Fix openrent scraping

finder/openrent.py

@@ -6,11 +6,20 @@ impersonation to make requests with those cookies.
 
 OpenRent is a rental-only platform, so this scraper only handles RENT channel.
 
-HTML parsing notes:
-  OpenRent server-renders property cards in the search results page.
-  Property cards use class "pli" (property list item). Each card contains
-  a link to the property detail page, price, bedrooms, and address info.
-  The CSS selectors below may need updating if OpenRent changes their markup.
+HTML structure (as of 2026-03):
+  Search results page renders property cards as <a class="pli search-property-card">.
+  Each card contains:
+    - Monthly price in <div class="pim"> with <span class="text-primary">£X,XXX</span>
+    - Weekly price in <div class="piw"> (hidden by Alpine.js)
+    - Title in <div class="fw-medium text-primary fs-3">N Bed Type, Location, OUTCODE</div>
+    - Features in <ul> with <li> items like "1 Bed", "1 Bath", "Furnished"
+    - Listing ID in data-listing-id on the .or-swiper div
+    - Description snippet in <div class="line-clamp-2">
+
+  Detail page has:
+    - <h1> with property title including outcode
+    - <div id="map" data-lat="..." data-lng="..."> for coordinates
+    - Tables with deposit, rent, furnishing, tenant preferences
 """
 
 import logging
@@ -70,7 +79,7 @@ def solve_waf() -> tuple[dict[str, str], str] | None:
             if "AwsWafIntegration" in content:
                 log.info("Got WAF challenge page, waiting for resolution...")
                 page.wait_for_selector(
-                    "a.pli, .pli, [class*=propertyListing]", timeout=30000,
+                    "a.pli, .pli, .search-property-card", timeout=30000,
                 )
 
             raw_cookies = context.cookies()
@@ -195,6 +204,17 @@ def fetch_page(
 # ---------------------------------------------------------------------------
 
 
+def _extract_price_from_element(el) -> tuple[int, str] | None:
+    """Extract price integer from a price element's text like '£2,100'."""
+    if not el:
+        return None
+    text = el.get_text(strip=True)
+    match = re.search(r"£([\d,]+)", text)
+    if not match:
+        return None
+    return int(match.group(1).replace(",", ""))
+
+
 def _extract_price(text: str) -> tuple[int, str] | None:
     """Extract price and frequency from text like '£1,500 pcm' or '£350 pw'.
     Returns (price_int, frequency) or None.
@@ -227,6 +247,25 @@ def _extract_bedrooms_from_title(title: str) -> int | None:
     return None
 
 
+def _extract_beds_baths_from_features(feature_items: list) -> tuple[int | None, int | None]:
+    """Extract bedrooms and bathrooms from feature list items.
+
+    OpenRent search cards have <ul> with items like:
+      <li>1 Bed</li>  <li>1 Bath</li>  <li>Furnished</li>
+    """
+    bedrooms = None
+    bathrooms = None
+    for li in feature_items:
+        text = li.get_text(strip=True).lower()
+        bed_match = re.search(r"(\d+)\s*bed", text)
+        if bed_match:
+            bedrooms = int(bed_match.group(1))
+        bath_match = re.search(r"(\d+)\s*bath", text)
+        if bath_match:
+            bathrooms = int(bath_match.group(1))
+    return bedrooms, bathrooms
+
+
 def _extract_postcode(text: str) -> str | None:
     """Extract full UK postcode from text like '2 Bed Flat, Pimlico, SW1V 2AA'."""
     match = re.search(r"([A-Z]{1,2}\d[A-Z0-9]?\s*\d[A-Z]{2})", text, re.IGNORECASE)
@@ -235,23 +274,69 @@ def _extract_postcode(text: str) -> str | None:
     return None
 
