perf(analysis): compare_responses — снимок и сравнение ответов тяжёлых GET

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

perf/analysis/compare_responses.py

@@ -0,0 +1,169 @@
+"""Сравнение ответов тяжёлых GET до и после задач A и B (AURORA-1848).
+
+Порядок использования (D7.1):
+  1. `dump` на базовом коммите (до A):      python analysis/compare_responses.py dump --scale 1 --out /path/base
+  2. `dump` на коммите A:                  ... dump --scale 1 --out /path/a
+  3. `compare /path/base /path/a --mode exact`        - A не меняет ответы ни на байт значения
+  4. `dump` на коммите B:                  ... dump --scale 1 --out /path/b
+  5. `compare /path/base /path/b --mode strip-nulls`  - B: листы и своды = «до» с вырезанными null в data
+     (семантика jsonb_strip_nulls PG 15), остальные ответы сравниваются точно.
+Снимок: <out>/<имя>.json.gz (распакованное тело) и <out>/index.json ({имя: статус}). Код возврата compare: 1 при расхождении.
+Стенд запускает человек; тесты работают на tmp_path и MockTransport.
+"""
+from __future__ import annotations
+
+import argparse
+import gzip
+import json
+import sys
+from pathlib import Path
+
+import httpx
+
+PERF = Path(__file__).resolve().parents[1]
+
+
+def build_requests(man: dict) -> list[tuple[str, str, str]]:
+    """Список (роль, имя, путь с query) по манифесту."""
+    out: list[tuple[str, str, str]] = []
+    sheets = man.get("sheets", [])
+    for s in sheets:
+        role = "rf" if s.get("owner", "rf") == "rf" else "dfip"
+        d = s.get("direction")
+        path = f"/api/v1/form/{s['form_id']}/sheet/{s['sheet']}" + (f"?direction={d}" if d is not None else "")
+        name = f"sheet-{s['form_type']}-{s['sheet']}-{d or 'none'}-{s['form_id']}"
+        out.append((role, name, path))
+        if s["form_type"] == "FORM_2" and s["sheet"] == "AHR" and role == "rf":
+            sep = "&" if "?" in path else "?"
+            out.append((role, f"{name}-sections", f"{path}{sep}sections=plan,q1"))
+    f2 = next((s for s in sheets if s["form_type"] == "FORM_2" and s["sheet"] == "AHR" and s.get("owner", "rf") == "rf"), None)
+    year = (f2 or sheets[0])["year"] if sheets else None
+    if year is not None:
+        for k in ("FORM_1", "FORM_2", "FORM_3", "FORM_4", "MAIN"):
+            out.append(("dfip", f"svod1-{k}", f"/api/v1/svod/1/{k}/{year}"))
+        for k in ("GO", "RF"):
+            out.append(("dfip", f"svod2-{k}", f"/api/v1/svod/2/{k}/{year}"))
+        for kind in ("plan", "fact", "detail"):
+            out.append(("dfip", f"projects-{kind}-summary", f"/api/v1/projects/{kind}-summary?year={year}"))
+    for r in man.get("project_reports", []):
+        out.append(("rf", f"report-{r['project_id']}-{r['year']}-{r['report_type']}",
+                    f"/api/v1/projects/{r['project_id']}/report/{r['year']}/{r['report_type']}"))
+    for pid in dict.fromkeys(r["project_id"] for r in man.get("project_reports", [])):
+        out.append(("rf", f"project-summary-{pid}", f"/api/v1/projects/{pid}/summary"))
+    return out
+
+
+def strip_nulls(value):
+    """Как jsonb_strip_nulls: убирает None-поля объектов рекурсивно; None в списках и {} остаются."""
+    if isinstance(value, dict):
+        return {k: strip_nulls(v) for k, v in value.items() if v is not None}
+    if isinstance(value, list):
+        return [strip_nulls(v) for v in value]
+    return value
+
+
+def strip_response(body: dict) -> dict:
+    """Копия тела, где strip_nulls применён только к row['data'] строк result."""
+    out = dict(body)
+    out["result"] = [{**row, "data": strip_nulls(row["data"])} if isinstance(row, dict) and "data" in row else row
+                     for row in body.get("result") or []]
+    return out
