IAAI scraper
This commit is contained in:
qananasikq
2
iaai_scraper/parsing/__init__.py
Normal file
2
iaai_scraper/parsing/__init__.py
Normal file
@@ -0,0 +1,2 @@
|
||||
from .mapper import * # noqa: F401,F403
|
||||
from .parser import * # noqa: F401,F403
|
||||
309
iaai_scraper/parsing/mapper.py
Normal file
309
iaai_scraper/parsing/mapper.py
Normal file
@@ -0,0 +1,309 @@
|
||||
import hashlib
|
||||
import json
|
||||
import re
|
||||
from datetime import datetime, timezone
|
||||
from typing import Any
|
||||
from urllib.parse import urlparse
|
||||
|
||||
from ..core.utils import first_non_empty
|
||||
from ..storage.enums import (
|
||||
BODY_TYPE_ENUM_VALUES,
|
||||
COUNTRY_ENUM_VALUES,
|
||||
CURRENCY_ENUM_VALUES,
|
||||
DRIVE_ENUM_VALUES,
|
||||
GEARBOX_ENUM_VALUES,
|
||||
ORIGIN_ENUM_VALUES,
|
||||
SELLING_TYPE_ENUM_VALUES,
|
||||
STEERING_WHEEL_ENUM_VALUES,
|
||||
)
|
||||
from ..storage.schemas import CarRecord, ImageRecord
|
||||
|
||||
|
||||
class CarMapper:
|
||||
"""IAAI data → CarRecord."""
|
||||
|
||||
BODY_MAP = {
|
||||
"sedan": "SEDAN", "coupe": "COUPE", "hatchback": "HATCHBACK", "sport utility": "SUV",
|
||||
"suv": "SUV", "wagon": "STATION_WAGON", "station wagon": "STATION_WAGON", "pickup": "PICKUP",
|
||||
"pickup truck": "PICKUP", "crew cab": "PICKUP", "extended cab": "PICKUP", "regular cab": "PICKUP",
|
||||
"quad cab": "PICKUP", "double cab": "PICKUP", "king cab": "PICKUP", "mega cab": "PICKUP",
|
||||
"supercab": "PICKUP", "supercrew": "PICKUP", "truck": "TRUCK", "van": "MINIVAN",
|
||||
"minivan": "MINIVAN", "convertible": "OPEN", "cabriolet": "OPEN", "rv": "RV", "crossover": "SUV",
|
||||
}
|
||||
DRIVE_MAP = {
|
||||
"front wheel drive": "FWD", "fwd": "FWD", "rear wheel drive": "RWD", "rwd": "RWD",
|
||||
"all wheel drive": "4WD", "awd": "4WD", "4x4": "4WD", "four wheel drive": "4WD",
|
||||
"4wd": "4WD", "2wd": "2WD", "two wheel drive": "2WD",
|
||||
}
|
||||
GEARBOX_MAP = {
|
||||
"automatic": "AT", "automatic transmission": "AT", "a/t": "AT", "aut": "AT",
|
||||
"manual": "MT", "manual transmission": "MT", "m/t": "MT", "cvt": "CVT",
|
||||
"continuously variable transmission": "CVT", "electric": "EV", "ev": "EV",
|
||||
}
|
||||
STEERING_MAP = {"left": "LEFT", "left hand drive": "LEFT", "right": "RIGHT", "right hand drive": "RIGHT"}
|
||||
COUNTRY_MAP = {
|
||||
"us": "US", "usa": "US", "united states": "US", "ca": "CA", "canada": "CA",
|
||||
"jp": "JP", "japan": "JP", "kr": "KR", "korea": "KR", "south korea": "KR",
|
||||
}
|
||||
NO_DAMAGE_MARKERS = {"normal wear", "normal wear & tear", "normal wear and tear", "n/a", "na", "none", "no damage", "minor dents/scratches"}
|
||||
|
||||
def map_to_car_record(self, vehicle_url: str, vehicle_summary: dict[str, Any], payload_insights: dict[str, Any]) -> CarRecord:
|
||||
vehicle_summary = vehicle_summary or {}
|
||||
payload_insights = payload_insights or {}
|
||||
notes: list[str] = []
|
||||
core = payload_insights.get("vehicle_core", {})
|
||||
pricing = payload_insights.get("pricing", {})
|
||||
damage = payload_insights.get("damage", {})
|
||||
auction = payload_insights.get("auction", {})
|
||||
images = payload_insights.get("images", {})
|
||||
|
||||
origin_id = self._build_origin_id(vehicle_url, vehicle_summary, core)
|
||||
parser_id = f"iaai:{origin_id}"
|
||||
brand = self._as_str(first_non_empty([core.get("make"), vehicle_summary.get("make"), "UNKNOWN"]))
|
||||
model = self._as_str(first_non_empty([core.get("model"), vehicle_summary.get("model"), "UNKNOWN"]))
|
||||
year = self._to_int(first_non_empty([core.get("year"), vehicle_summary.get("year")]))
|
||||
price = self._to_int(first_non_empty([pricing.get("buy_now"), pricing.get("current_bid"), vehicle_summary.get("buy_now"), vehicle_summary.get("current_bid")]))
|
||||
mileage = self._to_int(first_non_empty([core.get("odometer"), vehicle_summary.get("odometer"), 0])) or 0
|
||||
color = self._normalize_color(first_non_empty([core.get("color"), vehicle_summary.get("color"), "other"]))
|
||||
drive = self._normalize_drive(first_non_empty([core.get("drive"), vehicle_summary.get("drive")]))
