Fix exclude_pending and mls_only filters not working with raw return type

When return_type="raw" was specified, the exclude_pending and mls_only
parameters were ignored because these filters only existed in
process_property(), which is bypassed for raw data returns.

Changes:
- Added _apply_raw_data_filters() method to handle client-side filtering
  for raw data
- Applied the filter in search() method after sorting but before returning
- Fixed exclude_pending to check flags.is_pending and flags.is_contingent
- Fixed mls_only to check source.id (not mls.id which doesn't exist in raw data)
- Added comprehensive tests for both filters with raw data

Fixes #140

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

README.md

@@ -278,7 +278,9 @@ Optional
 │
 ├── exclude_pending (True/False): If set, excludes 'pending' properties from the 'for_sale' results unless listing_type is 'pending'
 │
-└── limit (integer): Limit the number of properties to fetch. Max & default is 10000.
+├── limit (integer): Limit the number of properties to fetch. Max & default is 10000.
+│
+└── offset (integer): Starting position for pagination within the 10k limit. Use with limit to fetch results in chunks.
 ```
 
 ### Property Schema

homeharvest/__init__.py

@@ -1,7 +1,7 @@
 import warnings
 import pandas as pd
 from .core.scrapers import ScraperInput
-from .utils import process_result, ordered_properties, validate_input, validate_dates, validate_limit, validate_datetime, validate_filters, validate_sort
+from .utils import process_result, ordered_properties, validate_input, validate_dates, validate_limit, validate_offset, validate_datetime, validate_filters, validate_sort
 from .core.scrapers.realtor import RealtorScraper
 from .core.scrapers.models import ListingType, SearchPropertyType, ReturnType, Property
 from typing import Union, Optional, List
@@ -21,6 +21,7 @@ def scrape_property(
     extra_property_data: bool = True,
     exclude_pending: bool = False,
     limit: int = 10000,
+    offset: int = 0,
     # New date/time filtering parameters
     past_hours: int = None,
     datetime_from: str = None,
@@ -61,6 +62,7 @@ def scrape_property(
     :param extra_property_data: Increases requests by O(n). If set, this fetches additional property data (e.g. agent, broker, property evaluations etc.)
     :param exclude_pending: If true, this excludes pending or contingent properties from the results, unless listing type is pending.
     :param limit: Limit the number of results returned. Maximum is 10,000.
+    :param offset: Starting position for pagination within the 10k limit (offset + limit cannot exceed 10,000). Use with limit to fetch results in chunks (e.g., offset=200, limit=200 fetches results 200-399). Should be a multiple of 200 (page size) for optimal performance. Default is 0. Note: Cannot be used to bypass the 10k API limit - use date ranges (date_from/date_to) to narrow searches and fetch more data.
 
     New parameters:
     :param past_hours: Get properties in the last _ hours (requires client-side filtering)
@@ -77,6 +79,7 @@ def scrape_property(
     validate_input(listing_type)
     validate_dates(date_from, date_to)
     validate_limit(limit)
+    validate_offset(offset, limit)
     validate_datetime(datetime_from)
     validate_datetime(datetime_to)
     validate_filters(
@@ -100,6 +103,7 @@ def scrape_property(
         extra_property_data=extra_property_data,
         exclude_pending=exclude_pending,
         limit=limit,
+        offset=offset,
         # New date/time filtering
         past_hours=past_hours,
         datetime_from=datetime_from,

homeharvest/core/scrapers/__init__.py

@@ -25,6 +25,7 @@ class ScraperInput(BaseModel):
     extra_property_data: bool | None = True
     exclude_pending: bool | None = False
     limit: int = 10000
+    offset: int = 0
     return_type: ReturnType = ReturnType.pandas
 
     # New date/time filtering parameters
@@ -106,6 +107,7 @@ class Scraper:
         self.extra_property_data = scraper_input.extra_property_data
         self.exclude_pending = scraper_input.exclude_pending
         self.limit = scraper_input.limit
+        self.offset = scraper_input.offset
         self.return_type = scraper_input.return_type
 
         # New date/time filtering

homeharvest/core/scrapers/realtor/__init__.py

@@ -405,13 +405,23 @@ class RealtorScraper(Scraper):
 
         if self.return_type != ReturnType.raw:
             with ThreadPoolExecutor(max_workers=self.NUM_PROPERTY_WORKERS) as executor:
-                futures = [executor.submit(process_property, result, self.mls_only, self.extra_property_data, 