+
+
+def is_stripped(name: str) -> bool:
+    return name.startswith(("sheet-", "svod1-", "svod2-"))
+
+
+def first_diff(a, b, path: str = "$") -> str | None:
+    """Путь до первого отличия с обоими значениями; None при равенстве."""
+    if isinstance(a, dict) and isinstance(b, dict):
+        if set(a) != set(b):
+            return f"{path}: ключи {sorted(set(a) ^ set(b))} есть только в одном из ответов"
+        for k in a:
+            d = first_diff(a[k], b[k], f"{path}.{k}")
+            if d:
+                return d
+        return None
+    if isinstance(a, list) and isinstance(b, list):
+        if len(a) != len(b):
+            return f"{path}: длина {len(a)} != {len(b)}"
+        for i, (x, y) in enumerate(zip(a, b)):
+            d = first_diff(x, y, f"{path}[{i}]")
+            if d:
+                return d
+        return None
+    if type(a) is type(b) and a == b:
+        return None
+    return f"{path}: {a!r} != {b!r}"
+
+
+def _load(d: Path, name: str):
+    return json.loads(gzip.decompress((d / f"{name}.json.gz").read_bytes()))
+
+
+def compare(before: Path, after: Path, mode: str) -> tuple[int, list[str]]:
+    bi = json.loads((before / "index.json").read_text())
+    ai_p = after / "index.json"
+    ai = json.loads(ai_p.read_text()) if ai_p.exists() else {}
+    ok, bad = 0, []
+    for name, st in bi.items():
+        if name not in ai or not (after / f"{name}.json.gz").exists():
+            bad.append(f"{name}: нет в after")
+            continue
+        if st != ai[name]:
+            bad.append(f"{name}: статус {st} != {ai[name]}")
+            continue
+        b, a = _load(before, name), _load(after, name)
+        if mode == "strip-nulls" and is_stripped(name):
+            b = strip_response(b)
+        d = first_diff(b, a)
+        if d:
+            bad.append(f"{name}: {d}")
+        else:
+            ok += 1
+    return ok, bad
+
+
+def dump(man: dict, api: str, out: Path) -> None:
+    out.mkdir(parents=True, exist_ok=True)
+    index, tokens = {}, {}
+    reqs = build_requests(man)
+    with httpx.Client(base_url=api, timeout=300) as c:
+        for role in sorted({r for r, _, _ in reqs}):
+            u = man["users"][role]
+            r = c.post("/api/v1/auth/login", json={"username": u["username"], "password": u["password"]})
+            r.raise_for_status()
+            tokens[role] = r.json()["access_token"]
+        for role, name, path in reqs:
+            r = c.get(path, headers={"Authorization": f"Bearer {tokens[role]}"})
+            (out / f"{name}.json.gz").write_bytes(gzip.compress(r.content))
+            index[name] = r.status_code
+            print(name, r.status_code, len(r.content), flush=True)
+    (out / "index.json").write_text(json.dumps(index, ensure_ascii=False, indent=1))
+
+
+def main(argv=None) -> int:
+    ap = argparse.ArgumentParser()
+    sub = ap.add_subparsers(dest="cmd", required=True)
+    d = sub.add_parser("dump")
+    d.add_argument("--scale", type=int, default=1)
+    d.add_argument("--out", required=True)
+    d.add_argument("--api", default="http://localhost:8000")
+    c = sub.add_parser("compare")
+    c.add_argument("before")
+    c.add_argument("after")
+    c.add_argument("--mode", choices=("exact", "strip-nulls"), default="exact")
+    a = ap.parse_args(argv)
+    if a.cmd == "dump":
+        man = json.loads((PERF / "results" / f"manifest-s{a.scale}.json").read_text())
+        dump(man, a.api, Path(a.out))
+        return 0
+    ok, bad = compare(Path(a.before), Path(a.after), a.mode)
+    print(f"совпало {ok} из {ok + len(bad)}")
+    for line in bad:
+        print(line)
+    return 1 if bad else 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