|
||||
gearbox = self._normalize_gearbox(first_non_empty([core.get("gearbox"), vehicle_summary.get("gearbox")]))
|
||||
steering = self._normalize_steering(first_non_empty([core.get("steering_wheel"), vehicle_summary.get("steering_wheel")])) or "LEFT"
|
||||
body_type = self._normalize_body_type(first_non_empty([core.get("body_type"), vehicle_summary.get("body_type")]))
|
||||
engine_volume = self._to_engine_cc(first_non_empty([core.get("engine"), core.get("engine_volume"), vehicle_summary.get("engine"), vehicle_summary.get("engine_volume")]))
|
||||
title_text = self._as_str(first_non_empty([core.get("title"), vehicle_summary.get("title"), ""]))
|
||||
seller = self._as_str(first_non_empty([core.get("seller"), vehicle_summary.get("seller"), ""]))
|
||||
location = self._as_str(first_non_empty([core.get("location"), vehicle_summary.get("location"), auction.get("branch"), ""]))
|
||||
country = self._normalize_country(first_non_empty([core.get("country"), location, "US"]))
|
||||
is_damaged = self._bool_damage(damage, vehicle_summary)
|
||||
is_sold = self._bool_sold(auction)
|
||||
one_owner = self._boolish(first_non_empty([core.get("one_owner"), vehicle_summary.get("one_owner"), False]))
|
||||
new_car = self._boolish(first_non_empty([core.get("new_car"), vehicle_summary.get("new_car"), False]))
|
||||
rental = self._boolish(first_non_empty([core.get("rental"), vehicle_summary.get("rental"), False]))
|
||||
repair_history = self._boolish(first_non_empty([core.get("repair_history"), vehicle_summary.get("repair_history"), False]))
|
||||
non_smoking = self._boolish(first_non_empty([core.get("non_smoking"), vehicle_summary.get("non_smoking"), True]))
|
||||
evaluation = self._as_str(first_non_empty([core.get("grade"), core.get("evaluation"), vehicle_summary.get("evaluation")])) or None
|
||||
currency = self._normalize_currency(first_non_empty([pricing.get("currency"), vehicle_summary.get("currency"), "USD"]))
|
||||
slug = self._slugify(" ".join(filter(None, [str(year or ""), brand, model, origin_id])))
|
||||
images_records = self._build_images(images.get("urls") or vehicle_summary.get("image_urls") or [])
|
||||
origin = "IAAI" if "IAAI" in ORIGIN_ENUM_VALUES else "NA"
|
||||
if not price:
|
||||
notes.append("Price is missing or not parseable from the observed payloads.")
|
||||
if not vehicle_summary.get("vin"):
|
||||
notes.append("VIN was not observed in the accessible payloads for this account/session.")
|
||||
if not images_records:
|
||||
notes.append("No image URLs were found in the captured payloads.")
|
||||
|
||||
raw_attributes = {
|
||||
"vin": vehicle_summary.get("vin"),
|
||||
"lot_number": first_non_empty([core.get("lot_number"), vehicle_summary.get("lot_number")]),
|
||||
"trim": first_non_empty([core.get("trim"), vehicle_summary.get("trim")]),
|
||||
"fuel_type": first_non_empty([core.get("fuel_type"), vehicle_summary.get("fuel_type")]),
|
||||
"cylinders": first_non_empty([core.get("cylinders"), vehicle_summary.get("cylinders")]),
|
||||
"engine": first_non_empty([core.get("engine"), vehicle_summary.get("engine")]),
|
||||
"manufactured_in": vehicle_summary.get("manufactured_in"),
|
||||
"vehicle_class": vehicle_summary.get("vehicle_class"),
|
||||
"run_and_drive": first_non_empty([core.get("run_and_drive"), vehicle_summary.get("run_and_drive")]),
|
||||
"keys": first_non_empty([core.get("keys"), vehicle_summary.get("keys")]),
|
||||
"title": title_text,
|
||||
"title_brand": vehicle_summary.get("title_brand"),
|
||||
"damage_primary": first_non_empty([damage.get("primary"), vehicle_summary.get("primary_damage")]),
|
||||
"damage_secondary": damage.get("secondary"),
|
||||
"damage_description": damage.get("description"),
|
||||
"buy_now": first_non_empty([pricing.get("buy_now"), vehicle_summary.get("buy_now")]),
|
||||
"current_bid": first_non_empty([pricing.get("current_bid"), vehicle_summary.get("current_bid")]),
|
||||
"actual_cash_value": pricing.get("actual_cash_value"),
|
||||
"estimated_repair_cost": pricing.get("estimated_repair_cost"),
|
||||
"seller": seller,
|