-                                         self.exclude_pending, self.listing_type, get_key, process_extra_property_details) for result in properties_list]
+                # Store futures with their indices to maintain sort order
+                futures_with_indices = [
+                    (i, executor.submit(process_property, result, self.mls_only, self.extra_property_data,
+                                       self.exclude_pending, self.listing_type, get_key, process_extra_property_details))
+                    for i, result in enumerate(properties_list)
+                ]
 
-                for future in as_completed(futures):
+                # Collect results and sort by index to preserve API sort order
+                results = []
+                for idx, future in futures_with_indices:
                     result = future.result()
                     if result:
-                        properties.append(result)
+                        results.append((idx, result))
+
+                # Sort by index and extract properties in correct order
+                results.sort(key=lambda x: x[0])
+                properties = [result for idx, result in results]
         else:
             properties = properties_list
 
@@ -428,7 +438,7 @@ class RealtorScraper(Scraper):
         location_type = location_info["area_type"]
 
         search_variables = {
-            "offset": 0,
+            "offset": self.offset,
         }
 
         search_type = (
@@ -473,21 +483,30 @@ class RealtorScraper(Scraper):
         homes = result["properties"]
 
         with ThreadPoolExecutor() as executor:
-            futures = [
-                executor.submit(
+            # Store futures with their offsets to maintain proper sort order
+            # Start from offset + page_size and go up to offset + limit
+            futures_with_offsets = [
+                (i, executor.submit(
                     self.general_search,
                     variables=search_variables | {"offset": i},
                     search_type=search_type,
-                )
+                ))
                 for i in range(
-                    self.DEFAULT_PAGE_SIZE,
-                    min(total, self.limit),
+                    self.offset + self.DEFAULT_PAGE_SIZE,
+                    min(total, self.offset + self.limit),
                     self.DEFAULT_PAGE_SIZE,
                 )
             ]
 
-            for future in as_completed(futures):
-                homes.extend(future.result()["properties"])
+            # Collect results and sort by offset to preserve API sort order across pages
+            results = []
+            for offset, future in futures_with_offsets:
+                results.append((offset, future.result()["properties"]))
+
+            # Sort by offset and concatenate in correct order
+            results.sort(key=lambda x: x[0])
+            for offset, properties in results:
+                homes.extend(properties)
 
         # Apply client-side hour-based filtering if needed
         # (API only supports day-level filtering, so we post-filter for hour precision)
@@ -498,6 +517,16 @@ class RealtorScraper(Scraper):
         elif self.listing_type == ListingType.PENDING and (self.last_x_days or self.date_from):
             homes = self._apply_pending_date_filter(homes)
 
+        # Apply client-side sort to ensure results are properly ordered
+        # This is necessary after filtering and to guarantee sort order across page boundaries
+        if self.sort_by:
+            homes = self._apply_sort(homes)
+
+        # Apply raw data filters (exclude_pending and mls_only) for raw return type
+        # These filters are normally applied in process_property() but are bypassed for raw data
+        if self.return_type == ReturnType.raw:
+            homes = self._apply_raw_data_filters(homes)
+
         return homes
 
     def _apply_hour_based_date_filter(self, homes):
@@ -722,6 +751,101 @@ class RealtorScraper(Scraper):
             return date_range['from_date'] <= date_obj <= date_range['to_date']
         return False
 
+    def _apply_sort(self, homes):
+        """Apply client-side sorting to ensure results are properly ordered.
+
+        This is necessary because:
+        1. Multi-page results need to be re-sorted after concatenation
+        2. Filtering operations may disrupt the original sort order
+
+        Args:
+            homes: List of properties (either dicts or Property objects)
+
+        Returns:
+            Sorted list of properties
+        """
+        if not homes or not self.sort_by:
+            return homes
+
+        def get_sort_key(home):
+            """Extract the sort field value from a home (handles both dict and Property object)."""
