[BW] switch uuid fallback ids to address hashes

CHANGELOG.md

@@ -1,5 +1,13 @@
 # Changelog
 
+## 2026-04-25
+- [BW]: Schools that do not have a Dienststellennummer change their ID
+  from an unstable UUID to a more stable hash of name, address, zip and city.
+  (that is schools that were `BW-UUID-<something>` become `BW-FB-<something>`)
+  ⚠️ This breaks existing ids.
+  Schools that are in the dataset right now with UUID ids will be dropped
+  from the data that is published through the API and CSV-export.
+
 ## 2025-05-12
 - [SL]: Switch to data from Geoportal instead of web scraping. We do not get 
   contact details such as email and phone anymore but we (might) get more

README.md

@@ -22,11 +22,11 @@ In details, the IDs are sourced as follows:
 
 |State| ID-Source                                                                                                    | example-id                                                                 |stable|
 |-----|--------------------------------------------------------------------------------------------------------------|----------------------------------------------------------------------------|------|
-|BW| DISCH (Dienststellenschlüssel) extracted from email, fallback to WFS UUID when not available                | `BW-04154817` or `BW-UUID-00000a15-a965-4999-b9ad-05895eb0fad2`            |✅ likely (~80% with DISCH, ~20% UUID fallback)|
+|BW| DISCH (Dienststellenschlüssel) extracted from email, fallback to address hash when not available (see below) | `BW-04154817` or `BW-FB-e5c29cbf7215726b4f3515cfad6bee63e2a0bb8ded432a34e9e51c4324ec52ea`            |✅ likely (~80% with DISCH, ~20% fallback)|
 |BY| id from the WFS service                                                                                      | `BY-SCHUL_SCHULSTANDORTEGRUNDSCHULEN_2acb7d31-915d-40a9-adcf-27b38251fa48` |❓ unlikely (although we reached out to ask for canonical IDs to be published)|
 |BE| Field `bsn` (Berliner Schulnummer) from the WFS Service                                                      | `BE-02K10`                                                                 |✅ likely|
 |BB| Field `schul_nr` (Schulnummer) from thw WFS Service                                                          | `BB-111430`                                                                |✅ likely|
-|HB| Field `snr_txt` (Schulnummer) from the INSPIRE shapefile - official 3-digit ID used in Bremen materials  | `HB-002`                                                                   |✅ likely|
+|HB| Field `snr_txt` (Schulnummer) from the INSPIRE shapefile - official 3-digit ID used in Bremen materials      | `HB-002`                                                                   |✅ likely|
 |HH| Field `schul_id` From the WFS Service                                                                        | `HH-7910-0`                                                                |✅ likely|
 |HE| `school_no` URL query param of the schools's details page (identical to the Dienststellennummer)             | `HE-4024`                                                                  |✅ likely|
 |MV| Field `dstnr` from the WFS                                                                                   | `MV-75130302`                                                              |✅ likely|
@@ -38,6 +38,9 @@ In details, the IDs are sourced as follows:
 |ST| `OBJECTID` from the ArcGIS FeatureServer API (prefixed with `ARC`)                                           | `ST-ARC00001`                                                              |❓ unlikely (OBJECTID may change on data reimport)|
 |TH| `Schulnummer` from the WFS service                                                                           | `TH-10601`                                                                 |✅ likely|
 
+For Baden-Württemberg, not all schools have a Dienststellenschlüssel that we can extract. For those who don't,
+we join name, address, zip and city with a " " between each part and generate the SHA256 hash.
+
 ## Geolocations
 When available, we try to use the geolocations provided by the data publishers.
 

jedeschule/spiders/baden_wuerttemberg.py

@@ -1,3 +1,5 @@
+import hashlib
+
 import re
 import scrapy
 from scrapy import Item
@@ -32,6 +34,8 @@ def extract_disch(email: str | None) -> str | None:
     match = DISCH_RE.search(email.strip())
     return match.group(1) if match else None
 
+def create_address_based_fallback(address, city, zip):
+    return hashlib.sha256(f"{address} {zip} {city}").hexdigest()
 
 class BadenWuerttembergSpider(SchoolSpider):
     name = "baden-wuerttemberg"
@@ -136,13 +140,16 @@ def parse(self, response):
 
     @staticmethod
     def normalize(item: Item) -> School:
-        # Prefer DISCH (stable government ID) over UUID when available
-        disch = item.get("disch")
-        uuid = item.get("uuid")
-        school_id = f"BW-{disch}" if disch else f"BW-UUID-{uuid}"
+        def id():
+            # Prefer DISCH (stable government ID) when available
+            if disch := item.get('disch'):
+                return f'{disch}'
+            key = " ".join([item.get(key) or "" for key in ['name', 'address', 'zip', 'city']])
+            key_hash = hashlib.sha256(key.encode('utf-8')).hexdigest()
+            return f'FB-{key_hash}'
 
         return School(
-            id=school_id,
+            id=f"BW-{id()}",
             name=item.get("name"),
             address=item.get("address"),
             zip=item.get("zip"),