a unitless slope of zero and no R2; that is a
+    # flat signal, not a gain.
+    assert by_axis["FY"]["status"] == "flat_signal"
+    assert by_axis["FY"]["R2"] is None
+    assert by_axis["FY"]["slope_raw_per_N"] == pytest.approx(0.0)
+    assert summary["windows"][0]["axis_samples"]["FX"] == 0