Different result between official Nominatim API and self-hosted version

[jp1337]

Hi everyone!
We have a problem with inaccurate addresses in our locally hosted version of Nominatim. A search query has returned a result that should not actually return a result. The official Nominatim API does not return a result at this search.
We have also made a debug request and compared these, but without success. We are therefore turning to the community with this issue.

What did you search for?

Official Nominatim API:
https://nominatim.openstreetmap.org/search?api_key=xxx&format=jsonv2&addressdetails=1&postalcode=71404&city=Korb&street=Berlinerstraße

Our Nominatim API:
https://xxx/nominatim/search?format=jsonv2&addressdetails=1&postalcode=71404&city=Korb&street=Berlinerstraße

What result did you get?

[
    {
        "place_id": 123353830,
        "licence": "Data © OpenStreetMap contributors, ODbL 1.0. https://osm.org/copyright",
        "osm_type": "way",
        "osm_id": 27779579,
        "boundingbox": [
            "51.2736754",
            "51.274833",
            "8.8566134",
            "8.862381"
        ],
        "lat": "51.2740548",
        "lon": "8.8593513",
        "display_name": "Berliner Straße, Korbach, Landkreis Waldeck-Frankenberg, Hessen, 34497, Deutschland",
        "place_rank": 26,
        "category": "highway",
        "type": "residential",
        "importance": 0.20000999999999994,
        "address": {
            "road": "Berliner Straße",
            "town": "Korbach",
            "municipality": "Korbach",
            "county": "Landkreis Waldeck-Frankenberg",
            "state": "Hessen",
            "ISO3166-2-lvl4": "DE-HE",
            "postcode": "34497",
            "country": "Deutschland",
            "country_code": "de"
        }
    },
    {
        "place_id": 123303706,
        "licence": "Data © OpenStreetMap contributors, ODbL 1.0. https://osm.org/copyright",
        "osm_type": "node",
        "osm_id": 5326236397,
        "boundingbox": [
            "51.2737054",
            "51.2738054",
            "8.8605081",
            "8.8606081"
        ],
        "lat": "51.2737554",
        "lon": "8.8605581",
        "display_name": "Berliner Straße, Korbach, Landkreis Waldeck-Frankenberg, Hessen, 34497, Deutschland",
        "place_rank": 30,
        "category": "highway",
        "type": "bus_stop",
        "importance": 0.10000999999999996,
        "address": {
            "highway": "Berliner Straße",
            "road": "Berliner Straße",
            "town": "Korbach",
            "municipality": "Korbach",
            "county": "Landkreis Waldeck-Frankenberg",
            "state": "Hessen",
            "ISO3166-2-lvl4": "DE-HE",
            "postcode": "34497",
            "country": "Deutschland",
            "country_code": "de"
        }
    }
]

What result did you expect?

[]

The query should not return any results because the street does not exist in the town and postcode. The result returned by Nominatim is in a different city and in a different federal state.

Further details

First, some information about our hosted version:

Version: (https://xxx/status?format=json)

{
    "status": 0,
    "message": "OK",
    "data_updated": "2023-11-28T00:00:00+00:00",
    "software_version": "4.3.2-0",
    "database_version": "4.3.2-0"
}

Version: Docker (mediagis/nominatim:4.3)

Environment Variables of Docker container:

TZ: "Europe/Berlin"
PBF_URL: "https://planet.openstreetmap.org/pbf/planet-latest.osm.pbf"
REPLICATION_URL: "https://planet.openstreetmap.org/replication/day/"
UPDATE_MODE: "continuous"
THREADS: "16"
NOMINATIM_PASSWORD: "xxx"
POSTGRES_MAINTENANCE_WORK_MEM: "20GB"
POSTGRES_WORK_MEM: "100MB"
POSTGRES_EFFECTIVE_CACHE_SIZE: "48GB"

Environment Variables inside the container (.env file):

NOMINATIM_TOKENIZER=icu
NOMINATIM_REPLICATION_URL=https://planet.openstreetmap.org/replication/day/
NOMINATIM_REPLICATION_UPDATE_INTERVAL=86400
NOMINATIM_REPLICATION_RECHECK_INTERVAL=900
NOMINATIM_IMPORT_STYLE=full
NOMINATIM_FLATNODE_FILE="/nominatim/flatnode/flatnode.file

Debug Data for self hosted request:
https://pastebin.com/nw3J5B4t

Debug Data for official Nominatim API:
https://pastebin.com/duh5E63T

Content of search.php file:

<?php

@define('CONST_Debug', $_GET['debug'] ?? false);
@define('CONST_LibDir', '/usr/local/lib/nominatim/lib-php');
@define('CONST_TokenizerDir', '/nominatim/tokenizer');
@define('CONST_NominatimVersion', '4.3.2-0');

@define('CONST_Database_DSN', 'pgsql:dbname=nominatim');
@define('CONST_Default_Language', false);
@define('CONST_Log_DB', false);
@define('CONST_Log_File', false);
@define('CONST_NoAccessControl', true);
@define('CONST_Places_Max_ID_count', 50);
@define('CONST_PolygonOutput_MaximumTypes', 1);
@define('CONST_Search_BatchMode', false);
@define('CONST_Search_NameOnlySearchFrequencyThreshold', '500');
@define('CONST_Use_US_Tiger_Data', true);
@define('CONST_MapIcon_URL', false);
@define('CONST_Search_WithinCountries', false);

require_once(CONST_LibDir.'/website/search.php');

[lonvia]

The official API uses the Python frontend and usually has a slightly different data state than other databases. Although in this case, the frontend is the more likely culprit.

[jp1337]

Thank you for your feedback on this long message. The result of the Python API is the correct one. It comes up with an empty result, as I would expect with the wrong query.
Should we wait for version 5 to implement the Python API in our production environment?