||||
"location": location,
|
||||
"vehicle_location": vehicle_summary.get("vehicle_location"),
|
||||
"auction_date": auction.get("auction_date"),
|
||||
"lane": auction.get("lane"),
|
||||
"branch": auction.get("branch"),
|
||||
"sale_status": auction.get("sale_status"),
|
||||
"source_endpoints": payload_insights.get("source_endpoints", {}),
|
||||
}
|
||||
|
||||
content_hash = hashlib.sha256(json.dumps({
|
||||
"brand": brand, "model": model, "year": year, "price": price, "mileage": mileage,
|
||||
"color": color, "drive": drive, "gearbox": gearbox, "body_type": body_type,
|
||||
"engine_volume": engine_volume, "is_damaged": is_damaged, "is_sold": is_sold,
|
||||
"image_count": len(images_records),
|
||||
}, sort_keys=True, default=str).encode()).hexdigest()
|
||||
|
||||
return CarRecord(
|
||||
parser_id=parser_id, brand=brand, model=model, year=year, price=price, currency=currency,
|
||||
mileage=mileage, country=country, is_sold=is_sold, color=color, drive=drive, gearbox=gearbox,
|
||||
steering_wheel=steering, body_type=body_type, engine_volume=engine_volume,
|
||||
selling_type=self._normalize_selling_type("AUCTION"), one_owner=one_owner, new_car=new_car,
|
||||
is_hidden=False, origin=origin, origin_url=vehicle_url, origin_id=origin_id, is_damaged=is_damaged,
|
||||
evaluation=evaluation, non_smoking=non_smoking, rental=rental, repair_history=repair_history,
|
||||
slug=slug, last_seen_at=datetime.now(timezone.utc), content_hash=content_hash, images=images_records,
|
||||
raw_attributes=raw_attributes, mapping_notes=notes,
|
||||
)
|
||||
|
||||
@staticmethod
|
||||
def _as_str(value: Any) -> str:
|
||||
return "" if value is None else str(value).strip()
|
||||
|
||||
@staticmethod
|
||||
def _to_int(value: Any) -> int | None:
|
||||
if value is None:
|
||||
return None
|
||||
if isinstance(value, bool):
|
||||
return int(value)
|
||||
if isinstance(value, (int, float)):
|
||||
return int(value)
|
||||
digits = re.sub(r"[^\d]", "", str(value))
|
||||
return int(digits) if digits else None
|
||||
|
||||
def _to_engine_cc(self, value: Any) -> int | None:
|
||||
text = str(value).lower().strip() if value is not None else ""
|
||||
if not text:
|
||||
return None
|
||||
if liters_match := re.search(r"(\d+(?:\.\d+)?)\s*l", text):
|
||||
return int(float(liters_match.group(1)) * 1000)
|
||||
if cubic_match := re.search(r"(\d{3,5})\s*(?:cc|cm3|cubic)", text):
|
||||
return int(cubic_match.group(1))
|
||||
return int(text) if re.match(r"^\d+$", text) else None
|
||||
|
||||
def _normalize_currency(self, value: Any) -> str:
|
||||
text = self._as_str(value).upper() or "USD"
|
||||
if text in CURRENCY_ENUM_VALUES:
|
||||
return text
|
||||
return "USD" if "$" in str(value) else "USD"
|
||||
|
||||
def _normalize_drive(self, value: Any) -> str | None:
|
||||
text = self._as_str(value).lower()
|
||||
if not text:
|
||||
return None
|
||||
if (mapped := self.DRIVE_MAP.get(text)) and mapped in DRIVE_ENUM_VALUES:
|
||||
return mapped
|
||||
for marker, mapped in self.DRIVE_MAP.items():
|
||||
if marker in text and mapped in DRIVE_ENUM_VALUES:
|
||||
return mapped
|
||||
return "NA"
|
||||
|
||||
def _normalize_gearbox(self, value: Any) -> str | None:
|
||||
text = self._as_str(value).lower()
|
||||
if not text:
|
||||
return None
|
||||
if (mapped := self.GEARBOX_MAP.get(text)) and mapped in GEARBOX_ENUM_VALUES:
|
||||
return mapped
|
||||
for marker, mapped in self.GEARBOX_MAP.items():
|
||||
if marker in text and mapped in GEARBOX_ENUM_VALUES:
|
||||
return mapped
|
||||
return "NA"
|
||||
|
||||
def _normalize_steering(self, value: Any) -> str | None:
|
||||
text = self._as_str(value).lower()
|
||||
if not text:
|
||||
return None
|
||||
if (mapped := self.STEERING_MAP.get(text)) and mapped in STEERING_WHEEL_ENUM_VALUES:
|
||||
return mapped
|
||||
for marker, mapped in self.STEERING_MAP.items():
|
||||
if marker in text and mapped in STEERING_WHEEL_ENUM_VALUES:
|
||||
return mapped
|
||||
return None
|
||||
|
||||
def _normalize_body_type(self, value: Any) -> str:
|
||||
text = self._as_str(value).lower()
|
||||
