Merge pull request #3980 from lonvia/security-smells

Improve SQL query assembly

.github/PULL_REQUEST_TEMPLATE.md

@@ -0,0 +1,12 @@
+## Summary
+<!-- Describe the purpose of your pull request and, if present, link to existing issues. -->
+
+## AI usage
+<!-- Please list where and to what extent AI was used. -->
+
+## Contributor guidelines (mandatory)
+<!-- We only accept pull requests that follow our guidelines. A deliberate violation may result in a ban. -->
+
+- [ ] I have adhered to the [coding style](https://github.com/osm-search/Nominatim/blob/master/CONTRIBUTING.md#coding-style)
+- [ ] I have [tested](https://github.com/osm-search/Nominatim/blob/master/CONTRIBUTING.md#testing) the proposed changes
+- [ ] I have [disclosed](https://github.com/osm-search/Nominatim/blob/master/CONTRIBUTING.md#using-ai-assisted-code-generators) above any use of AI to generate code, documentation, or the pull request description

.github/actions/build-nominatim/action.yml

@@ -22,7 +22,7 @@ runs:
 
         - name: Install prerequisites from apt
           run: |
-            sudo apt-get install -y -qq python3-icu python3-datrie python3-jinja2 python3-psutil python3-dotenv python3-yaml python3-sqlalchemy python3-psycopg python3-asyncpg
+            sudo apt-get install -y -qq python3-icu python3-datrie python3-jinja2 python3-psutil python3-dotenv python3-yaml python3-sqlalchemy python3-psycopg python3-asyncpg python3-mwparserfromhell
           shell: bash
           if: inputs.dependencies == 'apt'
 

CONTRIBUTING.md

@@ -4,7 +4,7 @@
 
 Bugs can be reported at https://github.com/openstreetmap/Nominatim/issues.
 Please always open a separate issue for each problem. In particular, do
-not add your bugs to closed issues. They may looks similar to you but
+not add your bugs to closed issues. They may look similar to you but
 often are completely different from the maintainer's point of view.
 
 ## Workflow for Pull Requests
@@ -21,7 +21,7 @@ that you are responsible for your pull requests. You should be prepared
 to get change requests because as the maintainers we have to make sure
 that your contribution fits well with the rest of the code. Please make
 sure that you have time to react to these comments and amend the code or
-engage in a conversion. Do not expect that others will pick up your code,
+engage in a conversation. Do not expect that others will pick up your code,
 it will almost never happen.
 
 Please open a separate pull request for each issue you want to address.
@@ -38,10 +38,19 @@ description or in documentation need to
 1. clearly mark the AI-generated sections as such, for example, by
    mentioning all use of AI in the PR description, and
 2. include proof that you have run the generated code on an actual
-   installation of Nominatim. Adding and excuting tests will not be
+   installation of Nominatim. Adding and executing tests will not be
    sufficient. You need to show that the code actually solves the problem
    the PR claims to solve.
 
+## Getting Started with Development
+
+Please see the development section of the Nominatim documentation for
+
+* [an architecture overview](https://nominatim.org/release-docs/develop/develop/overview/)
+  and backgrounds on some of the algorithms
+* [how to set up a development environment](https://nominatim.org/release-docs/develop/develop/Development-Environment/)
+* and background on [how tests are organised](https://nominatim.org/release-docs/develop/develop/Testing/)
+
 
 ## Coding style
 

docs/admin/Deployment-Python.md

@@ -10,14 +10,14 @@ Nominatim. Please refer to the documentation of
 [Nginx](https://nginx.org/en/docs/) for background information on how
 to configure it.
 
-!!! Note
-    Throughout this page, we assume your Nominatim project directory is
-    located in `/srv/nominatim-project`. If you have put it somewhere else,
-    you need to adjust the commands and configuration accordingly.
-
-
 ### Installing the required packages
 
+!!! warning
+    ASGI support in gunicorn requires at least version 25.0. If you need
+    to work with an older version of gunicorn, please refer to
+    [older Nominatim deployment documentation](https://nominatim.org/release-docs/5.2/admin/Deployment-Python/)
+    to learn how to run gunicorn with uvicorn.
+
 The Nominatim frontend is best run from its own virtual environment. If
 you have already created one for the database backend during the
 [installation](Installation.md#building-nominatim), you can use that. Otherwise
@@ -37,23 +37,27 @@ cd Nominatim
 ```
 
 The recommended way to deploy a Python ASGI application is to run
-the ASGI runner [uvicorn](https://www.uvicorn.org/)
-together with [gunicorn](https://gunicorn.org/) HTTP server. We use
+the [gunicorn](https://gunicorn.org/) HTTP server. We use
 Falcon here as the web framework.
 
 Add the necessary packages to your virtual environment:
 
