feat: allowing duplicate gbfs versions from different sources

Author: cka-y

api/src/feeds/impl/models/gbfs_version_impl.py

@@ -28,7 +28,7 @@ def from_orm(cls, version: GbfsVersionOrm | None) -> GbfsVersion | None:
             version=version.version,
             created_at=version.created_at,
             last_updated_at=latest_report.validated_at if latest_report else None,
-            latest=version.latest,
+            source=version.source,
             endpoints=[GbfsEndpointImpl.from_orm(item) for item in version.gbfsendpoints]
             if version.gbfsendpoints
             else [],

api/src/scripts/populate_db_test_data.py

@@ -135,7 +135,7 @@ def populate_test_datasets(self, filepath, db_session: "Session"):
                     self.logger.error(f"No feed found with stable_id: {version['feed_id']}")
                     continue
                 gbfs_version = Gbfsversion(
-                    id=version["id"], version=version["version"], url=version["url"], latest=version["latest"]
+                    id=version["id"], version=version["version"], url=version["url"]
                 )
                 if version.get("endpoints"):
                     for endpoint in version["endpoints"]:

api/tests/test_data/extra_test_data.json

@@ -766,7 +766,6 @@
       "id": "gbfs-system_id_1-2.3",
       "version": "2.3",
       "url": "https://www.example.com/gbfs_feed_1/2.3/gbfs.json",
-      "latest": false,
       "endpoints": [
         {
           "id": "gbfs-system_id_1-2.3-system_information",
@@ -789,7 +788,6 @@
       "id": "gbfs-system_id_1-3.0",
       "version": "3.0",
       "url": "https://www.example.com/gbfs_feed_1/3.0/gbfs.json",
-      "latest": true,
       "endpoints": [
         {
           "id": "gbfs-system_id_1-3.0-system_information",
@@ -810,7 +808,6 @@
       "id": "gbfs-system_id_2-2.3",
       "version": "2.3",
       "url": "https://www.example.com/gbfs_feed_1/2.3/gbfs.json",
-      "latest": false,
       "endpoints": [
         {
           "id": "gbfs-system_id_2-2.3-system_information",

docs/DatabaseCatalogAPI.yaml

@@ -527,11 +527,16 @@ components:
           type: string
           format: date-time
           example: 2023-07-10T22:06:00Z
-        latest:
-          description: >
-              A boolean value indicating if this is the latest version of the GBFS feed.
-          type: boolean
-          example: true
+        source:
+            description: >
+              The source from where the version was obtained. Possible values are: 
+              - `autodiscovery`: The version was obtained through the autodiscovery url.
+              - `gbfs_versions`: The version was obtained through the gbfs_versions endpoint.
+            type: string
+            enum:
+              - autodiscovery
+              - gbfs_versions
+
         endpoints:
           description: >
             A list of endpoints that are available in the version.

functions-python/gbfs_validator/src/gbfs_data_processor.py

@@ -101,7 +101,7 @@ def record_autodiscovery_request(
 
     def extract_gbfs_endpoints(
         self,
-        gbfs_json_url: str,
+        gbfs_json_url: str, extracted_from: str, latency: bool = True
     ) -> Tuple[Optional[List[GBFSEndpoint]], GBFSVersion]:
         """
         Extract GBFS endpoints from the GBFS JSON URL.
@@ -115,9 +115,11 @@ def extract_gbfs_endpoints(
             self.logger.warning(
                 "No version found in the GBFS data. Defaulting to version 1.0."
             )
-            gbfs_version = GBFSVersion("1.0", gbfs_json_url)
+            gbfs_version = GBFSVersion("1.0", gbfs_json_url, extracted_from)
         else:
-            gbfs_version = GBFSVersion(version_match[0].value, gbfs_json_url)
+            gbfs_version = GBFSVersion(
+                version_match[0].value, gbfs_json_url, extracted_from
+            )
         if not feeds_matches:
             self.logger.error(
                 "No feeds found in the GBFS data for version %s.", gbfs_version.version
@@ -133,9 +135,9 @@ def extract_gbfs_endpoints(
                 )
             except AttributeError:
                 language = None
-            endpoints += GBFSEndpoint.from_dict(feed_match.value, language)
+            endpoints += GBFSEndpoint.from_dict(feed_match.value, language, latency)
 