+            if isinstance(home, dict):
+                value = home.get(self.sort_by)
+            else:
+                # Property object
+                value = getattr(home, self.sort_by, None)
+
+            # Handle None values - push them to the end
+            if value is None:
+                # Use a sentinel value that sorts to the end
+                return (1, 0) if self.sort_direction == "desc" else (1, float('inf'))
+
+            # For datetime fields, convert string to datetime for proper sorting
+            if self.sort_by in ['list_date', 'sold_date', 'pending_date']:
+                if isinstance(value, str):
+                    try:
+                        from datetime import datetime
+                        # Handle timezone indicators
+                        date_value = value
+                        if date_value.endswith('Z'):
+                            date_value = date_value[:-1] + '+00:00'
+                        parsed_date = datetime.fromisoformat(date_value)
+                        return (0, parsed_date)
+                    except (ValueError, AttributeError):
+                        # If parsing fails, treat as None
+                        return (1, 0) if self.sort_direction == "desc" else (1, float('inf'))
+                return (0, value)
+
+            # For numeric fields, ensure we can compare
+            return (0, value)
+
+        # Sort the homes
+        reverse = (self.sort_direction == "desc")
+        sorted_homes = sorted(homes, key=get_sort_key, reverse=reverse)
+
+        return sorted_homes
+
+    def _apply_raw_data_filters(self, homes):
+        """Apply exclude_pending and mls_only filters for raw data returns.
+
+        These filters are normally applied in process_property(), but that function
+        is bypassed when return_type="raw", so we need to apply them here instead.
+
+        Args:
+            homes: List of properties (either dicts or Property objects)
+
+        Returns:
+            Filtered list of properties
+        """
+        if not homes:
+            return homes
+
+        # Only filter raw data (dict objects)
+        # Property objects have already been filtered in process_property()
+        if homes and not isinstance(homes[0], dict):
+            return homes
+
+        filtered_homes = []
+
+        for home in homes:
+            # Apply exclude_pending filter
+            if self.exclude_pending and self.listing_type != ListingType.PENDING:
+                flags = home.get('flags', {})
+                is_pending = flags.get('is_pending', False)
+                is_contingent = flags.get('is_contingent', False)
+
+                if is_pending or is_contingent:
+                    continue  # Skip this property
+
+            # Apply mls_only filter
+            if self.mls_only:
+                source = home.get('source', {})
+                if not source or not source.get('id'):
+                    continue  # Skip this property
+
+            filtered_homes.append(home)
+
+        return filtered_homes
 
 
     @retry(

homeharvest/utils.py

@@ -1,5 +1,6 @@
 from __future__ import annotations
 import pandas as pd
+import warnings
 from datetime import datetime
 from .core.scrapers.models import Property, ListingType, Advertisers
 from .exceptions import InvalidListingType, InvalidDate
@@ -182,6 +183,36 @@ def validate_limit(limit: int) -> None:
         raise ValueError("Property limit must be between 1 and 10,000.")
 
 
+def validate_offset(offset: int, limit: int = 10000) -> None:
+    """Validate offset parameter for pagination.
+
+    Args:
+        offset: Starting position for results pagination
+        limit: Maximum number of results to fetch
+
+    Raises:
+        ValueError: If offset is invalid or if offset + limit exceeds API limit
+    """
+    if offset is not None and offset < 0:
+        raise ValueError("Offset must be non-negative (>= 0).")
+
+    # Check if offset + limit exceeds API's hard limit of 10,000
+    if offset is not None and limit is not None and (offset + limit) > 10000:
+        raise ValueError(
+            f"offset ({offset}) + limit ({limit}) = {offset + limit} exceeds API maximum of 10,000. "
+            f"The API cannot return results beyond position 10,000. "
+            f"To fetch more results, narrow your search."