 ``` sh
-/srv/nominatim-venv/bin/pip install falcon uvicorn gunicorn
+/srv/nominatim-venv/bin/pip install falcon gunicorn
 ```
 
 ### Setting up Nominatim as a systemd job
 
+!!! Note
+    These instructions assume your Nominatim project directory is
+    located in `/srv/nominatim-project`. If you have put it somewhere else,
+    you need to adjust the commands and configuration accordingly.
+
 Next you need to set up the service that runs the Nominatim frontend. This is
 easiest done with a systemd job.
 
 First you need to tell systemd to create a socket file to be used by
-hunicorn. Create the following file `/etc/systemd/system/nominatim.socket`:
+gunicorn. Create the following file `/etc/systemd/system/nominatim.socket`:
 
 ``` systemd
 [Unit]
@@ -81,10 +85,8 @@ Type=simple
 User=www-data
 Group=www-data
 WorkingDirectory=/srv/nominatim-project
-ExecStart=/srv/nominatim-venv/bin/gunicorn -b unix:/run/nominatim.sock -w 4 -k uvicorn.workers.UvicornWorker "nominatim_api.server.falcon.server:run_wsgi()"
+ExecStart=/srv/nominatim-venv/bin/gunicorn -b unix:/run/nominatim.sock -w 4 --worker-class asgi --protocol uwsgi --worker-connections 1000 "nominatim_api.server.falcon.server:run_wsgi()"
 ExecReload=/bin/kill -s HUP $MAINPID
-StandardOutput=append:/var/log/gunicorn-nominatim.log
-StandardError=inherit
 PrivateTmp=true
 TimeoutStopSec=5
 KillMode=mixed
@@ -96,7 +98,10 @@ WantedBy=multi-user.target
 This sets up gunicorn with 4 workers (`-w 4` in ExecStart). Each worker runs
 its own Python process using
 [`NOMINATIM_API_POOL_SIZE`](../customize/Settings.md#nominatim_api_pool_size)
-connections to the database to serve requests in parallel.
+connections to the database to serve requests in parallel. The parameter
+`--worker-connections` restricts how many requests gunicorn will queue for
+each worker. This can help distribute work better when the server is under
+high load.
 
 Make the new services known to systemd and start it:
 
@@ -108,13 +113,15 @@ sudo systemctl enable nominatim.service
 sudo systemctl start nominatim.service
 ```
 
-This sets the service up, so that Nominatim is automatically started
+This sets the service up so that Nominatim is automatically started
 on reboot.
 
 ### Configuring nginx
 
 To make the service available to the world, you need to proxy it through
-nginx. Add the following definition to the default configuration:
+nginx. We use the binary uwsgi protocol to speed up communication
+between nginx and gunicorn. Add the following definition to the default
+configuration:
 
 ``` nginx
 upstream nominatim_service {
@@ -129,11 +136,8 @@ server {
     index /search;
 
     location / {
-            proxy_set_header Host $http_host;
-            proxy_set_header X-Forwarded-For $proxy_add_x_forwarded_for;
-            proxy_set_header X-Forwarded-Proto $scheme;
-            proxy_redirect off;
-            proxy_pass http://nominatim_service;
+        uwsgi_pass nominatim_service;
+        include uwsgi_params;
     }
 }
 ```

docs/admin/Installation.md

@@ -37,6 +37,7 @@ Furthermore the following Python libraries are required:
   * [Jinja2](https://palletsprojects.com/p/jinja/)
   * [PyICU](https://pypi.org/project/PyICU/)
   * [PyYaml](https://pyyaml.org/) (5.1+)
+  * [mwparserfromhell](https://github.com/earwig/mwparserfromhell/)
 
 These will be installed automatically when using pip installation.
 

docs/customize/Import-Styles.md

@@ -113,6 +113,7 @@ The following classifications are recognized:
 | named           | Consider as main tag, when the object has a primary name (see [names](#name-tags) below) |
 | named_with_key  | Consider as main tag, when the object has a primary name with a domain prefix. For example, if the main tag is  `bridge=yes`, then it will only be added as an extra entry, if there is a tag `bridge:name[:XXX]` for the same object. If this property is set, all names that are not domain-specific are ignored. |
 | fallback        | Consider as main tag only when no other main tag was found. Fallback always implies `named`, i.e. fallbacks are only tried for objects with primary names. |
+| postcode_area   | Tag indicates a postcode area. Copy area into the table of postcodes but only when the object is a relation and has a postcode tagged. |
 | delete          | Completely ignore the tag in any further processing |
 | extra           | Move the tag to extratags and then ignore it for further processing |
 | `<function>`| Advanced handling, see [below](#advanced-main-tag-handling) |

docs/customize/Result-Formatting.md

@@ -229,7 +229,7 @@ _None._
 
 | Option          | Description |
 |-----------------|-------------|
-| locales         | [Locale](../library/Result-Handling.md#locale) object for the requested language(s) |
+| locales         | [Locales](../library/Result-Handling.md#locale) object for the requested language(s) |
 | group_hierarchy | Setting of [group_hierarchy](../api/Details.md#output-details) parameter |
 | icon_base_url   | (optional) URL pointing to icons as set in [NOMINATIM_MAPICON_URL](Settings.md#nominatim_mapicon_url) |
 

docs/develop/Database-Layout.md

@@ -79,7 +79,7 @@ the placex table. Only three columns are special:
 Address interpolations are always ways in OSM, which is why there is no column
 `osm_type`.
 
-The **location_postcode** table holds computed centroids of all postcodes that
+The **location_postcodes** table holds computed centroids of all postcodes that
 can be found in the OSM data. The meaning of the columns is again the same
 as that of the placex table.
 

docs/develop/Development-Environment.md

@@ -56,7 +56,7 @@ The easiest way, to handle these Python dependencies is to run your
 development from within a virtual environment.
 
