🌊 Волна 17: backend #8 — баги + N+1 (16 багов)

Продуктовые:
- hint_consumed_before_word_check: сначала проверка слов, потом инкремент (hints.py)
- hint_fake_first_cell: 404 вместо [0,0] (hints.py)
- leaderboard_rank_wrong_for_uncached_user: корректный rank (leaderboard.py)
- teacher_ai_explanation_empty_on_failure: fallback-текст (teacher.py)

Данные:
- duplicate_delete_keeps_wrong_row: оставляем лучшую строку (database.py)

Security:
- fastapi_docs_exposed: /docs скрыт в проде (main.py)
- cf_ip_check_per_request: предкомпиляция CF_IPS (main.py)

N+1:
- admin_content_lesson_levels_n_plus_1: 3 GROUP BY (admin_content.py)
- admin_content_lesson_topics_n_plus_1: 2 GROUP BY (admin_content.py)
- admin_content_get_lesson_n_plus_1: ANY + группировка (admin_content.py)
- teacher_progress_n_plus_1: 2 GROUP BY (teacher.py)
- teacher_diagnostic_test_n_plus_1: ANY + группировка (teacher.py)
- orchestrator_n_plus_1_options: ANY + группировка (orchestrator.py)

Качество:
- cleanup_task_leak_on_startup_failure: cleanup после init (main.py)
- unused_first_name: удалена мёртвая переменная (auth.py)

Пропущено осознанно:
- initdata_in_query_param (fallback deleteFeedPost)
- init_db_slow_startup (Alembic, отдельная задача)

Аудит v5: Issue #8, backend 16/17 задач
This commit is contained in:
2026-09-24 17:35:14 +00:00
parent f655a44a84
commit 3d75a00046
8 changed files with 272 additions and 131 deletions
+1 -1
View File
@@ -73,7 +73,7 @@ async def verify_telegram_initdata(init_data: str):
user_id = user_data.get('id')
username = user_data.get('username', '')
first_name = user_data.get('first_name', '')
# 🔧 ФИКС (25.09.2026): unused_first_name — удалена мёртвая переменная.
if not user_id:
raise HTTPException(401, "User ID not found in initData")
+11 -8
View File
@@ -131,15 +131,18 @@ async def init_db():
ALTER COLUMN user_id TYPE BIGINT
""")
# 🔧 ФИКС (24.09.2026): add_pk_on_duplicates —
# удаляем дубли перед ADD PRIMARY KEY.
# Оставляем строку с максимальным completed_at (или первую).
# 🔧 ФИКС (25.09.2026): duplicate_delete_keeps_wrong_row —
# оставляем строку с максимальным completed_at, затем score.
# Удаляем все остальные дубли по (user_id, level_id).
await conn.execute("""
DELETE FROM user_progress a
USING user_progress b
WHERE a.ctid < b.ctid
AND a.user_id = b.user_id
AND a.level_id = b.level_id
DELETE FROM user_progress
WHERE ctid NOT IN (
SELECT DISTINCT ON (user_id, level_id) ctid
FROM user_progress
ORDER BY user_id, level_id,
completed_at DESC NULLS LAST,
score DESC
)
""")
await conn.execute("""
+30 -15
View File
@@ -126,23 +126,25 @@ class RateLimiter:
rate_limiter = RateLimiter()
# 🔧 ФИКС (24.09.2026): cf_connecting_ip_spoof —
# проверяем, что IP клиента входит в список Cloudflare.
# 🔧 ФИКС (25.09.2026): cf_ip_check_per_request —
# компилируем CF_IPS один раз при старте, а не парсим 22 CIDR на каждый запрос.
_CF_NETWORKS = []
for _cidr in CF_IPS:
try:
_CF_NETWORKS.append(ipaddress.ip_network(_cidr, strict=False))
except ValueError:
logger.warning(f"Invalid CF_IPS entry skipped: {_cidr}")
def _is_cloudflare_ip(ip_str: str) -> bool:
"""Проверяет, что IP входит в CF_IPS (Cloudflare)."""
