refactor mobile.de parser, fix country mapping, update README

This commit is contained in:
qananasikq
2026-05-04 18:12:10 +03:00
parent edddfc884c
commit 800867b4e6
46 changed files with 2887 additions and 10900 deletions

View File

@@ -21,6 +21,17 @@ CAR_DB_FIELDS = {
_IN_CHUNK_SIZE = 5000
CAR_TABLE_NAME = Car.__tablename__
MOBILEDE_ORIGIN_PREFIXES = ("mobile.de:", "mobilede:")
def _origin_prefix_filter(column, prefixes: tuple[str, ...] = MOBILEDE_ORIGIN_PREFIXES):
"""Match all supported mobile.de origin_id prefixes.
Older records were stored as ``mobile.de:<id>`` while some newer helper code
used ``mobilede:<id>``. Cleanup and refresh queries must include both to avoid
leaving stale active cars in the DB.
"""
return or_(*[column.like(f"{prefix}%") for prefix in prefixes])
class PersistenceService:
@@ -502,7 +513,7 @@ class PersistenceService:
update(Car)
.where(Car.origin_id.notin_(active_origin_ids))
.where(Car.is_sold == False) # noqa: E712
.where(Car.origin_id.like("mobilede:%"))
.where(_origin_prefix_filter(Car.origin_id))
.values(is_sold=True)
)
result = session.execute(stmt)
@@ -533,7 +544,12 @@ class PersistenceService:
if is_postgres:
# Считаем кандидатов на sold.
total_active = session.execute(
text(f"SELECT count(*) FROM {CAR_TABLE_NAME} WHERE is_sold = FALSE AND origin_id LIKE 'mobilede:%%'")
text(f"""
SELECT count(*)
FROM {CAR_TABLE_NAME}
WHERE is_sold = FALSE
AND (origin_id LIKE 'mobile.de:%%' OR origin_id LIKE 'mobilede:%%')
""")
).scalar() or 0
if total_active == 0:
@@ -561,7 +577,7 @@ class PersistenceService:
LEFT JOIN _active_urls a ON replace(c.origin_url, '~', '-') = a.url
WHERE a.url IS NULL
AND c.is_sold = FALSE
AND c.origin_id LIKE 'mobilede:%%'
AND (c.origin_id LIKE 'mobile.de:%%' OR c.origin_id LIKE 'mobilede:%%')
""".format(car_table=CAR_TABLE_NAME))).scalar() or 0
# Защита от аномалии.
@@ -582,7 +598,7 @@ class PersistenceService:
LEFT JOIN _active_urls a ON replace(c.origin_url, '~', '-') = a.url
WHERE a.url IS NULL
AND c.is_sold = FALSE
AND c.origin_id LIKE 'mobilede:%%'
AND (c.origin_id LIKE 'mobile.de:%%' OR c.origin_id LIKE 'mobilede:%%')
) sub
WHERE {car_table}.id = sub.id
""".format(car_table=CAR_TABLE_NAME)))
@@ -592,13 +608,13 @@ class PersistenceService:
stmt = (
update(Car)
.where(Car.is_sold == False) # noqa: E712
.where(Car.origin_id.like("mobilede:%"))
.where(_origin_prefix_filter(Car.origin_id))
.values(is_sold=True)
)
# Загружаем active URL.
all_active = session.execute(
select(Car.id, Car.origin_url).where(
Car.is_sold == False, Car.origin_id.like("mobilede:%") # noqa: E712
Car.is_sold == False, _origin_prefix_filter(Car.origin_id) # noqa: E712
)
).all()
mark_ids = [row[0] for row in all_active if _norm(row[1]) not in normalized_urls]
@@ -622,35 +638,99 @@ class PersistenceService:
logger.info("Marked %d cars as sold by URL (no longer in listing)", count)
return count
def get_all_origin_ids_for_lane(self, prefix: str = "mobilede:") -> set[str]:
def mark_sold_not_seen_since(
self,
since_ts: datetime,
*,
prefix: str | tuple[str, ...] = MOBILEDE_ORIGIN_PREFIXES,
safety_ratio: float = 0.8,
) -> int:
"""Mark active cars as sold when they were not seen during a full refresh cycle.
Any unsold mobile.de car with ``last_seen_at < since_ts`` is considered absent
from the latest completed refresh cycle and can be marked as sold.
A safety guard prevents anomalous bulk updates.
"""
prefixes = (prefix,) if isinstance(prefix, str) else tuple(prefix)
with self.session_scope() as session:
total_active = int(
session.execute(
select(text("count(*)")).select_from(Car).where(
