datamarket 0.9.28__py3-none-any.whl → 0.9.29__py3-none-any.whl

datamarket/interfaces/nominatim.py

@@ -3,12 +3,20 @@
 
 import gettext
 import logging
-import pycountry
-from geopy.distance import geodesic
+from typing import Any, Dict, List, Optional, Tuple, Union
 
+import pycountry
 import requests
+from geopy.distance import geodesic
+from jellyfish import jaro_winkler_similarity
 
 from ..params.nominatim import POSTCODES
+from ..utils.strings import normalize
+
+########################################################################################################################
+# PARAMETERS
+
+JARO_WINKLER_THRESHOLD = 0.85
 
 ########################################################################################################################
 # CLASSES
@@ -19,11 +27,11 @@ spanish.install()
 
 
 class GeoNames:
-    def __init__(self, endpoint):
+    def __init__(self, endpoint: str) -> None:
         self.endpoint = endpoint
 
     @staticmethod
-    def validate_postcode(postcode):
+    def validate_postcode(postcode: Union[int, str]) -> Optional[str]:
         if isinstance(postcode, int):
             postcode = str(postcode)
 
@@ -36,148 +44,245 @@ class GeoNames:
                 return postcode
 
     @staticmethod
-    def get_province_from_postcode(postcode):
+    def get_province_from_postcode(postcode: Optional[str]) -> Optional[str]:
         if postcode:
             return POSTCODES[postcode[:2]]
 
-    def reverse(self, lat, lon):
-        return requests.get(f"{self.endpoint}/reverse?lat={lat}&lon={lon}").json()
+    def reverse(self, lat: Union[float, str], lon: Union[float, str]) -> Dict[str, Any]:
+        return requests.get(f"{self.endpoint}/reverse?lat={lat}&lon={lon}", timeout=30).json()
 
 
 class Nominatim:
-    def __init__(self, nominatim_endpoint, geonames_endpoint):
+    def __init__(self, nominatim_endpoint: str, geonames_endpoint: str) -> None:
         self.endpoint = nominatim_endpoint
         self.geonames = GeoNames(geonames_endpoint)
 
     @staticmethod
-    def get_attribute(raw_json, keys):
+    def _get_attribute(raw_json: Dict[str, Any], keys: List[str]) -> Any:
         for key in keys:
             if key in raw_json:
                 return raw_json[key]
 