"""Проверяет, что IP входит в предкомпилированные сети Cloudflare."""
if not ip_str:
return False
try:
ip = ipaddress.ip_address(ip_str)
except ValueError:
return False
for cidr in CF_IPS:
try:
if ip in ipaddress.ip_network(cidr, strict=False):
return True
except ValueError:
continue
return False
return any(ip in net for net in _CF_NETWORKS)
# ============ Фоновое обновление кэша рейтинга ============
@@ -171,10 +173,9 @@ async def lifespan(app: FastAPI):
await init_achievements()
await init_learning_data()
# 🔧 ФИКС (24.09.2026): cleanup_task_not_cancelled —
# сохраняем ссылку и отменяем в shutdown.
cleanup_task = asyncio.create_task(rate_limiter.start_cleanup())
# 🔧 ФИКС (25.09.2026): cleanup_task_leak_on_startup_failure —
# создаём cleanup_task ПОСЛЕ успешной инициализации БД и matview.
# Если что-то упадёт до этого — задача не создастся, утечки не будет.
from database import get_pool
pool = await get_pool()
async with pool.acquire() as conn:
@@ -227,6 +228,9 @@ async def lifespan(app: FastAPI):
logger.info("Leaderboard cache created")
refresh_task = asyncio.create_task(refresh_leaderboard_cache_loop())
# 🔧 ФИКС (25.09.2026): cleanup_task_leak_on_startup_failure —
# создаём cleanup_task только после успешной инициализации.
cleanup_task = asyncio.create_task(rate_limiter.start_cleanup())
logger.info(f"Filwords API v{APP_VERSION} started")
@@ -251,7 +255,18 @@ async def lifespan(app: FastAPI):
logger.info(f"Filwords API v{APP_VERSION} stopped")
app = FastAPI(title="Filwords PL API", lifespan=lifespan)
# 🔧 ФИКС (25.09.2026): fastapi_docs_exposed —
# в проде скрываем /docs, /redoc, /openapi.json.
_ENV = os.getenv("ENV", "production")
_is_prod = _ENV == "production"
app = FastAPI(
title="Filwords PL API",
lifespan=lifespan,
docs_url=None if _is_prod else "/docs",
redoc_url=None if _is_prod else "/redoc",
openapi_url=None if _is_prod else "/openapi.json",
)
# ============ Rate Limiting Middleware ============
+84 -45
View File
@@ -647,37 +647,53 @@ async def get_lesson_levels(request: Request):
ORDER BY sort_order
""")
if not level_rows:
return {"levels": []}
level_codes = [lvl["code"] for lvl in level_rows]
# 🔧 ФИКС (25.09.2026): admin_content_lesson_levels_n_plus_1 —
# 3 GROUP BY запроса вместо 3×N.
topics_rows = await conn.fetch("""
SELECT level_code, COUNT(*) as cnt
FROM learning_topics
WHERE level_code = ANY($1::text[]) AND is_active = TRUE
GROUP BY level_code
""", level_codes)
topics_map = {r["level_code"]: r["cnt"] for r in topics_rows}
lessons_rows = await conn.fetch("""
SELECT t.level_code, COUNT(*) as cnt
FROM learning_lessons l
JOIN learning_topics t ON l.topic_id = t.topic_id
WHERE t.level_code = ANY($1::text[]) AND l.is_active = TRUE
GROUP BY t.level_code
""", level_codes)
lessons_map = {r["level_code"]: r["cnt"] for r in lessons_rows}
last_edited_rows = await conn.fetch("""
SELECT level_code, MAX(edited_at) as last_edited
FROM (
SELECT t.level_code, t.edited_at FROM learning_topics t
WHERE t.level_code = ANY($1::text[])
UNION ALL
SELECT t.level_code, l.edited_at FROM learning_lessons l
JOIN learning_topics t ON l.topic_id = t.topic_id
WHERE t.level_code = ANY($1::text[])
) sub
GROUP BY level_code
""", level_codes)
last_edited_map = {r["level_code"]: r["last_edited"] for r in last_edited_rows}
result = []
for lvl in level_rows:
code = lvl["code"]
topics_count = await conn.fetchval("""
SELECT COUNT(*) FROM learning_topics
WHERE level_code = $1 AND is_active = TRUE
""", code) or 0
lessons_count = await conn.fetchval("""
SELECT COUNT(*) FROM learning_lessons l
JOIN learning_topics t ON l.topic_id = t.topic_id
WHERE t.level_code = $1 AND l.is_active = TRUE
""", code) or 0
last_edited = await conn.fetchval("""
SELECT MAX(edited_at) FROM (
SELECT t.edited_at FROM learning_topics t
WHERE t.level_code = $1
UNION ALL
SELECT l.edited_at FROM learning_lessons l
JOIN learning_topics t ON l.topic_id = t.topic_id
WHERE t.level_code = $1
) sub
""", code)
last_edited = last_edited_map.get(code)
result.append({
"code": code,
"name_ru": lvl["name_ru"],
"topics_count": topics_count,
"lessons_count": lessons_count,
"topics_count": topics_map.get(code, 0),
"lessons_count": lessons_map.get(code, 0),
"last_edited": last_edited.isoformat() if last_edited else None,
})
@@ -709,20 +725,32 @@ async def get_lesson_topics(request: Request, level: str):
ORDER BY order_number ASC
""", level)
if not rows:
return {"level": level, "topics": []}
topic_ids = [r["topic_id"] for r in rows]