_origin_prefix_filter(Car.origin_id, prefixes),
Car.is_sold == False, # noqa: E712
)
).scalar()
or 0
)
if total_active <= 0:
return 0
would_mark = int(
session.execute(
select(text("count(*)")).select_from(Car).where(
_origin_prefix_filter(Car.origin_id, prefixes),
Car.is_sold == False, # noqa: E712
Car.last_seen_at < since_ts,
)
).scalar()
or 0
)
if would_mark <= 0:
return 0
if total_active > 100 and would_mark > int(total_active * max(0.0, min(1.0, safety_ratio))):
logger.error(
"mark_sold(last_seen) safety abort: would mark %d/%d (%.0f%%) as sold",
would_mark,
total_active,
(would_mark / max(1, total_active)) * 100,
)
return 0
result = session.execute(
update(Car)
.where(_origin_prefix_filter(Car.origin_id, prefixes))
.where(Car.is_sold == False) # noqa: E712
.where(Car.last_seen_at < since_ts)
.values(is_sold=True)
)
count = int(result.rowcount or 0)
if count:
logger.info("Marked %d cars as sold by last_seen cutoff=%s", count, since_ts.isoformat())
return count
def get_all_origin_ids_for_lane(self, prefix: str | tuple[str, ...] = MOBILEDE_ORIGIN_PREFIXES) -> set[str]:
"""Возвращает все известные origin_id для заданного префикса.
Использует yield_per для потоковой загрузки при большом количестве записей.
"""
prefixes = (prefix,) if isinstance(prefix, str) else tuple(prefix)
with self.session_scope() as session:
result = session.execute(
select(Car.origin_id).where(Car.origin_id.like(f"{prefix}%")).execution_options(yield_per=10000)
select(Car.origin_id).where(_origin_prefix_filter(Car.origin_id, prefixes)).execution_options(yield_per=10000)
)
return {str(row[0]) for row in result if row and row[0]}
def get_all_active_origin_urls_for_lane(self, prefix: str = "mobilede:") -> set[str]:
def get_all_active_origin_urls_for_lane(self, prefix: str | tuple[str, ...] = MOBILEDE_ORIGIN_PREFIXES) -> set[str]:
"""Возвращает origin_url всех активных (не проданных) авто для заданного lane-префикса."""
prefixes = (prefix,) if isinstance(prefix, str) else tuple(prefix)
with self.session_scope() as session:
result = session.execute(
select(Car.origin_url).where(
Car.origin_id.like(f"{prefix}%"),
_origin_prefix_filter(Car.origin_id, prefixes),
Car.is_sold == False, # noqa: E712
).execution_options(yield_per=10000)
)
return {str(row[0]) for row in result if row and row[0]}
def count_active_cars_for_lane(self, prefix: str = "mobilede:") -> int:
def count_active_cars_for_lane(self, prefix: str | tuple[str, ...] = MOBILEDE_ORIGIN_PREFIXES) -> int:
"""Возвращает количество активных (не проданных) авто для заданного lane-префикса."""
from sqlalchemy import func as sa_func
prefixes = (prefix,) if isinstance(prefix, str) else tuple(prefix)
with self.session_scope() as session:
result = session.execute(
select(sa_func.count()).select_from(Car).where(
Car.origin_id.like(f"{prefix}%"),
_origin_prefix_filter(Car.origin_id, prefixes),
Car.is_sold == False, # noqa: E712
)
)
@@ -658,7 +738,7 @@ class PersistenceService:
def get_active_origin_urls_batch_for_refresh(
self,
prefix: str = "mobilede:",
prefix: str | tuple[str, ...] = MOBILEDE_ORIGIN_PREFIXES,
offset: int = 0,
limit: int = 500,
) -> list[str]:
@@ -666,10 +746,11 @@ class PersistenceService:
Сортировка по last_seen_at ASC — давно не обновлённые идут первыми.
"""
prefixes = (prefix,) if isinstance(prefix, str) else tuple(prefix)
with self.session_scope() as session:
result = session.execute(
select(Car.origin_url).where(
Car.origin_id.like(f"{prefix}%"),
_origin_prefix_filter(Car.origin_id, prefixes),
Car.is_sold == False, # noqa: E712
).order_by(Car.last_seen_at.asc()).offset(offset).limit(limit)
)