-    def geocode(self, address):
-        return requests.get(f"{self.endpoint}/search?q={address}&format=json").json()
+    def _calculate_distance(self, lat_str: Optional[str], lon_str: Optional[str], input_coords: Tuple[float, float]) -> float:
+        dist = float("inf")
+        if lat_str and lon_str:
+            try:
+                coords = (float(lat_str), float(lon_str))
+                dist = geodesic(input_coords, coords).km
+            except (ValueError, TypeError):
+                logger.warning("Invalid coordinates for distance calculation.")
+        return dist
+
+    def _parse_nominatim_result(self, nominatim_raw_json: Dict[str, Any]) -> Dict[str, Optional[str]]:
+        raw_address = nominatim_raw_json.get("address", {})
+
+        postcode_str = str(raw_address.get("postcode", ""))
+        postcode = self.geonames.validate_postcode(postcode_str)
+
+        city = self._get_attribute(raw_address, ["city", "town", "village"])
+        district, quarter = self._get_district_quarter(raw_address)
+
+        return {
+            "country": raw_address.get("country"),
+            "country_code": (raw_address.get("country_code") or "").lower(),
+            "state": raw_address.get("state"),
+            "province": raw_address.get("province"),
+            "city": city,
+            "postcode": postcode,
+            "district": district,
+            "quarter": quarter,
+            "street": raw_address.get("road"),
+            "number": raw_address.get("house_number"),
+        }
+
+    def _parse_geonames_result(self, geonames_raw_json: Dict[str, Any]) -> Dict[str, Optional[str]]:
+        geonames_country_code_str = geonames_raw_json.get("country_code")
+        country_name = None
+        if geonames_country_code_str:
+            try:
+                country_obj = pycountry.countries.get(alpha_2=geonames_country_code_str.upper())
+                if country_obj:
+                    country_name = spanish.gettext(country_obj.name)
+            except LookupError:
+                logger.warning(f"Country name not found for code: {geonames_country_code_str} using pycountry.")
+
+        postcode_str = str(geonames_raw_json.get("postal_code", ""))
+        postcode = self.geonames.validate_postcode(postcode_str)
+        province = self.geonames.get_province_from_postcode(postcode) if postcode else None
+        city = geonames_raw_json.get("place_name")
+
+        return {
+            "country": country_name,
+            "country_code": (geonames_country_code_str or "").lower(),
+            "state": geonames_raw_json.get("community"),
+            "province": province,
+            "city": city,
+            "postcode": postcode,
+            "district": None,
+            "quarter": None,
+            "street": None,
+            "number": None,
+        }
+
+    def _get_empty_address_result(self) -> Dict[str, None]:
+        return {
+            "country": None,
+            "country_code": None,
+            "state": None,
+            "province": None,
+            "city": None,
+            "postcode": None,
+            "district": None,
+            "quarter": None,
+            "street": None,
+            "number": None,
+        }
+
+    def _select_postcode_and_derived_province(
+        self, parsed_nominatim_result: Dict[str, Optional[str]], parsed_geonames_result: Dict[str, Optional[str]], nominatim_address_province_raw: Optional[str]
+    ) -> Tuple[Optional[str], Optional[str]]:
+        """
+        Determines the postcode and its derived province based on comparisons
+        between Nominatim and GeoNames data, and Nominatim's raw address province.
+        """
+        nominatim_postcode = parsed_nominatim_result.get("postcode")
+        geonames_postcode = parsed_geonames_result.get("postcode")
+
+        province_from_nominatim_postcode = self.geonames.get_province_from_postcode(nominatim_postcode)
+        province_from_geonames_postcode = self.geonames.get_province_from_postcode(geonames_postcode)
+
+        norm_raw_nominatim_province = (
+            normalize(nominatim_address_province_raw) if nominatim_address_province_raw else ""
+        )
+        norm_province_from_nominatim_postcode = (
+            normalize(province_from_nominatim_postcode) if province_from_nominatim_postcode else ""
+        )
+        norm_province_from_geonames_postcode = (
+            normalize(province_from_geonames_postcode) if province_from_geonames_postcode else ""
+        )
+
+        selected_postcode = None
+        selected_province_from_postcode = None
+
+        # If provinces derived from Nominatim and GeoNames postcodes differ
+        nominatim_postcode_province_matches = False
+        if norm_province_from_nominatim_postcode and norm_raw_nominatim_province:
+            nominatim_postcode_province_matches = (
+                jaro_winkler_similarity(norm_province_from_nominatim_postcode, norm_raw_nominatim_province)
+                > JARO_WINKLER_THRESHOLD
+            )
+
+        geonames_postcode_province_matches = False
+        if norm_province_from_geonames_postcode and norm_raw_nominatim_province:
+            geonames_postcode_province_matches = (
+                jaro_winkler_similarity(norm_province_from_geonames_postcode, norm_raw_nominatim_province)
+                > JARO_WINKLER_THRESHOLD
+            )
+
+        # Prefer GeoNames postcode if its province matches Nominatim's raw address province,
+        # and Nominatim's own postcode-derived province does not.
+        if nominatim_postcode_province_matches:
+            selected_postcode = nominatim_postcode
+            selected_province_from_postcode = province_from_nominatim_postcode
+        if geonames_postcode_province_matches and not nominatim_postcode_province_matches:
+            selected_postcode = geonames_postcode
+            selected_province_from_postcode = province_from_geonames_postcode
+
+        return selected_postcode, selected_province_from_postcode
+
+    def _select_final_result(
+        self,
+        parsed_nominatim_result: Dict[str, Optional[str]],
+        parsed_geonames_result: Dict[str, Optional[str]],
+        dist_nominatim: float,
+        dist_geonames: float,
+        authoritative_postcode: Optional[str],
+        authoritative_province_from_postcode: Optional[str],
+        nominatim_address_province_raw: Optional[str],
+    ) -> Dict[str, Optional[str]]:
+        """
+        Selects the final address result based on distances and applies the authoritative postcode/province.
+        """
+        if dist_nominatim <= dist_geonames and dist_nominatim != float("inf"):
+            final_result = parsed_nominatim_result
+            final_result["postcode"] = authoritative_postcode
+            final_result["province"] = nominatim_address_province_raw
+        elif dist_geonames < dist_nominatim and dist_geonames != float("inf"):
+            final_result = parsed_geonames_result
+            final_result["postcode"] = authoritative_postcode
+            final_result["province"] = authoritative_province_from_postcode
+        else:
+            final_result = self._get_empty_address_result()
+        return final_result
+
+    def _get_district_quarter(self, raw_json: Dict[str, Any]) -> Tuple[Optional[str], Optional[str]]:
+        district = self._get_attribute(raw_json, ["city_district", "suburb", "borough"])
+        quarter = self._get_attribute(raw_json, ["quarter", "neighbourhood"])
 