if not text:
|
||||
return "OTHER"
|
||||
for marker, mapped in self.BODY_MAP.items():
|
||||
if marker in text and mapped in BODY_TYPE_ENUM_VALUES:
|
||||
return mapped
|
||||
return "OTHER"
|
||||
|
||||
def _normalize_country(self, value: Any) -> str:
|
||||
text = self._as_str(value).lower()
|
||||
if not text:
|
||||
return "US"
|
||||
for marker, mapped in self.COUNTRY_MAP.items():
|
||||
if marker in text and mapped in COUNTRY_ENUM_VALUES:
|
||||
return mapped
|
||||
return "US" if "US" in COUNTRY_ENUM_VALUES else "NA"
|
||||
|
||||
def _normalize_selling_type(self, value: Any) -> str:
|
||||
text = self._as_str(value) or "AUCTION"
|
||||
return text if text in SELLING_TYPE_ENUM_VALUES else "AUCTION"
|
||||
|
||||
@staticmethod
|
||||
def _normalize_color(value: Any) -> str:
|
||||
text = str(value).strip() if value is not None else "other"
|
||||
if not text:
|
||||
return "other"
|
||||
if "/" in text:
|
||||
text = text.split("/")[0].strip()
|
||||
return text.lower() or "other"
|
||||
|
||||
@staticmethod
|
||||
def _boolish(value: Any) -> bool:
|
||||
return value if isinstance(value, bool) else str(value).strip().lower() in {"1", "true", "yes", "y", "owner", "one owner", "new"}
|
||||
|
||||
def _bool_damage(self, damage: dict[str, Any], vehicle_summary: dict[str, Any]) -> bool:
|
||||
for value in [damage.get("primary"), damage.get("secondary"), damage.get("description"), vehicle_summary.get("primary_damage")]:
|
||||
text = self._as_str(value).lower()
|
||||
if text and text not in self.NO_DAMAGE_MARKERS:
|
||||
return True
|
||||
return False
|
||||
|
||||
def _bool_sold(self, auction: dict[str, Any]) -> bool:
|
||||
return any(token in self._as_str(auction.get("sale_status")).lower() for token in ["sold", "closed", "ended"])
|
||||
|
||||
def _build_images(self, urls: list[Any]) -> list[ImageRecord]:
|
||||
best_by_key: dict[str, str] = {}
|
||||
key_order: list[str] = []
|
||||
non_keyed: list[str] = []
|
||||
for item in urls:
|
||||
if not isinstance(item, str):
|
||||
continue
|
||||
url = item.strip()
|
||||
if not url:
|
||||
continue
|
||||
if m := re.search(r'imageKeys=([^&]+)', url):
|
||||
img_key = m.group(1)
|
||||
w = int(re.search(r'width=(\d+)', url).group(1)) if re.search(r'width=(\d+)', url) else 0
|
||||
existing = best_by_key.get(img_key)
|
||||
if existing is None:
|
||||
best_by_key[img_key] = url
|
||||
key_order.append(img_key)
|
||||
else:
|
||||
existing_w = int(re.search(r'width=(\d+)', existing).group(1)) if re.search(r'width=(\d+)', existing) else 0
|
||||
if w > existing_w:
|
||||
best_by_key[img_key] = url
|
||||
elif url not in non_keyed:
|
||||
non_keyed.append(url)
|
||||
deduped = [best_by_key[k] for k in key_order] + non_keyed
|
||||
return [ImageRecord(fullres_image=self._make_fullres_url(url), preview_image=self._make_preview_url(url), order_index=index) for index, url in enumerate(deduped)]
|
||||
|
||||
@staticmethod
|
||||
def _make_fullres_url(url: str) -> str:
|
||||
if "vis.iaai.com" in url:
|
||||
return re.sub(r'height=\d+', 'height=633', re.sub(r'width=\d+', 'width=845', url))
|
||||
return url
|
||||
|
||||
@staticmethod
|
||||
def _make_preview_url(url: str) -> str:
|
||||
if "vis.iaai.com" in url:
|
||||
preview = re.sub(r'height=\d+', 'height=300', re.sub(r'width=\d+', 'width=400', url))
|
||||
if preview != url:
|
||||
return preview
|
||||
return url
|
||||
|
||||
def _build_origin_id(self, vehicle_url: str, vehicle_summary: dict[str, Any], core: dict[str, Any]) -> str:
|
||||
for value in [core.get("lot_number"), vehicle_summary.get("lot_number"), vehicle_summary.get("vin")]:
|
||||
text = self._as_str(value)
|
||||
if text:
|
||||
return text
|
||||
tail = urlparse(vehicle_url).path.rstrip("/").split("/")[-1]
|
||||
if "~" in tail:
|
||||
tail = tail.split("~")[0]
|
||||
return tail or self._slugify(vehicle_url)
|
||||
|
||||
@staticmethod
|
||||
def _slugify(value: str) -> str:
|
||||
return re.sub(r"[^a-z0-9]+", "-", value.lower()).strip("-") or "car"
|
||||
|
||||
|
||||
class IAAICarMapper(CarMapper):
|
||||
"""Совместимое имя маппера."""