+        )
+
+    # Warn if offset is not a multiple of 200 (API page size)
+    if offset is not None and offset > 0 and offset % 200 != 0:
+        warnings.warn(
+            f"Offset should be a multiple of 200 (page size) for optimal performance. "
+            f"Using offset {offset} may result in less efficient pagination.",
+            UserWarning
+        )
+
+
 def validate_datetime(datetime_str: str | None) -> None:
     """Validate ISO 8601 datetime format."""
     if not datetime_str:

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "homeharvest"
-version = "0.7.1"
+version = "0.7.3"
 description = "Real estate scraping library"
 authors = ["Zachary Hampton <zachary@bunsly.com>", "Cullen Watson <cullen@bunsly.com>"]
 homepage = "https://github.com/ZacharyHampton/HomeHarvest"

tests/test_realtor.py

@@ -870,66 +870,111 @@ def test_combined_filters():
 
 
 def test_sorting_by_price():
-    """Test sorting by list_price - note API sorting may not be perfect"""
+    """Test sorting by list_price with actual sort order validation"""
 
-    # Sort ascending (cheapest first)
+    # Sort ascending (cheapest first) with multi-page limit to test concatenation
     result_asc = scrape_property(
         location="Orlando, FL",
         listing_type="for_sale",
         sort_by="list_price",
         sort_direction="asc",
-        limit=20
+        limit=250  # Multi-page to test concatenation logic
     )
 
     assert result_asc is not None and len(result_asc) > 0
 
+    # Verify ascending sort order (allow for None/NA values at the end)
+    prices_asc = result_asc["list_price"].dropna().tolist()
+    assert len(prices_asc) > 0, "No properties with prices found"
+    assert prices_asc == sorted(prices_asc), f"Prices not in ascending order: {prices_asc[:10]}"
+
     # Sort descending (most expensive first)
     result_desc = scrape_property(
         location="San Antonio, TX",
         listing_type="for_sale",
         sort_by="list_price",
         sort_direction="desc",
-        limit=20
+        limit=250  # Multi-page to test concatenation logic
     )
 
     assert result_desc is not None and len(result_desc) > 0
 
-    # Note: Realtor API sorting may not be perfectly reliable for all search types
-    # The test ensures the sort parameters don't cause errors, actual sort order may vary
+    # Verify descending sort order (allow for None/NA values at the end)
+    prices_desc = result_desc["list_price"].dropna().tolist()
+    assert len(prices_desc) > 0, "No properties with prices found"
+    assert prices_desc == sorted(prices_desc, reverse=True), f"Prices not in descending order: {prices_desc[:10]}"
 
 
 def test_sorting_by_date():
-    """Test sorting by list_date - note API sorting may not be perfect"""
+    """Test sorting by list_date with actual sort order validation"""
 
-    result = scrape_property(
+    # Test descending (newest first) with multi-page limit
+    result_desc = scrape_property(
         location="Columbus, OH",
         listing_type="for_sale",
         sort_by="list_date",
         sort_direction="desc",  # Newest first
-        limit=20
+        limit=250  # Multi-page to test concatenation logic
     )
 
-    assert result is not None and len(result) > 0
+    assert result_desc is not None and len(result_desc) > 0
 
-    # Test ensures sort parameter doesn't cause errors
-    # Note: Realtor API sorting may not be perfectly reliable for all search types
+    # Verify descending sort order (allow for None/NA values at the end)
+    dates_desc = result_desc["list_date"].dropna().tolist()
+    assert len(dates_desc) > 0, "No properties with dates found"
+    assert dates_desc == sorted(dates_desc, reverse=True), f"Dates not in descending order (newest first): {dates_desc[:10]}"
+
+    # Test ascending (oldest first)
+    result_asc = scrape_property(
+        location="Columbus, OH",
+        listing_type="for_sale",
+        sort_by="list_date",
+        sort_direction="asc",  # Oldest first
+        limit=250
+    )
+
+    assert result_asc is not None and len(result_asc) > 0
+
+    # Verify ascending sort order
+    dates_asc = result_asc["list_date"].dropna().tolist()
+    assert len(dates_asc) > 0, "No properties with dates found"
+    assert dates_asc == sorted(dates_asc), f"Dates not in ascending order (oldest first): {dates_asc[:10]}"
 
 
 def test_sorting_by_sqft():
-    """Test sorting by square footage - note API sorting may not be perfect"""
+    """Test sorting by square footage with actual sort order validation"""
 
-    result = scrape_property(
+    # Test descending (largest first) with multi-page limit
+    result_desc = scrape_property(
