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

@@ -314,9 +314,7 @@ class MobileDeScraper:
max_pages,
)
try:
data = self.collect_search(
start_page=start_page,
max_pages=max_pages,
params = self._build_search_params(
search_url=search_url,
make_id=make_id,
model_id=model_id,
@@ -328,96 +326,121 @@ class MobileDeScraper:
mileage_max=mileage_max,
sort_by=sort_by,
sort_order=sort_order,
progress_callback=progress_callback,
)
pages_payload = list(data.get("pages", []))
listing_count = int(data.get("listing_count", 0) or 0)
unique_ids = set(data.get("unique_listing_ids", []))
pages_collected = len(pages_payload)
def _on_page(page, meta: dict[str, int | None]) -> None:
if progress_callback is None:
return
payload = {
**meta,
"page_url": page.url,
"unique_ids_seen": len(unique_ids) + len({listing.id for listing in page.listings if listing.id}),
}
progress_callback("page_collected", payload)
records: list[CarRecord] = []
for page in pages_payload:
page_records = [self.mapper.listing_to_car_record(MobileDeListing(**item)) for item in page.get("listings", [])]
records.extend(self._dedupe_page_records(page_records, seen_record_keys))
pages_collected = 0
early_stopped = False
for page in self.client.iter_search_pages(
start_page=start_page,
max_pages=max_pages,
search_url=search_url,
progress_callback=_on_page,
**params,
):
pages_collected += 1
pages_payload.append(asdict(page))
listing_count += len(page.listings)
unique_ids.update(str(listing.id) for listing in page.listings if listing.id)
if only_new and records:
existing_origin_ids = self.persistence.get_existing_origin_ids(
[record.origin_id for record in records if record.origin_id]
)
should_cut_tail = self._should_cut_only_new_tail(sort_by, sort_order)
if should_cut_tail:
filtered_records: list[CarRecord] = []
for record_index, record in enumerate(records):
is_existing = bool(record.origin_id and record.origin_id in existing_origin_ids)
filtered_records.append(record)
if is_existing:
existing_streak += 1
if (
existing_streak >= MOBILEDE_ONLY_NEW_STOP_ON_EXISTING_STREAK
and new_records_kept >= MOBILEDE_ONLY_NEW_MIN_NEW_RECORDS
):
head_cut_triggered = True
skipped_existing += max(0, len(records) - record_index - 1)
break
else:
existing_streak = 0
new_records_kept += 1
records = filtered_records
if progress_callback is not None:
progress_callback(
"records_mapped",
{
"record_count": len(records),
"skipped_existing": skipped_existing,
"only_new": bool(only_new),
"run_id": run_id,
"pages_collected": pages_collected,
},
page_records = [self.mapper.listing_to_car_record(listing) for listing in page.listings]
page_records = self._dedupe_page_records(page_records, seen_record_keys)
page_records, skipped_existing, existing_streak, new_records_kept, head_cut_triggered = (
self._apply_only_new_page_policy(
page_records=page_records,
only_new=only_new,
sort_by=sort_by,
sort_order=sort_order,
skipped_existing=skipped_existing,
existing_streak=existing_streak,
new_records_kept=new_records_kept,
)
)
upsert = self.persistence.upsert_cars_batch(records) if records else {
"inserted": 0,
"updated": 0,
"images_upserted": 0,
if progress_callback is not None:
progress_callback(
"records_mapped",
{
"record_count": len(page_records),
"skipped_existing": skipped_existing,
"only_new": bool(only_new),
"run_id": run_id,
"pages_collected": pages_collected,
"page_number": page.page_number,
},
)
upsert = self.persistence.upsert_cars_batch(page_records) if page_records else {
"inserted": 0,
"updated": 0,
"images_upserted": 0,
}
page_inserted = int(upsert.get("inserted", 0))
page_updated = int(upsert.get("updated", 0))
page_images = int(upsert.get("images_upserted", 0))
ids_fetched += len(page_records)
inserted_total += page_inserted
updated_total += page_updated
images_upserted += page_images
cars_upserted = inserted_total + updated_total
logger.debug(
"mobile.de sync_search page upsert: run_id=%s page=%s inserted=%s updated=%s images=%s",
run_id,
page.page_number,
page_inserted,
page_updated,
page_images,
)
if progress_callback is not None:
progress_callback(
"db_upsert_done",
{
"run_id": run_id,
"pages_collected": pages_collected,
"pages_in_batch": 1,
"page_number": page.page_number,
"inserted": page_inserted,
"updated": page_updated,
"images_upserted": page_images,
},
)
if head_cut_triggered:
early_stopped = True
break
data = {
"source": "mobile.de",
"strategy_note": MOBILEDE_SEARCH_STRATEGY_NOTE,
"search_url": search_url,
"pages": pages_payload,
"listing_count": listing_count,
"unique_listing_count": len(unique_ids),
"unique_listing_ids": sorted(unique_ids),
"early_stopped": early_stopped,
}
ids_fetched = len(records)
inserted_total = int(upsert.get("inserted", 0))
updated_total = int(upsert.get("updated", 0))
images_upserted = int(upsert.get("images_upserted", 0))
cars_upserted = inserted_total + updated_total
logger.debug(
"mobile.de sync_search batch upsert: run_id=%s pages=%s inserted=%s updated=%s images=%s",
run_id,
pages_collected,
inserted_total,
updated_total,
images_upserted,
)
if progress_callback is not None:
progress_callback(
"db_upsert_done",
"search_collection_done",
{
"run_id": run_id,
"pages_collected": pages_collected,
"pages_in_batch": pages_collected,
"inserted": inserted_total,
"updated": updated_total,
"images_upserted": images_upserted,
"listing_count": listing_count,
"unique_listing_count": len(unique_ids),
},
)
if head_cut_triggered:
logger.info(
"mobile.de only_new head-cut applied: kept=%s skipped_existing=%s streak=%s",
len(records),
skipped_existing,
MOBILEDE_ONLY_NEW_STOP_ON_EXISTING_STREAK,
)
data["early_stopped"] = head_cut_triggered
self.persistence.finish_sync_run(
run_id,
status="success",
@@ -432,7 +455,16 @@ class MobileDeScraper:
data.get("listing_count", 0),
data.get("unique_listing_count", 0),
)
return {"run_id": run_id, "upsert": upsert, "skipped_existing": skipped_existing, **data}
return {
"run_id": run_id,
"upsert": {
"inserted": inserted_total,
"updated": updated_total,
"images_upserted": images_upserted,
},
"skipped_existing": skipped_existing,
**data,
}
except Exception as exc:
self.persistence.finish_sync_run(
run_id,