|
||||
358
iaai_scraper/parsing/parser.py
Normal file
358
iaai_scraper/parsing/parser.py
Normal file
@@ -0,0 +1,358 @@
|
||||
import html as html_module
|
||||
import json
|
||||
import logging
|
||||
import re
|
||||
from typing import Any
|
||||
|
||||
from ..core.utils import LOT_RE, PRICE_RE, VIN_RE, deep_find_key, first_non_empty
|
||||
|
||||
logger = logging.getLogger("iaai_scraper.parsers")
|
||||
|
||||
|
||||
class VehicleParser:
|
||||
"""DOM + JSON парсер страницы авто."""
|
||||
|
||||
SUMMARY_KEY_MAP = {
|
||||
"vin": {"vin", "vehicleidentificationnumber"},
|
||||
"lot_number": {"lotnumber", "stockno", "itemid", "lotid", "itemnumber"},
|
||||
"year": {"year"},
|
||||
"make": {"make", "manufacturer", "brand"},
|
||||
"model": {"model"},
|
||||
"trim": {"trim", "series"},
|
||||
"odometer": {"odometer", "odometermiles", "mileage", "actualcashvalueodometer"},
|
||||
"primary_damage": {"primarydamage", "damage", "damagetype", "loss"},
|
||||
"secondary_damage": {"secondarydamage"},
|
||||
"run_and_drive": {"runanddrive", "canrunanddrive", "rundrive"},
|
||||
"buy_now": {"buynowprice", "buyitnowprice", "instantpurchaseprice"},
|
||||
"current_bid": {"currentbid", "highbid", "bidamount", "currenthighbid"},
|
||||
"actual_cash_value": {"actualcashvalue", "acv"},
|
||||
"estimated_repair_cost": {"estimatedrepaircost", "repaircost"},
|
||||
"keys": {"keys", "keystatus"},
|
||||
"title": {"titletype", "title", "documenttype"},
|
||||
"seller": {"seller", "sellername"},
|
||||
"location": {"location", "branchname", "auctionlocation", "branch"},
|
||||
"auction_date": {"auctiondate", "saledate", "liveauctiondate"},
|
||||
"body_type": {"bodytype", "bodystyle"},
|
||||
"drive": {"driveline", "drive"},
|
||||
"gearbox": {"transmission", "gearbox"},
|
||||
"engine": {"engine", "enginevolume"},
|
||||
"fuel_type": {"fueltype", "fuel"},
|
||||
"cylinders": {"cylinders", "cylindercount"},
|
||||
"color": {"color", "primarycolor", "exteriorcolor"},
|
||||
}
|
||||
|
||||
DOM_LABEL_MAP: dict[str, str] = {
|
||||
"stock #": "lot_number",
|
||||
"vin (status)": "vin",
|
||||
"vin": "vin",
|
||||
"primary damage": "primary_damage",
|
||||
"secondary damage": "secondary_damage",
|
||||
"odometer": "odometer",
|
||||
"body style": "body_type",
|
||||
"engine": "engine",
|
||||
"transmission": "gearbox",
|
||||
"drive line type": "drive",
|
||||
"fuel type": "fuel_type",
|
||||
"cylinders": "cylinders",
|
||||
"exterior/interior": "color",
|
||||
"exterior color": "color",
|
||||
"model": "model",
|
||||
"series": "trim",
|
||||
"selling branch": "location",
|
||||
"vehicle location": "vehicle_location",
|
||||
"auction date and time": "auction_date",
|
||||
"lane/run #": "lane",
|
||||
"actual cash value": "actual_cash_value",
|
||||
"estimated repair cost": "estimated_repair_cost",
|
||||
"seller": "seller",
|
||||
"title/sale doc": "title",
|
||||
"title/sale doc brand": "title_brand",
|
||||
"start code": "run_and_drive",
|
||||
"key": "keys",
|
||||
"manufactured in": "manufactured_in",
|
||||
"vehicle class": "vehicle_class",
|
||||
}
|
||||
|
||||
def _parse_dom_key_value_pairs(self, dom_text: str) -> dict[str, str]:
|
||||
result: dict[str, str] = {}
|
||||
if not dom_text:
|
||||
return result
|
||||
lines = [line.strip() for line in dom_text.split("\n") if line.strip()]
|
||||
known_labels = set(self.DOM_LABEL_MAP.keys())
|
||||
for i, line in enumerate(lines):
|
||||
clean = line.rstrip(":").lower().strip()
|
||||
if clean in known_labels and i + 1 < len(lines):
|
||||
value = lines[i + 1].strip()
|
||||
if value.rstrip(":").lower().strip() in known_labels:
|
||||
continue
|
||||
if value.lower() in ("more actions", "view", "print", "share", "back to results"):
|
||||
continue
|
||||
field_name = self.DOM_LABEL_MAP[clean]
|
||||
if field_name not in result or not result[field_name]:
|
||||
result[field_name] = value
|
||||
return result
|
||||
|
||||
def _parse_title_for_year_make_model(self, page_title: str, dom_text: str) -> dict[str, str | None]:
|
||||
result: dict[str, str | None] = {"year": None, "make": None, "model": None}
|
||||
