elliot-stack 1.0.36 → 1.0.37

package/skills/estack-read-claude-session-history/scripts/tests/test_engagement.py

@@ -1,239 +1,239 @@
-"""Tests for the engagement mode (attention-time accounting)."""
-
-import json
-import os
-import shutil
-import subprocess
-import sys
-from datetime import datetime, timedelta
-
-import pytest
-
-import read_transcript as RT
-from lib import parser as PR
-
-
-def _run_cli(cli_path, *args, env_overrides=None):
-    env = dict(os.environ)
-    env["PYTHONIOENCODING"] = "utf-8"
-    if env_overrides:
-        env.update(env_overrides)
-    return subprocess.run(
-        [sys.executable, str(cli_path), *args],
-        capture_output=True,
-        text=True,
-        encoding="utf-8",
-        env=env,
-    )
-
-
-@pytest.fixture
-def utc_tz():
-    PR.set_timezone("UTC")
-    yield
-    PR.set_timezone(None)
-
-
-def _fake_root(tmp_path, **projects):
-    """Build a fake projects root: {project_dir_name: [(fixture, uuid), ...]}."""
-    root = tmp_path / "projects"
-    for proj_name, files in projects.items():
-        pd = root / proj_name
-        pd.mkdir(parents=True)
-        for src, uuid in files:
-            shutil.copy(src, pd / f"{uuid}.jsonl")
-    return root
-
-
-def _build(root, **kw):
-    defaults = dict(
-        report_dirs=None,
-        report_file=None,
-        since=datetime(2026, 5, 1),
-        until=datetime(2026, 5, 2),
-        break_minutes=10,
-        current_uuid=None,
-    )
-    defaults.update(kw)
-    return RT.build_engagement(root, **defaults)
-
-
-# ── unit: gap math ───────────────────────────────────────────────────────────
-
-def test_gaps_straddle_threshold(fixtures_dir, tmp_path, utc_tz):
-    root = _fake_root(
-        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-gaps.jsonl", "aaaa1111")]}
-    )
-    data = _build(root)
-    (s,) = data["sessions"].values()
-    # 5m + 3m active, 32m break (Claude finished 10:09, reply 10:40 — too late),
-    # then 5m active = 13m.
-    assert s["active"] == timedelta(minutes=13)
-    assert s["user_messages"] == 5
-    assert len(data["breaks"]) == 1
-    assert data["breaks"][0][0] == datetime(2026, 5, 1, 10, 8)
-    assert data["breaks"][0][1] == datetime(2026, 5, 1, 10, 40)
-
-
-def test_waiting_on_claude_credit(fixtures_dir, tmp_path, utc_tz):
-    root = _fake_root(
-        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-waiting.jsonl", "bbbb2222")]}
-    )
-    data = _build(root)
-    (s,) = data["sessions"].values()
-    # 32m gap, but Claude's last event was 10:30 and the user replied 10:32
-    # (2m ≤ 10m) — the whole gap counts as waiting-on-Claude.
-    assert s["active"] == timedelta(minutes=32)
-    assert data["breaks"] == []
-
-
-def test_noise_exclusion(fixtures_dir, tmp_path, utc_tz):
-    root = _fake_root(
-        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-noise.jsonl", "cccc3333")]}
-    )
-    data = _build(root)
-    (s,) = data["sessions"].values()
-    # Only the two real prompts (10:01, 10:06) count: compact continuation,
-    # isMeta injections, and tool results are all excluded from the user stream.
-    assert s["user_messages"] == 2
-    assert s["first"] == datetime(2026, 5, 1, 10, 1)
-    assert s["active"] == timedelta(minutes=5)
-
-
-def test_parallel_sessions_never_double_count(fixtures_dir, tmp_path, utc_tz):
-    root = _fake_root(
-        tmp_path,
-        **{
-            "C--proj-a": [(fixtures_dir / "engagement-parallel-a.jsonl", "aaaa1111")],
-            "C--proj-b": [(fixtures_dir / "engagement-parallel-b.jsonl", "bbbb2222")],
-        },
-    )
-    data = _build(root)
-    by_uuid = {s["summary"]["uuid"]: s for s in data["sessions"].values()}
-    # Stream: 10:00 A, 10:10 B, 10:20 B, 10:30 A. Each segment goes to the
-    # session of the LATER prompt: B gets 10:00–10:20 (20m), A gets 10:20–10:30
-    # (10m). Total 30m == wall clock, not the naive 40m.
