sync only new cars

iaai_scraper/scraper.py

@@ -1,4 +1,5 @@
 import logging
+import re
 import signal
 import time
 import uuid
@@ -21,10 +22,18 @@ from .storage.listing import ListingCollector
 from .storage.schemas import CarRecord
 
 logger = logging.getLogger("iaai_scraper.scraper")
+VEHICLE_ID_RE = re.compile(r"/VehicleDetail/(\d+)(?:~[A-Z]{2})?", re.IGNORECASE)
 
 
 class IAAIScraper:
 
+    @staticmethod
+    def _extract_origin_id_from_url(vehicle_url: str) -> str | None:
+        match = VEHICLE_ID_RE.search(vehicle_url)
+        if not match:
+            return None
+        return match.group(1)
+
     def __init__(self, runtime_settings: Settings | None = None) -> None:
         # Базовые зависимости и сервисы скрапера.
         self.settings = runtime_settings or settings
@@ -260,7 +269,14 @@ class IAAIScraper:
                 error_summary=error_summary,
             )
 
-    def sync_listing(self, make: str | None = None, model: str | None = None, lane: str = "iaai_cars", limit: int | None = None):
+    def sync_listing(
+        self,
+        make: str | None = None,
+        model: str | None = None,
+        lane: str = "iaai_cars",
+        limit: int | None = None,
+        only_new: bool | None = None,
+    ):
         """Листинг + sync всех найденных машин."""
         # Массовая синхронизация с общим run_id и сбором ошибок.
         trace_id = self._new_trace_id("sync-listing")
@@ -271,6 +287,7 @@ class IAAIScraper:
         cars_failed = 0
         images_upserted = 0
         total = 0
+        skipped_existing = 0
         failures: list[dict[str, str]] = []
         listing: dict = {}
 
@@ -280,6 +297,27 @@ class IAAIScraper:
             if limit is not None:
                 vehicle_urls = vehicle_urls[:max(0, limit)]
 
+            effective_only_new = self.settings.sync_only_new if only_new is None else only_new
+            if effective_only_new:
+                existing_urls = self.persistence.get_existing_origin_urls(vehicle_urls)
+                url_to_origin_id = {
+                    url: self._extract_origin_id_from_url(url)
+                    for url in vehicle_urls
+                }
+                candidate_origin_ids = [origin_id for origin_id in url_to_origin_id.values() if origin_id]
+                existing_ids = self.persistence.get_existing_origin_ids(candidate_origin_ids)
+
+                known_urls = {
+                    url
+                    for url in vehicle_urls
+                    if (url in existing_urls) or (url_to_origin_id.get(url) in existing_ids)
+                }
+
+                skipped_existing = len(known_urls)
+                if skipped_existing:
+                    logger.info("Filtering already known vehicles: skipped %d", skipped_existing)
+                vehicle_urls = [url for url in vehicle_urls if url not in known_urls]
+
             total = len(vehicle_urls)
             logger.info("Starting sync: %d vehicles to process", total)
 
@@ -345,6 +383,7 @@ class IAAIScraper:
             "cars_upserted": cars_upserted,
             "cars_failed": cars_failed,
             "images_upserted": images_upserted,
+            "skipped_existing": skipped_existing,
             "elapsed_seconds": round(time.perf_counter() - started_at, 3),
             "failures": failures,
         }