-        # If the autodiscovery endpoint is not listed then add it
+        # If the autodiscovery endpoint is not listed, then add it
         if not any(endpoint.name == "gbfs" for endpoint in endpoints):
             endpoints += GBFSEndpoint.from_dict(
                 [{"name": "gbfs", "url": gbfs_json_url}], None
@@ -147,6 +149,11 @@ def extract_gbfs_endpoints(
                 for endpoint in endpoints
             }.values()
         )
+        if len(unique_endpoints) != len(endpoints):
+            self.logger.warning(
+                "Duplicate endpoints found. This is a spec violation. Duplicates have been ignored."
+            )
+
         self.logger.info("Found version %s.", gbfs_version.version)
         self.logger.info(
             "Found endpoints %s.", ", ".join([endpoint.name for endpoint in endpoints])
@@ -155,10 +162,13 @@ def extract_gbfs_endpoints(
 
     def extract_gbfs_versions(self, gbfs_json_url: str) -> Optional[List[GBFSVersion]]:
         """Extract GBFS versions from the autodiscovery URL"""
-        all_endpoints, version = self.extract_gbfs_endpoints(gbfs_json_url)
+        all_endpoints, version = self.extract_gbfs_endpoints(
+            gbfs_json_url, "autodiscovery"
+        )
         if not all_endpoints or not version:
             return None
-        self.gbfs_endpoints[version.version] = all_endpoints
+        version_id = f"{self.stable_id}_{version.version}_{version.extracted_from}"
+        self.gbfs_endpoints[version_id] = all_endpoints
 
         # Fetch GBFS Versions
         gbfs_versions_endpoint = next(
@@ -172,7 +182,22 @@ def extract_gbfs_versions(self, gbfs_json_url: str) -> Optional[List[GBFSVersion
             gbfs_versions_json = fetch_gbfs_data(gbfs_versions_endpoint.url)
             versions_matches = parse("$..versions").find(gbfs_versions_json)
             if versions_matches:
-                gbfs_versions = GBFSVersion.from_dict(versions_matches[0].value)
+                extracted_versions = GBFSVersion.from_dict(
+                    versions_matches[0].value, "gbfs_versions"
+                )
+                autodiscovery_url_in_extracted = any(
+                    version.url == gbfs_json_url for version in extracted_versions
+                )
+                if not autodiscovery_url_in_extracted:
+                    self.logger.warning(
+                        "The autodiscovery URL is not included in gbfs_versions. There could be duplication"
+                        " of versions."
+                    )
+                gbfs_versions = [
+                    version
+                    for version in extracted_versions
+                    if version.url != gbfs_json_url
+                ] + [version]
                 self.logger.info(
                     "Found versions %s",
                     ", ".join([version.version for version in gbfs_versions]),
@@ -186,29 +211,6 @@ def extract_gbfs_versions(self, gbfs_json_url: str) -> Optional[List[GBFSVersion
             version
         ]  # If no gbfs_versions endpoint, return the version from the autodiscovery URL
 
-    def get_latest_version(self) -> Optional[str]:
-        """Get the latest GBFS version."""
-        max_version = max(
-            (
-                version
-                for version in self.gbfs_versions
-                if not version.version.lower().endswith("RC")
-            ),
-            key=lambda version: version.version,
-            default=None,
-        )
-        if not max_version:
-            self.logger.error(
-                "No non-RC versions found. Trying to set the latest to a RC version."
-            )
-            max_version = max(
-                self.gbfs_versions, key=lambda version: version.version, default=None
-            )
-        if not max_version:
-            self.logger.error("No versions found.")
-            return None
-        return max_version.version
-
     @with_db_session()
     def update_database_entities(self, db_session: Session) -> None:
         """Update the database entities with the processed GBFS data."""
@@ -222,9 +224,6 @@ def update_database_entities(self, db_session: Session) -> None:
             self.logger.error("GBFS feed with ID %s not found.", self.feed_id)
             return
         gbfs_versions_orm = []
-        latest_version = self.get_latest_version()
-        if not latest_version:
-            return
 
         # Deactivate versions that are not in the current feed
         active_versions = [version.version for version in self.gbfs_versions]
@@ -236,28 +235,29 @@ def update_database_entities(self, db_session: Session) -> None:
         # Update or create GBFS versions and endpoints
         for gbfs_version in self.gbfs_versions:
             gbfs_version_orm = self.update_or_create_gbfs_version(
-                db_session, gbfs_version, latest_version
+                db_session, gbfs_version
             )
             gbfs_versions_orm.append(gbfs_version_orm)
 