-    assert by_uuid["bbbb2222"]["active"] == timedelta(minutes=20)
-    assert by_uuid["aaaa1111"]["active"] == timedelta(minutes=10)
-    total = sum((s["active"] for s in data["sessions"].values()), timedelta())
-    assert total == timedelta(minutes=30)
-
-
-def test_single_message_session(tmp_path, utc_tz):
-    root = tmp_path / "projects"
-    pd = root / "C--proj-a"
-    pd.mkdir(parents=True)
-    (pd / "dddd4444.jsonl").write_text(
-        '{"type":"user","timestamp":"2026-05-01T10:00:00Z",'
-        '"message":{"role":"user","content":"one and done"}}\n',
-        encoding="utf-8",
-    )
-    data = _build(root)
-    (s,) = data["sessions"].values()
-    assert s["active"] == timedelta(0)
-    assert s["user_messages"] == 1
-
-
-def test_report_scope_filters_but_stream_stays_global(fixtures_dir, tmp_path, utc_tz):
-    root = _fake_root(
-        tmp_path,
-        **{
-            "C--proj-a": [(fixtures_dir / "engagement-parallel-a.jsonl", "aaaa1111")],
-            "C--proj-b": [(fixtures_dir / "engagement-parallel-b.jsonl", "bbbb2222")],
-        },
-    )
-    data = _build(root, report_dirs=[root / "C--proj-a"])
-    # Only A is reported, but B's prompts still split the stream — A gets its
-    # interval-correct 10m, not a naive 30m.
-    (s,) = data["sessions"].values()
-    assert s["summary"]["uuid"] == "aaaa1111"
-    assert s["active"] == timedelta(minutes=10)
-
-
-# ── CLI ──────────────────────────────────────────────────────────────────────
-
-def test_engagement_cli_text(cli_path, fixtures_dir, tmp_path):
-    root = _fake_root(
-        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-gaps.jsonl", "aaaa1111")]}
-    )
-    r = _run_cli(
-        cli_path, "--root", str(root), "--tz", "UTC",
-        "--mode", "engagement", "--date", "2026-05-01",
-    )
-    assert r.returncode == 0
-    assert "13m" in r.stdout
-    assert "break=10m" in r.stdout
-    assert "Breaks >10m" in r.stdout
-
-
-def test_engagement_cli_json(cli_path, fixtures_dir, tmp_path):
-    root = _fake_root(
-        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-gaps.jsonl", "aaaa1111")]}
-    )
-    r = _run_cli(
-        cli_path, "--root", str(root), "--tz", "UTC",
-        "--mode", "engagement", "--date", "2026-05-01", "--format", "json",
-    )
-    assert r.returncode == 0
-    data = json.loads(r.stdout)
-    assert data["break_minutes"] == 10
-    assert data["totals"]["active_minutes"] == 13
-    assert data["totals"]["sessions"] == 1
-    s = data["sessions"][0]
-    assert s["uuid"] == "aaaa1111"
-    assert s["elapsed_minutes"] == 45
-    assert s["ratio"] == round(13 / 45, 2)
-    assert len(data["stream_breaks"]) == 1
-    assert data["stream_breaks"][0]["minutes"] == 32
-
-
-def test_engagement_cli_file_window_derived(cli_path, fixtures_dir, tmp_path):
-    root = _fake_root(
-        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-waiting.jsonl", "bbbb2222")]}
-    )
-    f = root / "C--proj-a" / "bbbb2222.jsonl"
-    r = _run_cli(
-        cli_path, "--root", str(root), "--tz", "UTC",
-        "--mode", "engagement", "--file", str(f),
-    )
-    assert r.returncode == 0
-    assert "32m" in r.stdout
-    assert "Prompt gaps" in r.stdout
-
-
-def test_engagement_cli_custom_break(cli_path, fixtures_dir, tmp_path):
-    root = _fake_root(
-        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-gaps.jsonl", "aaaa1111")]}
-    )
-    r = _run_cli(
-        cli_path, "--root", str(root), "--tz", "UTC",
-        "--mode", "engagement", "--date", "2026-05-01",
-        "--break", "1h", "--format", "json",
-    )
-    assert r.returncode == 0
-    data = json.loads(r.stdout)
-    # 1h threshold swallows the 32m gap — everything is active: 45m.
-    assert data["totals"]["active_minutes"] == 45
-    assert data["stream_breaks"] == []
-
-
-def test_engagement_cli_invalid_break(cli_path, tmp_path):
-    root = tmp_path / "projects"
-    root.mkdir()
-    r = _run_cli(
-        cli_path, "--root", str(root),
-        "--mode", "engagement", "--break", "soon",
-    )
-    assert r.returncode == 1
-
-
-def test_engagement_cli_empty_range(cli_path, fixtures_dir, tmp_path):
-    root = _fake_root(
-        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-gaps.jsonl", "aaaa1111")]}
-    )
-    r = _run_cli(
-        cli_path, "--root", str(root), "--tz", "UTC",
-        "--mode", "engagement", "--date", "2020-01-01",
-    )
-    assert r.returncode == 0
-    assert "no user messages" in r.stdout
+"""Tests for the engagement mode (attention-time accounting)."""