# 🔧 ФИКС (25.09.2026): admin_content_lesson_topics_n_plus_1 —
# 2 GROUP BY запроса вместо 2×N.
lessons_rows = await conn.fetch("""
SELECT topic_id, COUNT(*) as cnt
FROM learning_lessons
WHERE topic_id = ANY($1::text[]) AND is_active = TRUE
GROUP BY topic_id
""", topic_ids)
lessons_map = {r["topic_id"]: r["cnt"] for r in lessons_rows}
test_rows = await conn.fetch("""
SELECT topic_id, COUNT(*) as cnt
FROM learning_questions
WHERE topic_id = ANY($1::text[]) AND is_test_question = TRUE
GROUP BY topic_id
""", topic_ids)
test_map = {r["topic_id"]: r["cnt"] for r in test_rows}
result = []
for r in rows:
topic_id = r["topic_id"]
lessons_count = await conn.fetchval("""
SELECT COUNT(*) FROM learning_lessons
WHERE topic_id = $1 AND is_active = TRUE
""", topic_id) or 0
test_questions_count = await conn.fetchval("""
SELECT COUNT(*) FROM learning_questions
WHERE topic_id = $1 AND is_test_question = TRUE
""", topic_id) or 0
result.append({
"topic_id": topic_id,
"order_number": r["order_number"],
@@ -730,8 +758,8 @@ async def get_lesson_topics(request: Request, level: str):
"name_ru": r["name_ru"],
"name_uk": r["name_uk"],
"name_en": r["name_en"],
"lessons_count": lessons_count,
"test_questions_count": test_questions_count,
"lessons_count": lessons_map.get(topic_id, 0),
"test_questions_count": test_map.get(topic_id, 0),
})
return {"level": level, "topics": result}
@@ -876,15 +904,26 @@ async def get_lesson(request: Request, lesson_id: str):
""")
wb_map = {w["wrong_because"]: w["explanation_ru"] for w in wb_rows}
questions = []
for q in questions_rows:
options_rows = await conn.fetch("""
SELECT option_key, text_pl, text_ru, text_uk, text_en,
# 🔧 ФИКС (25.09.2026): admin_content_get_lesson_n_plus_1 —
# один запрос ANY + группировка.
question_ids = [q["question_id"] for q in questions_rows]
all_options = []
if question_ids:
all_options = await conn.fetch("""
SELECT question_id, option_key, text_pl, text_ru, text_uk, text_en,
is_correct, wrong_because
FROM question_options
WHERE question_id = $1
ORDER BY option_key ASC
""", q["question_id"])
WHERE question_id = ANY($1::text[])
ORDER BY question_id, option_key ASC
""", question_ids)
options_by_q: dict = {}
for o in all_options:
options_by_q.setdefault(o["question_id"], []).append(o)
questions = []
for q in questions_rows:
options_rows = options_by_q.get(q["question_id"], [])
options = []
for o in options_rows:
+15 -11
View File
@@ -38,6 +38,17 @@ async def get_first_letter_hint(request: Request, level_id: int):
if hints_already_used >= MAX_HINTS:
raise HTTPException(403, "Hint limit reached")