title_match = re.match(r"(\d{4})\s+(\S+)\s+(.+?)(?:\s+for\s+)", page_title or "")
|
||||
if title_match:
|
||||
result["year"] = title_match.group(1)
|
||||
result["make"] = title_match.group(2)
|
||||
result["model"] = title_match.group(3)
|
||||
return result
|
||||
dom_match = re.search(r"(?:Search|Log In)\s*\n\s*(\d{4})\s+(\S+)\s+(.+?)(?:\n|$)", dom_text or "")
|
||||
if dom_match:
|
||||
result["year"] = dom_match.group(1)
|
||||
result["make"] = dom_match.group(2)
|
||||
result["model"] = dom_match.group(3).strip()
|
||||
return result
|
||||
|
||||
def normalize(self, vehicle_url: str, page_html: str, dom_text: str, network_dump: dict[str, Any]) -> dict[str, Any]:
|
||||
page_html = page_html or ""
|
||||
dom_text = dom_text or ""
|
||||
network_dump = network_dump or {}
|
||||
responses = network_dump.get("json_responses", [])
|
||||
payloads = [item.get("payload") for item in responses if isinstance(item.get("payload"), (dict, list))]
|
||||
dom_kv = self._parse_dom_key_value_pairs(dom_text)
|
||||
page_title = ""
|
||||
title_match = re.search(r"<title[^>]*>(.*?)</title>", page_html or "", re.IGNORECASE | re.DOTALL)
|
||||
if title_match:
|
||||
page_title = title_match.group(1).strip()
|
||||
title_parsed = self._parse_title_for_year_make_model(page_title, dom_text)
|
||||
|
||||
summary: dict[str, Any] = {"source_url": vehicle_url}
|
||||
for field, candidate_keys in self.SUMMARY_KEY_MAP.items():
|
||||
values: list[Any] = []
|
||||
for payload in payloads:
|
||||
values.extend(deep_find_key(payload, candidate_keys))
|
||||
if field in dom_kv:
|
||||
values.append(dom_kv[field])
|
||||
summary[field] = first_non_empty(values)
|
||||
|
||||
summary["year"] = summary.get("year") or title_parsed.get("year")
|
||||
summary["make"] = summary.get("make") or title_parsed.get("make")
|
||||
summary["model"] = summary.get("model") or title_parsed.get("model")
|
||||
summary["trim"] = summary.get("trim") or dom_kv.get("trim")
|
||||
summary["vin"] = summary.get("vin") or self._extract_vin(dom_text) or self._extract_vin(page_html)
|
||||
if summary.get("vin") and isinstance(summary["vin"], str):
|
||||
vin_clean = re.sub(r"\s*\(.*?\)\s*$", "", summary["vin"]).strip()
|
||||
vin_match = VIN_RE.search(vin_clean)
|
||||
summary["vin"] = vin_match.group(1) if vin_match else vin_clean
|
||||
summary["lot_number"] = summary.get("lot_number") or self._extract_lot_number(dom_text)
|
||||
summary["image_urls"] = self._extract_image_urls(payloads, page_html, vehicle_url)
|
||||
for dom_field, dom_value in dom_kv.items():
|
||||
if dom_field not in summary or not summary[dom_field]:
|
||||
summary[dom_field] = dom_value
|
||||
prices = self._extract_prices_from_text(dom_text)
|
||||
if not summary.get("actual_cash_value") and prices:
|
||||
summary["actual_cash_value"] = prices[0]
|
||||
if not summary.get("buy_now"):
|
||||
buy_now_match = re.search(r"Buy\s+Now[:\s]*\$\s*([\d,]+(?:\.\d{1,2})?)", dom_text or "", re.IGNORECASE)
|
||||
if buy_now_match:
|
||||
summary["buy_now"] = buy_now_match.group(1)
|
||||
elif prices:
|
||||
summary["buy_now"] = prices[0]
|
||||
if not summary.get("current_bid") and len(prices) > 1:
|
||||
summary["current_bid"] = prices[1]
|
||||
|
||||
embedded = self._extract_embedded_json(page_html)
|
||||
for item in embedded:
|
||||
p = item.get("payload")
|
||||
if isinstance(p, (dict, list)):
|
||||
payloads.append(p)
|
||||
for field, candidate_keys in self.SUMMARY_KEY_MAP.items():
|
||||
if not summary.get(field):
|
||||
val = first_non_empty(deep_find_key(p, candidate_keys))
|
||||
if val:
|
||||
summary[field] = val
|
||||
|
||||
return {
|
||||
"vehicle_summary": summary,
|
||||
"payload_insights": self._build_payload_insights(summary, responses, payloads, vehicle_url),
|
||||
"embedded_json": embedded,
|
||||
"dom_hints": self._dom_hints(dom_text),
|
||||
"access_notes": self._build_access_notes(summary, responses),
|
||||
}
|
||||
|
||||
def _build_payload_insights(self, summary: dict[str, Any], responses: list[dict[str, Any]], payloads: list[Any], vehicle_url: str = "") -> dict[str, Any]:
|
||||
image_urls = self._extract_image_urls(payloads, "", vehicle_url)
|
||||
return {
|
||||