 
+def _extract_outcode(text: str) -> str | None:
+    """Extract UK outcode from text like '1 Bed Flat, Bank Chambers, SW1Y'.
+
+    Looks for an outcode pattern (e.g., SW1Y, E1, EC2A) at the end of the text
+    or after the last comma."""
+    # Try after last comma first (most reliable position in OpenRent titles)
+    parts = text.split(",")
+    if len(parts) > 1:
+        last_part = parts[-1].strip()
+        match = re.match(r"^([A-Z]{1,2}\d[A-Z0-9]?)$", last_part, re.IGNORECASE)
+        if match:
+            return match.group(1).upper()
+
+    # Fall back to searching anywhere in text
+    match = re.search(r"\b([A-Z]{1,2}\d[A-Z0-9]?)\b", text, re.IGNORECASE)
+    if match:
+        candidate = match.group(1).upper()
+        # Avoid matching things like "1 Bed" → "1B"
+        if len(candidate) >= 2 and not candidate[0].isdigit():
+            return candidate
+    return None
+
+
+def _infer_property_type(title: str) -> str:
+    """Infer property type from title text.
+
+    Order matters: "Room in a Shared Flat" should be "Room" not "Flat",
+    so check "room" before "flat"."""
+    lower = title.lower()
+    if "room in" in lower or "room " in lower:
+        return "Room"
+    if "studio" in lower:
+        return "Studio"
+    if "flat" in lower or "apartment" in lower:
+        return "Flat"
+    if "maisonette" in lower:
+        return "Maisonette"
+    if "house" in lower:
+        return "House"
+    if "bungalow" in lower:
+        return "Bungalow"
+    return ""
+
+
 def parse_search_results(html: str) -> list[dict]:
     """Parse property data from OpenRent search results HTML.
 
     Returns list of raw property dicts extracted from property cards.
-    Uses multiple fallback selectors for resilience against markup changes.
+
+    Current OpenRent card structure (2026-03):
+      <a class="pli search-property-card" href="/property-to-rent/.../ID">
+        <div class="or-swiper" data-listing-id="ID">
+        <div class="pim"><span class="text-primary">£2,100</span> per month</div>
+        <div class="piw"><span class="text-primary">£485</span> per week</div>
+        <div class="fw-medium text-primary fs-3">1 Bed Flat, Location, SW1Y</div>
+        <ul>...<li>1 Bed</li><li>1 Bath</li><li>Furnished</li>...</ul>
     """
     soup = BeautifulSoup(html, "html.parser")
     properties = []
 
-    # Try known selectors for property cards (most specific first)
+    # Property cards: <a class="pli search-property-card">
     cards = soup.select("a.pli")
     if not cards:
-        cards = soup.select(".pli.clearfix")
-    if not cards:
-        cards = soup.select("[class*='propertyListing']")
-    if not cards:
-        # Last resort: look for links that match property URL pattern
         cards = soup.find_all("a", href=re.compile(r"/property-to-rent/"))
 
     if not cards:
@@ -274,37 +359,87 @@ def parse_search_results(html: str) -> list[dict]:
         id_match = re.search(r"/(\d+)(?:\?|$|#)", href)
         if id_match:
             prop["id"] = id_match.group(1)
+        else:
+            # Try data-listing-id on the swiper element
+            swiper = card.select_one("[data-listing-id]")
+            if swiper:
+                prop["id"] = swiper["data-listing-id"]
             else:
                 continue  # can't use a property without an ID
 
-        # Extract card text for parsing
-        card_text = card.get_text(" ", strip=True)
+        # --- Price ---
+        # Prefer structured price elements over free-text parsing.
+        # Monthly price is in <div class="pim"><span class="text-primary">£X</span>
+        pim = card.select_one(".pim .text-primary, .pim span")
+        piw = card.select_one(".piw .text-primary, .piw span")
 
-        # Price
+        monthly_price = _extract_price_from_element(pim)
+        weekly_price = _extract_price_from_element(piw)
+
+        if monthly_price:
+            prop["price"] = monthly_price
+            prop["frequency"] = "monthly"
+        elif weekly_price:
+            prop["price"] = weekly_price
+            prop["frequency"] = "weekly"
+        else:
+            # Fall back to parsing card text
+            card_text = card.get_text(" ", strip=True)
             price_result = _extract_price(card_text)
             if price_result:
                 prop["price"], prop["frequency"] = price_result
 
-        # Title / address — try specific elements first, fall back to card text
-        title_el = card.select_one(
-            ".listing-title, .banda, h2, h3, [class*='title']"
-        )
-        prop["title"] = (
-            title_el.get_text(strip=True) if title_el
-            else card_text[:200]
+        # --- Title / Address ---
+        # The property title is in a div with classes "fw-medium text-primary fs-3"
+        # e.g., "1 Bed Flat, Bank Chambers, SW1Y"
+        title_el = card.select_one("div.fw-medium.fs-3")
+        if not title_el:
+            # Fallback: try image alt text which also has the title
+            img = card.select_one("img.propertyPic")
+            if img and img.get("alt"):
+                prop["title"] = img["alt"]
+            else:
+                # Last resort: extract from card text, excluding price/nav noise
+                prop["title"] = ""
+        else:
+            prop["title"] = title_el.get_text(strip=True)
+
+        # --- Bedrooms / Bathrooms from feature list ---
+        feature_list = card.select("ul li")
+        beds_from_features, baths_from_features = _extract_beds_baths_from_features(
+            feature_list,
         )
 