# 🔧 ФИКС (25.09.2026): hint_consumed_before_word_check —
# сначала проверяем, что у уровня ЕСТЬ слова, и только потом
# инкрементим hints_used. Раньше при 404 hints_used уже +1.
words = await conn.fetch(
"SELECT word_pl FROM level_words WHERE level_id = $1", level_id
)
if not words:
raise HTTPException(404, "No words found for this level")
# Теперь безопасно инкрементим счётчик
if progress:
await conn.execute(
"UPDATE user_progress SET hints_used = hints_used + 1 WHERE user_id = $1 AND level_id = $2",
@@ -50,14 +61,6 @@ async def get_first_letter_hint(request: Request, level_id: int):
)
hints_already_used += 1
# Получаем все слова уровня
words = await conn.fetch(
"SELECT word_pl FROM level_words WHERE level_id = $1", level_id
)
if not words:
raise HTTPException(404, "No words found for this level")
word_paths = level["word_paths"]
if isinstance(word_paths, str):
import json
@@ -76,10 +79,11 @@ async def get_first_letter_hint(request: Request, level_id: int):
first_cell = word_paths[word][0]
break
# 🔧 ФИКС (25.09.2026): hint_fake_first_cell —
# если ни у одного слова нет пути в word_paths, подсказка
# бессмысленна. Возвращаем 404, а не фейковую координату [0,0].
if not found_word:
# Если пути не нашлись — берём любое слово
found_word = all_level_words[0]
first_cell = [0, 0]
raise HTTPException(404, "No word paths found for this level")
return {
"status": "ok",
+18 -5
View File
@@ -52,13 +52,26 @@ async def get_leaderboard(request: Request, limit: int = 50):
"sentences_completed": 0,
})
# Место пользователя
# 🔧 ФИКС (25.09.2026): leaderboard_rank_wrong_for_uncached_user —
# если пользователя нет в leaderboard_cache, подзапрос вернёт NULL,
# и COUNT(*) + 1 = 1 (юзер видит себя первым, хотя у него 0 очков).
# Теперь: если user нет в кэше — считаем его место по количеству
# записей с total_score > 0 + 1 (в конец списка).
my_total_in_cache = await conn.fetchval("""
SELECT total_score FROM leaderboard_cache WHERE telegram_id = $1
""", telegram_id)
if my_total_in_cache is None:
# Пользователя нет в кэше — считаем по количеству с > 0
user_rank = await conn.fetchval("""
SELECT COUNT(*) + 1 FROM leaderboard_cache
WHERE total_score > (
SELECT COALESCE(total_score, 0) FROM leaderboard_cache WHERE telegram_id = $1
)
""", telegram_id)
WHERE total_score > 0
""") or 1
else:
user_rank = await conn.fetchval("""
SELECT COUNT(*) + 1 FROM leaderboard_cache
WHERE total_score > $1
""", my_total_in_cache)
# Персональная статистика пользователя
my_row = await conn.fetchrow("""
+41 -12
View File
@@ -170,14 +170,25 @@ async def get_lesson_by_id(conn, lesson_id: str) -> Optional[Dict[str, Any]]:
ORDER BY id
""", lesson_id)
# 🔧 ФИКС (25.09.2026): orchestrator_n_plus_1_options —
# один запрос ANY + группировка в Python.
question_ids = [q["question_id"] for q in questions]
all_options = []
if question_ids:
all_options = await conn.fetch("""
SELECT * FROM question_options
WHERE question_id = ANY($1::text[])
""", question_ids)
options_by_q: dict = {}
for o in all_options:
options_by_q.setdefault(o["question_id"], []).append(dict(o))
questions_data = []
for q in questions:
options = await conn.fetch("""
SELECT * FROM question_options WHERE question_id = $1
""", q["question_id"])
questions_data.append({
"question": dict(q),
"options": [dict(o) for o in options],
"options": options_by_q.get(q["question_id"], []),
})
return {
@@ -196,14 +207,23 @@ async def get_lesson_questions(conn, lesson_id: str) -> list:
ORDER BY id
""", lesson_id)
question_ids = [q["question_id"] for q in questions]
all_options = []
if question_ids:
all_options = await conn.fetch("""
SELECT * FROM question_options
WHERE question_id = ANY($1::text[])
""", question_ids)
options_by_q: dict = {}
for o in all_options:
options_by_q.setdefault(o["question_id"], []).append(dict(o))
questions_data = []
for q in questions:
options = await conn.fetch("""
SELECT * FROM question_options WHERE question_id = $1
""", q["question_id"])
questions_data.append({
"question": dict(q),
"options": [dict(o) for o in options],
"options": options_by_q.get(q["question_id"], []),
})
return questions_data
@@ -218,14 +238,23 @@ async def get_test_questions(conn, topic_id: str) -> list:
ORDER BY test_order
""", topic_id)
question_ids = [q["question_id"] for q in questions]
all_options = []
if question_ids:
all_options = await conn.fetch("""
SELECT * FROM question_options
WHERE question_id = ANY($1::text[])
""", question_ids)
options_by_q: dict = {}
for o in all_options:
options_by_q.setdefault(o["question_id"], []).append(dict(o))
questions_data = []
for q in questions:
options = await conn.fetch("""
SELECT * FROM question_options WHERE question_id = $1
""", q["question_id"])
questions_data.append({
"question": dict(q),
"options": [dict(o) for o in options],
"options": options_by_q.get(q["question_id"], []),
})
return questions_data
+59 -21
View File
@@ -288,7 +288,20 @@ async def get_wrong_answer_ai_explanation(telegram_id: int, lang: str, prompt_te
async with pool.acquire() as conn:
await increment_teacher_ai_usage(conn, telegram_id)
else:
ai_response = get_ai_limit_message(lang) if not await _check_ai_limit_for_fallback(telegram_id, lang) else ""