"vehicle_core": {
|
||||
"vin": summary.get("vin"), "lot_number": summary.get("lot_number"), "year": summary.get("year"),
|
||||
"make": summary.get("make"), "model": summary.get("model"), "trim": summary.get("trim"),
|
||||
"odometer": summary.get("odometer"), "run_and_drive": summary.get("run_and_drive"),
|
||||
"seller": summary.get("seller"), "location": summary.get("location"), "title": summary.get("title"),
|
||||
"body_type": summary.get("body_type"), "drive": summary.get("drive"), "gearbox": summary.get("gearbox"),
|
||||
"engine": summary.get("engine"), "fuel_type": summary.get("fuel_type"), "cylinders": summary.get("cylinders"),
|
||||
"color": summary.get("color"), "keys": summary.get("keys"),
|
||||
},
|
||||
"pricing": {
|
||||
"buy_now": summary.get("buy_now"), "current_bid": summary.get("current_bid"),
|
||||
"actual_cash_value": summary.get("actual_cash_value"), "estimated_repair_cost": summary.get("estimated_repair_cost"),
|
||||
"currency": self._guess_currency(summary),
|
||||
},
|
||||
"bids": self._build_bid_insights(summary, payloads),
|
||||
"damage": {
|
||||
"primary": summary.get("primary_damage"),
|
||||
"secondary": summary.get("secondary_damage"),
|
||||
"description": first_non_empty(self._find_in_payloads(payloads, {"damageDescription", "damageDetails"})),
|
||||
},
|
||||
"auction": {
|
||||
"auction_date": summary.get("auction_date"),
|
||||
"lane": first_non_empty(self._find_in_payloads(payloads, {"lane", "lanename"})),
|
||||
"branch": first_non_empty([summary.get("location"), *self._find_in_payloads(payloads, {"branch", "branchname"})]),
|
||||
"sale_status": first_non_empty(self._find_in_payloads(payloads, {"salestatus", "auctionstatus", "status"})),
|
||||
"item_number": first_non_empty([summary.get("lot_number"), *self._find_in_payloads(payloads, {"itemnumber", "lotnumber", "lotid"})]),
|
||||
},
|
||||
"images": {"count": len(image_urls), "urls": image_urls},
|
||||
"source_endpoints": self._build_source_endpoints(responses),
|
||||
}
|
||||
|
||||
def _build_bid_insights(self, summary: dict[str, Any], payloads: list[Any]) -> dict[str, Any]:
|
||||
return {
|
||||
"amount": summary.get("current_bid"),
|
||||
"currency": self._guess_currency(summary),
|
||||
"bid_count": first_non_empty(self._find_in_payloads(payloads, {"bidcount", "numberofbids"})),
|
||||
"status": first_non_empty(self._find_in_payloads(payloads, {"bidstatus", "biddingstatus"})),
|
||||
}
|
||||
|
||||
def _build_source_endpoints(self, responses: list[dict[str, Any]]) -> dict[str, list[str]]:
|
||||
mapping = {"vehicle": [], "pricing": [], "bids": [], "damage": [], "auction": [], "images": []}
|
||||
for item in responses:
|
||||
url = item.get("url", "")
|
||||
category = item.get("category", "other")
|
||||
if category == "vehicle":
|
||||
mapping["vehicle"].append(url)
|
||||
lowered = url.lower()
|
||||
if any(token in lowered for token in ["bid", "offer"]):
|
||||
mapping["bids"].append(url)
|
||||
if any(token in lowered for token in ["damage", "report"]):
|
||||
mapping["damage"].append(url)
|
||||
if any(token in lowered for token in ["auction", "sale", "lane", "branch"]):
|
||||
mapping["auction"].append(url)
|
||||
elif category in mapping:
|
||||
mapping[category].append(url)
|
||||
return {key: list(dict.fromkeys(urls)) for key, urls in mapping.items()}
|
||||
|
||||
def _build_access_notes(self, summary: dict[str, Any], responses: list[dict[str, Any]]) -> dict[str, Any]:
|
||||
endpoints = [item.get("url", "") for item in responses]
|
||||
return {
|
||||
"vin_visible": bool(summary.get("vin")),
|
||||
"images_visible": bool(summary.get("image_urls")),
|
||||
"network_json_count": len(responses),
|
||||
|
||||
"observed_endpoints": endpoints[:20],
|
||||
}
|
||||
|
||||
@staticmethod
|
||||
def _find_in_payloads(payloads: list[Any], keys: set[str]) -> list[Any]:
|
||||
lowered = {key.lower() for key in keys}
|
||||
values: list[Any] = []
|
||||
for payload in payloads:
|
||||
values.extend(deep_find_key(payload, lowered))
|
||||
return values
|
||||
|
||||
@staticmethod
|
||||
def _guess_currency(summary: dict[str, Any]) -> str:
|
||||