-    def geocode_parsed(self, address):
+        if not district and quarter:
+            district = quarter
+            quarter = None
+
+        return district, quarter
+
+    def geocode(self, address: str) -> List[Dict[str, Any]]:
+        return requests.get(f"{self.endpoint}/search?q={address}&format=json", timeout=30).json()
+
+    def geocode_parsed(self, address: str) -> Optional[Dict[str, Optional[str]]]:
         results = self.geocode(address)
 
         if results:
             return self.reverse_parsed(results[0]["lat"], results[0]["lon"])
 
-    def reverse(self, lat, lon):
-        return requests.get(f"{self.endpoint}/reverse?lat={lat}&lon={lon}&format=json").json()
+    def reverse(self, lat: Union[float, str], lon: Union[float, str]) -> Dict[str, Any]:
+        return requests.get(f"{self.endpoint}/reverse?lat={lat}&lon={lon}&format=json", timeout=30).json()
+
+    def reverse_parsed(self, lat: Union[float, str], lon: Union[float, str]) -> Dict[str, Optional[str]]:
+        nominatim_response = self.reverse(lat, lon)
+        geonames_response = self.geonames.reverse(lat, lon)
 
-    def reverse_parsed(self, lat, lon):
-        nominatim_raw_json = self.reverse(lat, lon)
-        geonames_raw_json = self.geonames.reverse(lat, lon)
+        # Initial parsing
+        parsed_nominatim_result = self._parse_nominatim_result(nominatim_response)
+        parsed_geonames_result = self._parse_geonames_result(geonames_response)
 
-        nominatim_res_lat_str = nominatim_raw_json.get("lat")
-        nominatim_res_lon_str = nominatim_raw_json.get("lon")
-        geonames_res_lat_str = geonames_raw_json.get("lat")
-        geonames_res_lon_str = geonames_raw_json.get("lon")
+        # Determine authoritative postcode
+        raw_nominatim_province = nominatim_response.get("address", {}).get("province")
+        selected_postcode, selected_province_from_postcode = self._select_postcode_and_derived_province(
+            parsed_nominatim_result, parsed_geonames_result, raw_nominatim_province
+        )
 
-        dist_nominatim = float("inf")
-        dist_geonames = float("inf")
+        # Calculate distances
+        nominatim_response_lat = nominatim_response.get("lat")
+        nominatim_response_lon = nominatim_response.get("lon")
+        geonames_response_lat = geonames_response.get("lat")
+        geonames_response_lon = geonames_response.get("lon")
 