+
+import json
+import os
+import shutil
+import subprocess
+import sys
+from datetime import datetime, timedelta
+
+import pytest
+
+import read_transcript as RT
+from lib import parser as PR
+
+
+def _run_cli(cli_path, *args, env_overrides=None):
+    env = dict(os.environ)
+    env["PYTHONIOENCODING"] = "utf-8"
+    if env_overrides:
+        env.update(env_overrides)
+    return subprocess.run(
+        [sys.executable, str(cli_path), *args],
+        capture_output=True,
+        text=True,
+        encoding="utf-8",
+        env=env,
+    )
+
+
+@pytest.fixture
+def utc_tz():
+    PR.set_timezone("UTC")
+    yield
+    PR.set_timezone(None)
+
+
+def _fake_root(tmp_path, **projects):
+    """Build a fake projects root: {project_dir_name: [(fixture, uuid), ...]}."""
+    root = tmp_path / "projects"
+    for proj_name, files in projects.items():
+        pd = root / proj_name
+        pd.mkdir(parents=True)
+        for src, uuid in files:
+            shutil.copy(src, pd / f"{uuid}.jsonl")
+    return root
+
+
+def _build(root, **kw):
+    defaults = dict(
+        report_dirs=None,
+        report_file=None,
+        since=datetime(2026, 5, 1),
+        until=datetime(2026, 5, 2),
+        break_minutes=10,
+        current_uuid=None,
+    )
+    defaults.update(kw)
+    return RT.build_engagement(root, **defaults)
+
+
+# ── unit: gap math ───────────────────────────────────────────────────────────
+
+def test_gaps_straddle_threshold(fixtures_dir, tmp_path, utc_tz):
+    root = _fake_root(
+        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-gaps.jsonl", "aaaa1111")]}
+    )
+    data = _build(root)
+    (s,) = data["sessions"].values()
+    # 5m + 3m active, 32m break (Claude finished 10:09, reply 10:40 — too late),
+    # then 5m active = 13m.
+    assert s["active"] == timedelta(minutes=13)
+    assert s["user_messages"] == 5
+    assert len(data["breaks"]) == 1
+    assert data["breaks"][0][0] == datetime(2026, 5, 1, 10, 8)
+    assert data["breaks"][0][1] == datetime(2026, 5, 1, 10, 40)
+
+
+def test_waiting_on_claude_credit(fixtures_dir, tmp_path, utc_tz):
+    root = _fake_root(
+        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-waiting.jsonl", "bbbb2222")]}
+    )
+    data = _build(root)
+    (s,) = data["sessions"].values()
+    # 32m gap, but Claude's last event was 10:30 and the user replied 10:32
+    # (2m ≤ 10m) — the whole gap counts as waiting-on-Claude.
+    assert s["active"] == timedelta(minutes=32)
+    assert data["breaks"] == []
+
+
+def test_noise_exclusion(fixtures_dir, tmp_path, utc_tz):
+    root = _fake_root(
+        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-noise.jsonl", "cccc3333")]}
+    )
+    data = _build(root)
+    (s,) = data["sessions"].values()
+    # Only the two real prompts (10:01, 10:06) count: compact continuation,
+    # isMeta injections, and tool results are all excluded from the user stream.
+    assert s["user_messages"] == 2
+    assert s["first"] == datetime(2026, 5, 1, 10, 1)
+    assert s["active"] == timedelta(minutes=5)
+
+
+def test_parallel_sessions_never_double_count(fixtures_dir, tmp_path, utc_tz):
+    root = _fake_root(
+        tmp_path,
+        **{
+            "C--proj-a": [(fixtures_dir / "engagement-parallel-a.jsonl", "aaaa1111")],
+            "C--proj-b": [(fixtures_dir / "engagement-parallel-b.jsonl", "bbbb2222")],
+        },
+    )
+    data = _build(root)
+    by_uuid = {s["summary"]["uuid"]: s for s in data["sessions"].values()}
+    # Stream: 10:00 A, 10:10 B, 10:20 B, 10:30 A. Each segment goes to the
+    # session of the LATER prompt: B gets 10:00–10:20 (20m), A gets 10:20–10:30
+    # (10m). Total 30m == wall clock, not the naive 40m.