-        # Bedrooms from title text
-        beds = _extract_bedrooms_from_title(prop["title"])
+        # Bedrooms: prefer feature list, fall back to title parsing
+        if beds_from_features is not None:
+            prop["bedrooms"] = beds_from_features
+        else:
+            beds = _extract_bedrooms_from_title(prop.get("title", ""))
             if beds is not None:
                 prop["bedrooms"] = beds
 
-        # Postcode from title
-        postcode = _extract_postcode(prop["title"])
+        if baths_from_features is not None:
+            prop["bathrooms"] = baths_from_features
+
+        # --- Property type from title ---
+        title = prop.get("title", "")
+        prop["property_type"] = _infer_property_type(title)
+
+        # --- Postcode / outcode from title ---
+        postcode = _extract_postcode(title)
         if postcode:
             prop["postcode"] = postcode
+        else:
+            outcode = _extract_outcode(title)
+            if outcode:
+                prop["outcode"] = outcode
 
-        # Coordinates from data attributes (if present on card or child elements)
+        # --- Description snippet ---
+        desc_el = card.select_one(".line-clamp-2")
+        if desc_el:
+            prop["description"] = desc_el.get_text(strip=True)
+
+        # --- Coordinates from data attributes (may not be present on cards) ---
         for el in [card] + card.select("[data-lat], [data-latitude]"):
             lat = el.get("data-lat") or el.get("data-latitude")
             lng = (
@@ -329,62 +464,35 @@ def parse_search_results(html: str) -> list[dict]:
 def parse_property_detail(html: str) -> dict:
     """Parse a single property detail page for additional data.
 
-    Extracts: bedrooms, bathrooms, price, property_type, postcode,
-    lat/lng (from map data), description (for floor area).
+    Current detail page structure (2026-03):
+      - <h1> has the full title (e.g., "Room in a Shared House, Lime Tree Court, AL2")
+      - <div id="map" data-lat="..." data-lng="..."> has coordinates
+      - Tables have "Rent PCM", "Deposit", "Bills Included", etc. (NOT bedrooms)
+      - Description in elements with class containing "description"
     """
     soup = BeautifulSoup(html, "html.parser")
     details: dict = {}
 
-    # Parse structured data tables (class "table table-striped")
-    for table in soup.select("table.table-striped, table.table"):
-        for row in table.select("tr"):
-            cells = row.select("td, th")
-            if len(cells) < 2:
-                continue
-            label = cells[0].get_text(strip=True).lower()
-            value = cells[1].get_text(strip=True)
-
-            if "bedroom" in label:
-                match = re.search(r"(\d+)", value)
-                if match:
-                    details["bedrooms"] = int(match.group(1))
-            elif "bathroom" in label:
-                match = re.search(r"(\d+)", value)
-                if match:
-                    details["bathrooms"] = int(match.group(1))
-            elif "rent" in label or "price" in label:
-                match = re.search(r"£([\d,]+)", value)
-                if match:
-                    details["price"] = int(match.group(1).replace(",", ""))
-            elif "type" in label:
-                details["property_type"] = value
-            elif "available" in label or "move" in label:
-                details["available_date"] = value
-            elif "furnish" in label:
-                details["furnished"] = value
-
-    # Extract postcode from page title / address heading.
-    # Prefer h1 (the actual property title) over generic [class*='title']
-    # which may match nav elements like "Log in".
-    title_tag = soup.select_one("h1")
-    if not title_tag:
-        title_tag = soup.select_one(".property-title, [class*='title']")
-    if title_tag:
-        title_text = title_tag.get_text(strip=True)
+    # --- Title from h1 ---
+    h1 = soup.select_one("h1")
+    if h1:
+        title_text = h1.get_text(strip=True)
+        # Validate it's not a nav/modal element (e.g. "Log in")
+        if len(title_text) > 10 and "log in" not in title_text.lower():
             details["title"] = title_text
             postcode = _extract_postcode(title_text)
             if postcode:
                 details["postcode"] = postcode
 