         location="Indianapolis, IN",
         listing_type="for_sale",
         sort_by="sqft",
         sort_direction="desc",  # Largest first
-        limit=20
+        limit=250  # Multi-page to test concatenation logic
     )
 
-    assert result is not None and len(result) > 0
+    assert result_desc is not None and len(result_desc) > 0
 
-    # Test ensures sort parameter doesn't cause errors
-    # Note: Realtor API sorting may not be perfectly reliable for all search types
+    # Verify descending sort order (allow for None/NA values at the end)
+    sqfts_desc = result_desc["sqft"].dropna().tolist()
+    assert len(sqfts_desc) > 0, "No properties with sqft found"
+    assert sqfts_desc == sorted(sqfts_desc, reverse=True), f"Square footages not in descending order: {sqfts_desc[:10]}"
+
+    # Test ascending (smallest first)
+    result_asc = scrape_property(
+        location="Indianapolis, IN",
+        listing_type="for_sale",
+        sort_by="sqft",
+        sort_direction="asc",  # Smallest first
+        limit=250
+    )
+
+    assert result_asc is not None and len(result_asc) > 0
+
+    # Verify ascending sort order
+    sqfts_asc = result_asc["sqft"].dropna().tolist()
+    assert len(sqfts_asc) > 0, "No properties with sqft found"
+    assert sqfts_asc == sorted(sqfts_asc), f"Square footages not in ascending order: {sqfts_asc[:10]}"
 
 
 def test_filter_validation_errors():
@@ -1224,4 +1269,84 @@ def test_last_status_change_date_hour_filtering():
                     assert pending_date >= cutoff_time, \
                         f"PENDING property pending_date {pending_date} should be within 48 hours of {cutoff_time}"
                 except (ValueError, TypeError):
-                    pass  # Skip if parsing fails
+                    pass  # Skip if parsing fails
+
+
+def test_exclude_pending_with_raw_data():
+    """Test that exclude_pending parameter works correctly with return_type='raw'"""
+
+    # Query for sale properties with exclude_pending=True and raw data
+    result = scrape_property(
+        location="Phoenix, AZ",
+        listing_type="for_sale",
+        exclude_pending=True,
+        return_type="raw",
+        limit=50
+    )
+
+    assert result is not None and len(result) > 0
+
+    # Verify that no pending or contingent properties are in the results
+    for prop in result:
+        flags = prop.get('flags', {})
+        is_pending = flags.get('is_pending', False)
+        is_contingent = flags.get('is_contingent', False)
+
+        assert not is_pending, f"Property {prop.get('property_id')} should not be pending when exclude_pending=True"
+        assert not is_contingent, f"Property {prop.get('property_id')} should not be contingent when exclude_pending=True"
+
+
+def test_mls_only_with_raw_data():
+    """Test that mls_only parameter works correctly with return_type='raw'"""
+
+    # Query with mls_only=True and raw data
+    result = scrape_property(
+        location="Dallas, TX",
+        listing_type="for_sale",
+        mls_only=True,
+        return_type="raw",
+        limit=50
+    )
+
+    assert result is not None and len(result) > 0
+
+    # Verify that all properties have MLS IDs (stored in source.id)
+    for prop in result:
+        source = prop.get('source', {})
+        mls_id = source.get('id') if source else None
+
+        assert mls_id is not None and mls_id != "", \
+            f"Property {prop.get('property_id')} should have an MLS ID (source.id) when mls_only=True, got: {mls_id}"
+
+
+def test_combined_filters_with_raw_data():
+    """Test that both exclude_pending and mls_only work together with return_type='raw'"""
+
+    # Query with both filters enabled and raw data
+    result = scrape_property(
+        location="Austin, TX",
+        listing_type="for_sale",
+        exclude_pending=True,
+        mls_only=True,
+        return_type="raw",
+        limit=30
+    )
+
+    assert result is not None and len(result) > 0
+
+    # Verify both filters are applied
+    for prop in result:
+        # Check exclude_pending filter
+        flags = prop.get('flags', {})
+        is_pending = flags.get('is_pending', False)
+        is_contingent = flags.get('is_contingent', False)
+
+        assert not is_pending, f"Property {prop.get('property_id')} should not be pending"
+        assert not is_contingent, f"Property {prop.get('property_id')} should not be contingent"
+
+        # Check mls_only filter
+        source = prop.get('source', {})
+        mls_id = source.get('id') if source else None
+
+        assert mls_id is not None and mls_id != "", \
+            f"Property {prop.get('property_id')} should have an MLS ID (source.id)"