+    assert by_uuid["bbbb2222"]["active"] == timedelta(minutes=20)
+    assert by_uuid["aaaa1111"]["active"] == timedelta(minutes=10)
+    total = sum((s["active"] for s in data["sessions"].values()), timedelta())
+    assert total == timedelta(minutes=30)
+
+
+def test_single_message_session(tmp_path, utc_tz):
+    root = tmp_path / "projects"
+    pd = root / "C--proj-a"
+    pd.mkdir(parents=True)
+    (pd / "dddd4444.jsonl").write_text(
+        '{"type":"user","timestamp":"2026-05-01T10:00:00Z",'
+        '"message":{"role":"user","content":"one and done"}}\n',
+        encoding="utf-8",
+    )
+    data = _build(root)
+    (s,) = data["sessions"].values()
+    assert s["active"] == timedelta(0)
+    assert s["user_messages"] == 1
+
+
+def test_report_scope_filters_but_stream_stays_global(fixtures_dir, tmp_path, utc_tz):
+    root = _fake_root(
+        tmp_path,
+        **{
+            "C--proj-a": [(fixtures_dir / "engagement-parallel-a.jsonl", "aaaa1111")],
+            "C--proj-b": [(fixtures_dir / "engagement-parallel-b.jsonl", "bbbb2222")],
+        },
+    )
+    data = _build(root, report_dirs=[root / "C--proj-a"])
+    # Only A is reported, but B's prompts still split the stream — A gets its
+    # interval-correct 10m, not a naive 30m.
+    (s,) = data["sessions"].values()
+    assert s["summary"]["uuid"] == "aaaa1111"
+    assert s["active"] == timedelta(minutes=10)
+
+
+# ── CLI ──────────────────────────────────────────────────────────────────────
+
+def test_engagement_cli_text(cli_path, fixtures_dir, tmp_path):
+    root = _fake_root(
+        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-gaps.jsonl", "aaaa1111")]}
+    )
+    r = _run_cli(
+        cli_path, "--root", str(root), "--tz", "UTC",
+        "--mode", "engagement", "--date", "2026-05-01",
+    )
+    assert r.returncode == 0
+    assert "13m" in r.stdout
+    assert "break=10m" in r.stdout
+    assert "Breaks >10m" in r.stdout
+
+
+def test_engagement_cli_json(cli_path, fixtures_dir, tmp_path):
+    root = _fake_root(
+        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-gaps.jsonl", "aaaa1111")]}
+    )
+    r = _run_cli(
+        cli_path, "--root", str(root), "--tz", "UTC",
+        "--mode", "engagement", "--date", "2026-05-01", "--format", "json",
+    )
+    assert r.returncode == 0
+    data = json.loads(r.stdout)
+    assert data["break_minutes"] == 10
+    assert data["totals"]["active_minutes"] == 13
+    assert data["totals"]["sessions"] == 1
+    s = data["sessions"][0]
+    assert s["uuid"] == "aaaa1111"
+    assert s["elapsed_minutes"] == 45
+    assert s["ratio"] == round(13 / 45, 2)
+    assert len(data["stream_breaks"]) == 1
+    assert data["stream_breaks"][0]["minutes"] == 32
+
+
+def test_engagement_cli_file_window_derived(cli_path, fixtures_dir, tmp_path):
+    root = _fake_root(
+        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-waiting.jsonl", "bbbb2222")]}
+    )
+    f = root / "C--proj-a" / "bbbb2222.jsonl"
+    r = _run_cli(
+        cli_path, "--root", str(root), "--tz", "UTC",
+        "--mode", "engagement", "--file", str(f),
+    )
+    assert r.returncode == 0
+    assert "32m" in r.stdout
+    assert "Prompt gaps" in r.stdout
+
+
+def test_engagement_cli_custom_break(cli_path, fixtures_dir, tmp_path):
+    root = _fake_root(
+        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-gaps.jsonl", "aaaa1111")]}
+    )
+    r = _run_cli(
+        cli_path, "--root", str(root), "--tz", "UTC",
+        "--mode", "engagement", "--date", "2026-05-01",
+        "--break", "1h", "--format", "json",
+    )
+    assert r.returncode == 0
+    data = json.loads(r.stdout)
+    # 1h threshold swallows the 32m gap — everything is active: 45m.
+    assert data["totals"]["active_minutes"] == 45
+    assert data["stream_breaks"] == []
+
+
+def test_engagement_cli_invalid_break(cli_path, tmp_path):
+    root = tmp_path / "projects"
+    root.mkdir()
+    r = _run_cli(
+        cli_path, "--root", str(root),
+        "--mode", "engagement", "--break", "soon",
+    )
+    assert r.returncode == 1
+
+
+def test_engagement_cli_empty_range(cli_path, fixtures_dir, tmp_path):
+    root = _fake_root(
+        tmp_path, **{"C--proj-a": [(fixtures_dir / "engagement-gaps.jsonl", "aaaa1111")]}
+    )
+    r = _run_cli(
+        cli_path, "--root", str(root), "--tz", "UTC",
+        "--mode", "engagement", "--date", "2020-01-01",
+    )
+    assert r.returncode == 0
+    assert "no user messages" in r.stdout