-    # Extract coordinates from map element data attributes
-    map_el = soup.select_one("[data-lat], [data-latitude]")
+    # --- Coordinates from map element ---
+    # The map div has id="map" with data-lat and data-lng
+    map_el = soup.select_one("#map[data-lat]")
+    if not map_el:
+        # Fallback: any element with data-lat (but prefer #map)
+        map_el = soup.select_one("[data-lat]")
     if map_el:
-        lat = map_el.get("data-lat") or map_el.get("data-latitude")
-        lng = (
-            map_el.get("data-lng")
-            or map_el.get("data-longitude")
-            or map_el.get("data-lon")
-        )
+        lat = map_el.get("data-lat")
+        lng = map_el.get("data-lng") or map_el.get("data-lon")
         if lat and lng:
             try:
                 details["lat"] = float(lat)
@@ -392,7 +500,35 @@ def parse_property_detail(html: str) -> dict:
             except ValueError:
                 pass
 
-    # Also check for coordinates in JSON-LD or inline JavaScript
+    # --- Parse tables for rent and property details ---
+    for table in soup.select("table"):
+        for row in table.select("tr"):
+            cells = row.select("td")
+            if len(cells) < 2:
+                continue
+            label = cells[0].get_text(strip=True).lower()
+            value = cells[1].get_text(strip=True)
+
+            if "rent" in label and "pcm" in label:
+                match = re.search(r"£([\d,]+)", value)
+                if match:
+                    details["price"] = int(match.group(1).replace(",", ""))
+            elif "bedroom" in label:
+                match = re.search(r"(\d+)", value)
+                if match:
+                    details["bedrooms"] = int(match.group(1))
+            elif "bathroom" in label:
+                match = re.search(r"(\d+)", value)
+                if match:
+                    details["bathrooms"] = int(match.group(1))
+            elif "type" in label and "property" in label:
+                details["property_type"] = value
+            elif "available" in label or "move" in label:
+                details["available_date"] = value
+            elif "furnish" in label:
+                details["furnished"] = value
+
+    # --- Coordinates from inline JavaScript (last resort) ---
     if "lat" not in details:
         for script in soup.select("script"):
             text = script.string or ""
@@ -406,7 +542,7 @@ def parse_property_detail(html: str) -> dict:
                     pass
                 break
 
-    # Extract description for floor area parsing
+    # --- Description for floor area ---
     desc_el = soup.select_one(
         ".description, [class*='description'], #description"
     )
@@ -464,6 +600,16 @@ def parse_floor_area(description: str | None) -> float | None:
 # ---------------------------------------------------------------------------
 
 
+def _resolve_outcode_postcodes(
+    outcode: str,
+    pc_coords: dict[str, tuple[float, float]],
+) -> list[str]:
+    """Get all postcodes for an outcode from the postcode coordinates lookup."""
+    prefix = outcode + " "
+    # Also try without space for non-standard format (e.g. "SW1Y" matches "SW1Y 4AA")
+    return [pcd for pcd in pc_coords if pcd.startswith(prefix)]
+
+
 def transform_property(
     search_data: dict,
     detail_data: dict | None,
@@ -473,7 +619,7 @@ def transform_property(
     """Transform OpenRent property data into our output schema.
 
     Merges data from the search results page and (optionally) the detail page.
-    Uses pc_coords (postcode → lat/lng) as a fallback when coordinates are
+    Uses pc_coords (postcode -> lat/lng) as a fallback when coordinates are
     missing but a postcode is available.
     """
     detail = detail_data or {}
@@ -487,7 +633,7 @@ def transform_property(
 
     frequency = search_data.get("frequency", "monthly")
 