+        input_coords = None
         try:
             input_coords = (float(lat), float(lon))
         except (ValueError, TypeError):
             logger.error(f"Invalid input coordinates for distance calculation: lat={lat}, lon={lon}")
-        else:
-            if nominatim_res_lat_str and nominatim_res_lon_str:
-                try:
-                    nominatim_coords = (float(nominatim_res_lat_str), float(nominatim_res_lon_str))
-                    dist_nominatim = geodesic(input_coords, nominatim_coords).km
-                except (ValueError, TypeError):
-                    logger.warning("Invalid Nominatim coordinates for distance calculation.")
-
-            if geonames_res_lat_str and geonames_res_lon_str:
-                try:
-                    geonames_coords = (float(geonames_res_lat_str), float(geonames_res_lon_str))
-                    dist_geonames = geodesic(input_coords, geonames_coords).km
-                except (ValueError, TypeError):
-                    logger.warning("Invalid GeoNames coordinates for distance calculation.")
-
-        if dist_nominatim <= dist_geonames and nominatim_res_lat_str is not None and nominatim_res_lon_str is not None:
-            # Use Nominatim data
-            raw_address = nominatim_raw_json.get("address", {})
-            postcode_str = str(raw_address.get("postcode", ""))
-            postcode = self.geonames.validate_postcode(postcode_str)
-            province = self.geonames.get_province_from_postcode(postcode) if postcode else None
-            city = self.get_attribute(raw_address, ["city", "town", "village"])
-            district, quarter = self.get_district_quarter(raw_address)
-
-            return {
-                "country": raw_address.get("country"),
-                "country_code": (raw_address.get("country_code") or "").lower(),
-                "state": raw_address.get("state"),
-                "province": province,
-                "city": city,
-                "postcode": postcode,
-                "district": district,
-                "quarter": quarter,
-                "street": raw_address.get("road"),
-                "number": raw_address.get("house_number"),
-            }
-
-        elif dist_geonames < dist_nominatim and geonames_res_lat_str is not None and geonames_res_lon_str is not None:
-            # Use GeoNames data
-            geonames_country_code_str = geonames_raw_json.get("country_code")
-            country_name = None
-            if geonames_country_code_str:
-                try:
-                    country_obj = pycountry.countries.get(alpha_2=geonames_country_code_str.upper())
-                    if country_obj:
-                        country_name = spanish.gettext(country_obj.name)
-                except LookupError:
-                    logger.warning(f"Country name not found for code: {geonames_country_code_str} using pycountry.")
-
-            postcode_str = str(geonames_raw_json.get("postal_code", ""))
-            postcode = self.geonames.validate_postcode(postcode_str)
-            province = self.geonames.get_province_from_postcode(postcode) if postcode else None
-            city = geonames_raw_json.get("place_name")
-
-            return {
-                "country": country_name,
-                "country_code": (geonames_country_code_str or "").lower(),
-                "state": geonames_raw_json.get("community"),
-                "province": province,
-                "city": city,
-                "postcode": postcode,
-                "district": None,
-                "quarter": None,
-                "street": None,
-                "number": None,
-            }
+            return self._get_empty_address_result()
 
-        else:
-            # Neither source provided valid coordinates
-            return {
-                "country": None,
-                "country_code": None,
-                "state": None,
-                "province": None,
-                "city": None,
-                "postcode": None,
-                "district": None,
-                "quarter": None,
-                "street": None,
-                "number": None,
-            }
-
-    def get_district_quarter(self, raw_json):
-        district = self.get_attribute(raw_json, ["city_district", "suburb", "borough"])
-        quarter = self.get_attribute(raw_json, ["quarter", "neighbourhood"])
+        dist_nominatim = self._calculate_distance(nominatim_response_lat, nominatim_response_lon, input_coords)
+        dist_geonames = self._calculate_distance(geonames_response_lat, geonames_response_lon, input_coords)
 
-        if not district and quarter:
-            district = quarter
-            quarter = None
+        # Select final result
+        final_result = self._select_final_result(
+            parsed_nominatim_result,
+            parsed_geonames_result,
+            dist_nominatim,
+            dist_geonames,
+            selected_postcode,
+            selected_province_from_postcode,
+            raw_nominatim_province,
+        )
 
-        return district, quarter
+        return final_result
 
 
 class NominatimInterface(Nominatim):
-    def __init__(self, config):
+    def __init__(self, config: Dict[str, Any]) -> None:
         if "osm" in config:
             self.config = config["osm"]
 