perf/tests/test_compare_responses.py

@@ -0,0 +1,94 @@
+"""Самопроверка compare_responses (без стенда)."""
+import gzip
+import json
+import sys
+from pathlib import Path
+
+sys.path.insert(0, str(Path(__file__).resolve().parents[1] / "analysis"))
+
+import compare_responses as cr  # noqa: E402
+
+
+def test_strip_nulls_pg_semantics():
+    v = {"a": None, "b": {"c": None}, "d": [None, {"e": None, "f": 1}], "g": {}}
+    assert cr.strip_nulls(v) == {"b": {}, "d": [None, {"f": 1}], "g": {}}
+
+
+def test_strip_response_keeps_wrapper_and_sort_order():
+    body = {"success": True, "message": None, "count": 1,
+            "result": [{"row_type": "INPUT", "depth": 3, "sort_order": None, "data": {"x": None, "y": 1}}]}
+    out = cr.strip_response(body)
+    assert out["message"] is None
+    assert out["result"][0]["sort_order"] is None
+    assert out["result"][0]["data"] == {"y": 1}
+    assert body["result"][0]["data"] == {"x": None, "y": 1}  # исходное не изменено
+
+
+def test_first_diff_path():
+    a = {"result": [{"data": {"q1": {"m1": 1}}}]}
+    b = {"result": [{"data": {"q1": {"m1": 2}}}]}
+    assert cr.first_diff(a, b).startswith("$.result[0].data.q1.m1")
+    assert cr.first_diff(a, a) is None
+    assert cr.first_diff({"a": [1]}, {"a": [1, 2]}).startswith("$.a")
+    assert cr.first_diff({"a": 1}, {"a": 1, "b": 2}).startswith("$")
+
+
+def _write(d: Path, name: str, body: dict, status: int = 200):
+    d.mkdir(parents=True, exist_ok=True)
+    (d / f"{name}.json.gz").write_bytes(gzip.compress(json.dumps(body).encode()))
+    idx = d / "index.json"
+    cur = json.loads(idx.read_text()) if idx.exists() else {}
+    cur[name] = status
+    idx.write_text(json.dumps(cur))
+
+
+def _body(data):
+    return {"success": True, "message": None, "count": 1,
+            "result": [{"row_type": "INPUT", "depth": 1, "sort_order": 1, "data": data}]}
+
+
+def test_compare_exact_and_strip_modes(tmp_path):
+    b, a = tmp_path / "before", tmp_path / "after"
+    for name in ("sheet-FORM_2-AHR-none-1", "report-5-2026-LIMIT"):
+        _write(b, name, _body({"x": None, "y": 1}))
+        _write(a, name, _body({"y": 1}))
+    n, bad = cr.compare(b, a, "exact")
+    assert n == 0 and len(bad) == 2
+    n, bad = cr.compare(b, a, "strip-nulls")
+    assert n == 1 and len(bad) == 1 and "report-5-2026-LIMIT" in bad[0]
+
+
+def test_compare_status_mismatch(tmp_path):
+    b, a = tmp_path / "before", tmp_path / "after"
+    _write(b, "svod1-MAIN", _body({"y": 1}), 200)
+    _write(a, "svod1-MAIN", _body({"y": 1}), 500)
+    n, bad = cr.compare(b, a, "exact")
+    assert n == 0 and len(bad) == 1
+
+
+def test_compare_missing_after(tmp_path):
+    b, a = tmp_path / "before", tmp_path / "after"
+    _write(b, "svod1-MAIN", _body({}))
+    a.mkdir()
+    n, bad = cr.compare(b, a, "exact")
+    assert n == 0 and len(bad) == 1
+
+
+def test_build_requests_minimal_manifest():
+    man = {"sheets": [
+        {"form_id": 1, "form_type": "FORM_2", "sheet": "AHR", "direction": None, "year": 2026, "url": "", "rows": 1, "owner": "rf"},
+        {"form_id": 2, "form_type": "FORM_1", "sheet": "AHR", "direction": "Support", "year": 2026, "url": "", "rows": 1, "owner": "ssp"},
+    ], "project_reports": [{"project_id": 5, "year": 2026, "report_type": "LIMIT", "url": "", "rows": 1}]}
+    reqs = cr.build_requests(man)
+    by = {n: (r, p) for r, n, p in reqs}
+    assert by["sheet-FORM_2-AHR-none-1"] == ("rf", "/api/v1/form/1/sheet/AHR")
+    assert by["sheet-FORM_2-AHR-none-1-sections"] == ("rf", "/api/v1/form/1/sheet/AHR?sections=plan,q1")
+    assert by["sheet-FORM_1-AHR-Support-2"] == ("dfip", "/api/v1/form/2/sheet/AHR?direction=Support")
+    assert sum(n.startswith("svod1-") for n in by) == 5
+    assert sum(n.startswith("svod2-") for n in by) == 2
+    assert by["svod1-MAIN"] == ("dfip", "/api/v1/svod/1/MAIN/2026")
+    assert by["report-5-2026-LIMIT"][0] == "rf"
+    assert by["project-summary-5"] == ("rf", "/api/v1/projects/5/summary")
+    assert {"projects-plan-summary", "projects-fact-summary", "projects-detail-summary"} <= set(by)
+    assert not any("rollup" in n for n in by)
+    assert cr.is_stripped("sheet-x") and cr.is_stripped("svod2-GO") and not cr.is_stripped("report-1")