for key in ["buy_now", "current_bid", "actual_cash_value", "estimated_repair_cost"]:
|
||||
value = str(summary.get(key) or "")
|
||||
if "$" in value:
|
||||
return "USD"
|
||||
if "€" in value:
|
||||
return "EUR"
|
||||
if "¥" in value:
|
||||
return "JPY"
|
||||
return "USD"
|
||||
|
||||
@staticmethod
|
||||
def _extract_vin(text: str) -> str | None:
|
||||
match = VIN_RE.search(text or "")
|
||||
return match.group(1) if match else None
|
||||
|
||||
@staticmethod
|
||||
def _extract_lot_number(text: str) -> str | None:
|
||||
match = LOT_RE.search(text or "")
|
||||
return match.group(1) if match else None
|
||||
|
||||
@staticmethod
|
||||
def _extract_prices_from_text(text: str) -> list[str]:
|
||||
return [match.group(1) for match in PRICE_RE.finditer(text or "")]
|
||||
|
||||
@staticmethod
|
||||
def _extract_embedded_json(html: str) -> list[dict[str, Any]]:
|
||||
scripts = re.findall(r"<script[^>]*>(.*?)</script>", html or "", flags=re.DOTALL | re.IGNORECASE)
|
||||
extracted: list[dict[str, Any]] = []
|
||||
for script_text in scripts:
|
||||
if "{" not in script_text and "[" not in script_text:
|
||||
continue
|
||||
try:
|
||||
parsed = json.loads(script_text.strip())
|
||||
except Exception:
|
||||
continue
|
||||
extracted.append({"type": "inline_json", "payload": parsed})
|
||||
return extracted
|
||||
|
||||
@staticmethod
|
||||
def _extract_image_urls(payloads: list[Any], html: str, vehicle_url: str = "") -> list[str]:
|
||||
vehicle_key = ""
|
||||
key_match = re.search(r"VehicleDetail/(\d+)", vehicle_url or "")
|
||||
if key_match:
|
||||
vehicle_key = key_match.group(1)
|
||||
found: list[str] = []
|
||||
for payload in payloads:
|
||||
found.extend(deep_find_key(payload, {"imageurl", "imageurls", "url", "fullsizeurl", "thumbnailurl", "originalurl"}))
|
||||
flat: list[str] = []
|
||||
for item in found:
|
||||
if isinstance(item, str) and item.startswith("http"):
|
||||
cleaned = html_module.unescape(item)
|
||||
lowered = cleaned.lower()
|
||||
if vehicle_key and "vis.iaai.com" in lowered and vehicle_key not in cleaned:
|
||||
continue
|
||||
if cleaned not in flat:
|
||||
flat.append(cleaned)
|
||||
elif isinstance(item, list):
|
||||
for child in item:
|
||||
if isinstance(child, str) and child.startswith("http"):
|
||||
cleaned = html_module.unescape(child)
|
||||
lowered = cleaned.lower()
|
||||
if vehicle_key and "vis.iaai.com" in lowered and vehicle_key not in cleaned:
|
||||
continue
|
||||
if cleaned not in flat:
|
||||
flat.append(cleaned)
|
||||
for pattern in [r'<img[^>]+(?:src|data-src)\s*=\s*["\']([^"\']+)["\']', r'data-src\s*=\s*["\']([^"\']+)["\']']:
|
||||
for match in re.finditer(pattern, html or "", re.IGNORECASE):
|
||||
url = html_module.unescape(match.group(1).strip())
|
||||
if not url.startswith("http") or url in flat:
|
||||
continue
|
||||
lowered = url.lower()
|
||||
if vehicle_key and vehicle_key in url:
|
||||
flat.append(url)
|
||||
elif any(token in lowered for token in ["vis.iaai.com", "anvis", "vehicleimage"]):
|
||||
if vehicle_key and vehicle_key not in url:
|
||||
continue
|
||||
flat.append(url)
|
||||
if vehicle_key:
|
||||
for url in re.findall(r'https?://vis\.iaai\.com[^\s"\'<>]+', html or ""):
|
||||
cleaned = html_module.unescape(url)
|
||||
if cleaned not in flat and vehicle_key in cleaned:
|
||||
flat.append(cleaned)
|
||||
filtered: list[str] = []
|
||||
for url in flat:
|
||||
lowered = url.lower()
|
||||
if any(pat in lowered for pat in {"dimensions", "threesixty", "360view", ".js", ".css", ".svg", "/home/", "iframeview"}):
|
||||
continue
|
||||
if "vis.iaai.com" in lowered and "/resizer" not in lowered:
|
||||
continue
|
||||
filtered.append(url)
|
||||
return filtered
|
||||
|
||||
@staticmethod
|
||||
def _dom_hints(text: str) -> dict[str, Any]:
|
||||
lowered = (text or "").lower()
|
||||
return {
|
||||
"has_buy_now_text": "buy now" in lowered,
|
||||
"has_run_and_drive_text": "run & drive" in lowered or "run and drive" in lowered,
|
||||
"has_damage_text": "damage" in lowered,
|
||||
"has_vin_text": "vin" in lowered,
|
||||
"has_title_text": "title" in lowered,
|
||||
}
|
||||
Reference in New Issue
Block a user