{datamarket-0.9.28.dist-info → datamarket-0.9.29.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: datamarket
-Version: 0.9.28
+Version: 0.9.29
 Summary: Utilities that integrate advanced scraping knowledge into just one library.
 License: GPL-3.0-or-later
 Author: DataMarket
@@ -83,6 +83,7 @@ Requires-Dist: google-auth-oauthlib (>=1.0.0,<2.0.0) ; extra == "google-auth-oau
 Requires-Dist: html2text (>=2024.0.0,<2025.0.0) ; extra == "html2text"
 Requires-Dist: httpx[http2] (>=0.28.0,<0.29.0) ; extra == "httpx"
 Requires-Dist: inflection (>=0.5.0,<0.6.0)
+Requires-Dist: jellyfish (>=1.0.0,<2.0.0)
 Requires-Dist: jinja2 (>=3.0.0,<4.0.0)
 Requires-Dist: json5 (>=0.10.0,<0.11.0) ; extra == "json5"
 Requires-Dist: lxml[html-clean] (>=5.0.0,<6.0.0) ; extra == "lxml"

{datamarket-0.9.28.dist-info → datamarket-0.9.29.dist-info}/RECORD

@@ -4,7 +4,7 @@ datamarket/interfaces/alchemy.py,sha256=4q_gLKCKPK437VKOpdBKSrCyy42P_yWxIhE7KuvH
 datamarket/interfaces/aws.py,sha256=7KLUeBxmPN7avEMPsu5HC_KHB1N7W6Anp2X8fo43mlw,2383
 datamarket/interfaces/drive.py,sha256=shbV5jpQVe_KPE-8Idx6Z9te5Zu1SmVfrvSAyd9ZIgE,2915
 datamarket/interfaces/ftp.py,sha256=o0KlJxtksbop9OjCiQRzyAa2IeG_ExVXagS6apwrAQo,1881
-datamarket/interfaces/nominatim.py,sha256=TD4OhWdIwn53Va41BS1ugogHEZw0ANKxWIfc9G2JWLU,7280
+datamarket/interfaces/nominatim.py,sha256=rUnodcRKyZ_reBtyfFFjXNqP1TN0NMScW7zSGiJQ10I,12380
 datamarket/interfaces/peerdb.py,sha256=cwYwvO740GyaPo9zLAwJsf3UeJDGDiYzjQVM9Q6s-_g,23652
 datamarket/interfaces/proxy.py,sha256=updoOStKd8-nQBbxWbnD9eOt6HksnYi-5dQ0rEySf5M,3152
 datamarket/interfaces/tinybird.py,sha256=AYrcRGNOCoCt7ojilkWa27POROee9sTCwZ61GGHEPeM,2698
@@ -19,7 +19,7 @@ datamarket/utils/soda.py,sha256=eZTXFbI1P3WoMd1MM-YjoVTpdjTcDSWuvBb7ViBMhSQ,941
 datamarket/utils/strings.py,sha256=rEX9NeBG4C7RECgT0EQebgoFoxgZMy9-7EcBSxgBANU,5654
 datamarket/utils/typer.py,sha256=FDF3l6gh3UlAFPsHCtesnekvct2rKz0oFn3uKARBQvE,814
 datamarket/utils/types.py,sha256=vxdQZdwdXrfPR4Es52gBgol-tMRIOD6oK9cBo3rB0JQ,74
-datamarket-0.9.28.dist-info/LICENSE,sha256=OXLcl0T2SZ8Pmy2_dmlvKuetivmyPd5m1q-Gyd-zaYY,35149
-datamarket-0.9.28.dist-info/METADATA,sha256=5V61eLQ9HR_8Tl7J0sRKwR_mzKbJctIjvEebNtPtqug,6546
-datamarket-0.9.28.dist-info/WHEEL,sha256=b4K_helf-jlQoXBBETfwnf4B04YC67LOev0jo4fX5m8,88
-datamarket-0.9.28.dist-info/RECORD,,
+datamarket-0.9.29.dist-info/LICENSE,sha256=OXLcl0T2SZ8Pmy2_dmlvKuetivmyPd5m1q-Gyd-zaYY,35149
+datamarket-0.9.29.dist-info/METADATA,sha256=h4DuPT0ToLAN6vSLidYyriB9gtKjjaDPcf2MWH5fm44,6588
+datamarket-0.9.29.dist-info/WHEEL,sha256=b4K_helf-jlQoXBBETfwnf4B04YC67LOev0jo4fX5m8,88
+datamarket-0.9.29.dist-info/RECORD,,