-            gbfs_endpoints = self.gbfs_endpoints.get(gbfs_version.version, [])
+            gbfs_endpoints = self.gbfs_endpoints.get(gbfs_version_orm.id, [])
             gbfs_endpoints_orm = []
-            features = self.validation_reports.get(gbfs_version.version, {}).get(
+            features = self.validation_reports.get(gbfs_version_orm.id, {}).get(
                 "features", []
             )
             for endpoint in gbfs_endpoints:
                 gbfs_endpoint_orm = self.update_or_create_gbfs_endpoint(
-                    db_session, gbfs_version.version, endpoint, features
+                    db_session, gbfs_version_orm.id, endpoint, features
                 )
-                gbfs_endpoint_orm.httpaccesslogs.append(
-                    Httpaccesslog(
-                        request_method=HTTPMethod.GET.value,
-                        request_url=endpoint.url,
-                        status_code=endpoint.status_code,
-                        latency_ms=endpoint.latency,
-                        response_size_bytes=endpoint.response_size_bytes,
+                if endpoint.status_code is not None:
+                    gbfs_endpoint_orm.httpaccesslogs.append(
+                        Httpaccesslog(
+                            request_method=HTTPMethod.GET.value,
+                            request_url=endpoint.url,
+                            status_code=endpoint.status_code,
+                            latency_ms=endpoint.latency,
+                            response_size_bytes=endpoint.response_size_bytes,
+                        )
                     )
-                )
                 gbfs_endpoints_orm.append(gbfs_endpoint_orm)
 
             # Deactivate endpoints that are not in the current feed
@@ -269,41 +269,40 @@ def update_database_entities(self, db_session: Session) -> None:
             gbfs_version_orm.gbfsendpoints = gbfs_endpoints_orm
 
             validation_report_orm = self.create_validation_report_entities(
-                gbfs_version_orm, self.validation_reports.get(gbfs_version.version, {})
+                gbfs_version_orm, self.validation_reports.get(gbfs_version_orm.id, {})
             )
             if validation_report_orm:
                 gbfs_version_orm.gbfsvalidationreports.append(validation_report_orm)
         gbfs_feed.gbfsversions = gbfs_versions_orm
         db_session.commit()
 
     def update_or_create_gbfs_version(
-        self, db_session: Session, gbfs_version: GBFSVersion, latest_version: str
+        self, db_session: Session, gbfs_version: GBFSVersion
     ) -> Gbfsversion:
         """Update or create a GBFS version entity."""
-        formatted_id = f"{self.stable_id}_{gbfs_version.version}"
+        formatted_id = (
+            f"{self.stable_id}_{gbfs_version.version}_{gbfs_version.extracted_from}"
+        )
         gbfs_version_orm = (
             db_session.query(Gbfsversion).filter(Gbfsversion.id == formatted_id).first()
         )
         if not gbfs_version_orm:
             gbfs_version_orm = Gbfsversion(
-                id=formatted_id, version=gbfs_version.version
+                id=formatted_id, version=gbfs_version.version, source=gbfs_version.extracted_from
             )
 
         gbfs_version_orm.url = gbfs_version.url  # Update the URL
-        gbfs_version_orm.latest = (
-            gbfs_version.version == latest_version
-        )  # Update the latest flag
         return gbfs_version_orm
 
     def update_or_create_gbfs_endpoint(
         self,
         db_session: Session,
-        version: str,
+        version_id: str,
         endpoint: GBFSEndpoint,
         features: List[str],
     ) -> Gbfsendpoint:
         """Update or create a GBFS endpoint entity."""
-        formatted_id = f"{self.stable_id}_{version}_{endpoint.name}"
+        formatted_id = f"{version_id}_{endpoint.name}"
         if endpoint.language:
             formatted_id += f"_{endpoint.language}"
         gbfs_endpoint_orm = (
@@ -346,7 +345,8 @@ def validate_gbfs_feed_versions(self) -> None:
                 json.dumps(json_report_summary), content_type="application/json"
             )
             report_summary_blob.make_public()
-            self.validation_reports[version.version] = {
+            version_id = f"{self.stable_id}_{version.version}_{version.extracted_from}"
+            self.validation_reports[version_id] = {
                 "report_summary_url": report_summary_blob.public_url,
                 "json_report_summary": json_report_summary,
                 "validation_time": date_time_utc,
@@ -356,6 +356,9 @@ def validate_gbfs_feed_versions(self) -> None:
                     if not obj.get("required", True) and obj.get("exists", False)
                 ],
             }
+            self.logger.info(
+                f"Validated GBFS feed version: {version.version} with URL: {version.url}"
+            )
 