# 🔧 ФИКС (25.09.2026): teacher_ai_explanation_empty_on_failure —
# если AI упал и лимит НЕ исчерпан, возвращаем нейтральный fallback,
# а не пустую строку (иначе ученик не понимает, почему ошибся).
if not await _check_ai_limit_for_fallback(telegram_id, lang):
ai_response = get_ai_limit_message(lang)
else:
# Нейтральный fallback на 4 языках
fallbacks = {
"pl": "Zastanów się nad tym pytaniem jeszcze raz.",
"ru": "Подумай над этим вопросом ещё раз.",
"uk": "Поміркуй над цим питанням ще раз.",
"en": "Think about this question again.",
}
ai_response = fallbacks.get(lang, fallbacks["en"])
return ai_response
@@ -1029,14 +1042,26 @@ async def teacher_diagnostic_test(request: Request):
) or "A1"
session_id = await start_learning_session(conn, telegram_id, user_level, "", "diagnostic_test")
# 🔧 ФИКС (25.09.2026): teacher_diagnostic_test_n_plus_1 —
# один запрос ANY + группировка в Python.
question_ids = [q["question_id"] for q in questions]
all_options = []
if question_ids:
all_options = await conn.fetch("""
SELECT * FROM question_options
WHERE question_id = ANY($1::text[])
""", question_ids)
options_by_q: dict = {}
for o in all_options:
options_by_q.setdefault(o["question_id"], []).append(dict(o))
questions_data = []
for q in questions:
options = await conn.fetch("""
SELECT * FROM question_options WHERE question_id = $1
""", q["question_id"])
questions_data.append({
"question": dict(q),
"options": [dict(o) for o in options],
"options": options_by_q.get(q["question_id"], []),
})
return {
@@ -1104,31 +1129,44 @@ async def teacher_progress(request: Request):
ORDER BY sort_order
""")
result = []
if not levels:
return {"levels": []}
level_codes = [lvl["code"] for lvl in levels]
# 🔧 ФИКС (25.09.2026): teacher_progress_n_plus_1 —
# 2 запроса вместо 2×N.
topics_total_rows = await conn.fetch("""
SELECT level_code, COUNT(*) as cnt
FROM learning_topics
WHERE level_code = ANY($1::text[]) AND is_active = TRUE
GROUP BY level_code
""", level_codes)
topics_total_map = {r["level_code"]: r["cnt"] for r in topics_total_rows}
topics_completed_rows = await conn.fetch("""
SELECT t.level_code, COUNT(*) as cnt
FROM user_topic_progress utp
JOIN learning_topics t ON utp.topic_id = t.topic_id
WHERE utp.user_id = $1
AND utp.test_completed = TRUE
AND t.is_active = TRUE
AND t.level_code = ANY($2::text[])
GROUP BY t.level_code
""", telegram_id, level_codes)
topics_completed_map = {r["level_code"]: r["cnt"] for r in topics_completed_rows}
result = []
for level in levels:
code = level["code"]
topics_total = await conn.fetchval("""
SELECT COUNT(*) FROM learning_topics
WHERE level_code = $1 AND is_active = TRUE
""", code) or 0
topics_completed = await conn.fetchval("""
SELECT COUNT(*) FROM user_topic_progress
WHERE user_id = $1 AND topic_id IN (
SELECT topic_id FROM learning_topics WHERE level_code = $2 AND is_active = TRUE
) AND test_completed = TRUE
""", telegram_id, code) or 0
result.append({
"code": code,
"name_pl": level["name_pl"],
"name_ru": level["name_ru"],
"name_uk": level["name_uk"],
"name_en": level["name_en"],
"topics_total": topics_total,
"topics_completed": topics_completed,
"topics_total": topics_total_map.get(code, 0),
"topics_completed": topics_completed_map.get(code, 0),
})
return {"levels": result}