-    # Get postcode: detail page > search card > spatial index
+    # Get postcode: detail page > search card
     postcode = detail.get("postcode") or search_data.get("postcode")
 
     if lat is not None and lng is not None:
@@ -496,7 +642,15 @@ def transform_property(
             log.debug("Coords outside England: lat=%.4f lng=%.4f — skipping", lat, lng)
             return None
         if not postcode:
+            if pc_index:
                 postcode = pc_index.nearest(lat, lng)
+            elif search_data.get("outcode"):
+                # No spatial index — try outcode lookup as fallback
+                outcode_pcs = _resolve_outcode_postcodes(
+                    search_data["outcode"], pc_coords,
+                )
+                if outcode_pcs:
+                    postcode = outcode_pcs[0]
     elif postcode:
         # Have postcode but no coordinates — look up centroid from arcgis data
         coords = pc_coords.get(postcode)
@@ -505,6 +659,17 @@ def transform_property(
         else:
             log.debug("Postcode %s not in arcgis data — skipping", postcode)
             return None
+    elif search_data.get("outcode"):
+        # Have only outcode — find postcodes in that outcode and use centroid
+        outcode = search_data["outcode"]
+        outcode_postcodes = _resolve_outcode_postcodes(outcode, pc_coords)
+        if outcode_postcodes:
+            # Use the first postcode as a rough approximation
+            postcode = outcode_postcodes[0]
+            lat, lng = pc_coords[postcode]
+        else:
+            log.debug("No postcodes found for outcode %s — skipping", outcode)
+            return None
     else:
         return None
 
@@ -513,37 +678,45 @@ def transform_property(
         return None
 
     bedrooms = detail.get("bedrooms") or search_data.get("bedrooms", 0) or 0
-    bathrooms = detail.get("bathrooms", 0) or 0
+    bathrooms = detail.get("bathrooms") or search_data.get("bathrooms", 0) or 0
 
+    # Title: prefer detail page (has h1 with full title)
     title = detail.get("title") or search_data.get("title", "")
-    address = title.split(",")[0].strip() if title else ""
 
-    property_type = detail.get("property_type", "")
-    # Infer from title if not found in detail page
+    # Address: take the middle part of the title (skip the "N Bed Type" prefix
+    # and the outcode suffix). E.g., "1 Bed Flat, Bank Chambers, SW1Y" -> "Bank Chambers"
+    address = ""
+    if title:
+        parts = [p.strip() for p in title.split(",")]
+        if len(parts) >= 3:
+            # Skip first (type) and last (outcode), join the middle
+            address = ", ".join(parts[1:-1])
+        elif len(parts) == 2:
+            # Could be "Location, OUTCODE" or "Type, Location"
+            # If last part looks like an outcode, use the first part
+            if re.match(r"^[A-Z]{1,2}\d", parts[-1].strip()):
+                address = parts[0]
+            else:
+                address = parts[1]
+        else:
+            address = title
+
+    # Property type: prefer detail, then search card, then infer from title
+    property_type = detail.get("property_type") or search_data.get("property_type", "")
     if not property_type and title:
-        lower = title.lower()
-        if "flat" in lower or "apartment" in lower:
-            property_type = "Flat"
-        elif "studio" in lower:
-            property_type = "Studio"
-        elif "maisonette" in lower:
-            property_type = "Maisonette"
-        elif "house" in lower:
-            property_type = "House"
-        elif "room" in lower:
-            property_type = "Room"
+        property_type = _infer_property_type(title)
 
     prop_id = search_data.get("id", "")
     listing_url = search_data.get(
         "url", f"{OPENRENT_BASE}/{prop_id}" if prop_id else "",
     )
-    description = detail.get("description", "")
+    description = detail.get("description") or search_data.get("description", "")
 
     return {
         "id": f"or_{prop_id}",
         "Bedrooms": bedrooms,
         "Bathrooms": bathrooms,
-        "Number of bedrooms & living rooms": bedrooms + bathrooms,
+        "Number of bedrooms & living rooms": bedrooms,
         "lon": lng,
         "lat": lat,
         "Postcode": postcode,
@@ -571,14 +744,14 @@ def search_outcode(
     """Search OpenRent for rental properties in one outcode.
 
     1. Fetches the search results page for the outcode
-    2. Parses property cards from the HTML
-    3. Optionally fetches each property's detail page for full data
+    2. Parses property cards from the HTML (title, price, beds, baths)
+    3. Fetches each property's detail page for coordinates
     4. Transforms to common output schema
 
-    Args:
-        fetch_details: If True, visits each property's detail page for
-            coordinates and extra data. Slower but more complete.
-            If False, relies only on search card data + postcode lookup.
+    The search card provides most data (price, bedrooms, bathrooms, title,
+    property type). Detail pages are needed primarily for precise coordinates
+    and full postcodes. When detail pages fail, we fall back to outcode-level
+    coordinates from the postcode lookup.
     """
     search_url = f"{OPENRENT_BASE}/properties-to-rent/?term={outcode}&isLive=true"