     def create_validation_report_entities(
         self, gbfs_version_orm: Gbfsversion, validation_report_data: Dict
@@ -373,7 +376,7 @@ def create_validation_report_entities(
             return None
 
         validation_report_id = (
-            f"{self.stable_id}_v{gbfs_version_orm.version}_{validation_time}"
+            f"{self.stable_id}_v{gbfs_version_orm.id}_{validation_time}"
         )
         validation_report = Gbfsvalidationreport(
             id=validation_report_id,
@@ -401,21 +404,37 @@ def create_validation_report_entities(
     def extract_endpoints_for_all_versions(self):
         """Extract endpoints for all versions of the GBFS feed."""
         for version in self.gbfs_versions:
-            if version.version in self.gbfs_endpoints:
+            version_id = f"{self.stable_id}_{version.version}_{version.extracted_from}"
+            if version_id in self.gbfs_endpoints:
                 continue
-            endpoints, _ = self.extract_gbfs_endpoints(version.url)
+            self.logger.info(f"Extracting endpoints for version {version.version}.")
+            # Avoid fetching latency data for 'gbfs_versions' endpoint
+            endpoints, _ = self.extract_gbfs_endpoints(
+                version.url, "gbfs_versions", latency=False
+            )
             if endpoints:
-                self.gbfs_endpoints[version.version] = endpoints
+                self.gbfs_endpoints[version_id] = endpoints
             else:
                 self.logger.error("No endpoints found for version %s.", version.version)
 
     def trigger_location_extraction(self):
         """Trigger the location extraction process."""
-        latest_version = self.get_latest_version()
-        if not latest_version:
-            self.logger.error("No latest version found.")
+        autodiscovery_version = next(
+            (
+                version
+                for version in self.gbfs_versions
+                if version.extracted_from == "autodiscovery"
+            ),
+            None,
+        )
+        if not autodiscovery_version:
+            self.logger.error(
+                "No autodiscovery version found. Cannot trigger location extraction."
+            )
             return
-        endpoints = self.gbfs_endpoints.get(latest_version, [])
+        version_id = f"{self.stable_id}_{autodiscovery_version.version}_{autodiscovery_version.extracted_from}"
+        endpoints = self.gbfs_endpoints.get(version_id, [])
+
         # Find the station_information_url endpoint
         station_information_url = next(
             (

functions-python/gbfs_validator/src/gbfs_utils.py

@@ -47,13 +47,20 @@ def get_request_metadata(
 
     @staticmethod
     def from_dict(
-        data: List[Dict[str, Any]], language: Optional[str]
+        data: List[Dict[str, Any]], language: Optional[str], latency: bool = True
     ) -> List["GBFSEndpoint"]:
         """Creates a list of GBFSEndpoint objects from a list of dictionaries."""
         endpoints = []
         for file in data:
             if "name" in file and "url" in file:
-                metadata = GBFSEndpoint.get_request_metadata(file["url"])
+                if not latency:
+                    metadata = {
+                        "latency": None,
+                        "status_code": None,
+                        "response_size_bytes": None,
+                    }
+                else:
+                    metadata = GBFSEndpoint.get_request_metadata(file["url"])
                 if metadata:
                     endpoints.append(
                         GBFSEndpoint(
@@ -72,12 +79,15 @@ def from_dict(
 class GBFSVersion:
     version: str
     url: str
+    extracted_from: str
 
     @staticmethod
-    def from_dict(data: List[Dict[str, Any]]) -> List["GBFSVersion"]:
+    def from_dict(
+        data: List[Dict[str, Any]], extracted_from: str
+    ) -> List["GBFSVersion"]:
         """Creates a list of GBFSFile objects from a list of dictionaries."""
         return [
-            GBFSVersion(version["version"], version["url"])
+            GBFSVersion(version["version"], version["url"], extracted_from)
             for version in data
             if "version" in version and "url" in version
         ]