 ```sh
-sudo apt install libsqlite3-mod-spatialite osm2pgsql \
+sudo apt install build-essential libsqlite3-mod-spatialite osm2pgsql \
                  postgresql-postgis postgresql-postgis-scripts \
                  pkg-config libicu-dev virtualenv
 ```
@@ -68,12 +68,12 @@ virtualenv ~/nominatim-dev-venv
 ~/nominatim-dev-venv/bin/pip install\
     psutil 'psycopg[binary]' PyICU SQLAlchemy \
     python-dotenv jinja2 pyYAML \
-    mkdocs 'mkdocstrings[python]' mkdocs-gen-files \
+    mkdocs 'mkdocstrings[python]' mkdocs-gen-files mkdocs-material \
     pytest pytest-asyncio pytest-bdd flake8 \
     types-jinja2 types-markupsafe types-psutil types-psycopg2 \
     types-pygments types-pyyaml types-requests types-ujson \
-    types-urllib3 typing-extensions unicorn falcon starlette \
-    uvicorn mypy osmium aiosqlite
+    types-urllib3 typing-extensions gunicorn falcon starlette \
+    uvicorn mypy osmium aiosqlite mwparserfromhell
 ```
 
 Now enter the virtual environment whenever you want to develop:

docs/develop/Testing.md

@@ -52,6 +52,15 @@ To run the functional tests, do
 
     pytest test/bdd
 
+You can run a single feature file using expression matching:
+
+    pytest test/bdd -k osm2pgsql/import/entrances.feature
+
+This even works for running single tests by adding the line number of the
+scenario header like that:
+
+    pytest test/bdd -k 'osm2pgsql/import/entrances.feature and L4'
+
 The BDD tests create databases for the tests. You can set name of the databases
 through configuration variables in your `pytest.ini`:
 

docs/develop/search-tables.plantuml

@@ -74,15 +74,16 @@ map place_addressline {
   isaddress => BOOLEAN
 }
 
-map location_postcode {
+map location_postcodes {
   place_id => BIGINT
+  osm_id => BIGINT
   postcode => TEXT
   parent_place_id => BIGINT
   rank_search => SMALLINT
-  rank_address => SMALLINT
   indexed_status => SMALLINT
   indexed_date => TIMESTAMP
   geometry => GEOMETRY
+  centroid -> GEOMETRY
 }
 
 placex::place_id <-- search_name::place_id
@@ -94,6 +95,6 @@ search_name::nameaddress_vector --> word::word_id
 
 place_addressline -[hidden]> location_property_osmline
 search_name -[hidden]> place_addressline
-location_property_osmline -[hidden]-> location_postcode
+location_property_osmline -[hidden]-> location_postcodes
 
 @enduml

docs/library/Getting-Started.md

@@ -248,19 +248,19 @@ of the result. To do that, you first need to decide in which language the
 results should be presented. As with the names in the result itself, the
 places in `address_rows` contain all possible name translation for each row.
 
-The library has a helper class `Locale` which helps extracting a name of a
+The library has a helper class `Locales` which helps extracting a name of a
 place in the preferred language. It takes a single parameter with a list
 of language codes in the order of preference. So
 
 ``` python
-locale = napi.Locale(['fr', 'en'])
+locale = napi.Locales(['fr', 'en'])
 ```
 
 creates a helper class that returns the name preferably in French. If that is
 not possible, it tries English and eventually falls back to the default `name`
 or `ref`.
 
-The `Locale` object can be applied to a name dictionary to return the best-matching
+The `Locales` object can be applied to a name dictionary to return the best-matching
 name out of it:
 
 ``` python
@@ -273,7 +273,7 @@ component based on its `local_name` field. This is then utilized by the overall
 which has a helper function to apply the function to all its ‘address_row’ members and saves
 the result in the `locale_name` field. 
 
-However, in order to set this `local_name` field in a preferred language, you must use the `Locale`
+However, in order to set this `local_name` field in a preferred language, you must use the `Locales`
 object which contains the function `localize_results`, which explicitly sets each `local_name field`.
 
 ``` python

docs/mk_install_instructions.py

@@ -13,7 +13,8 @@ for infile in VAGRANT_PATH.glob('Install-on-*.sh'):
     outfile = f"admin/{infile.stem}.md"
     title = infile.stem.replace('-', ' ')
 
-    with mkdocs_gen_files.open(outfile, "w") as outfd, infile.open() as infd:
+    with mkdocs_gen_files.open(outfile, "w", encoding='utf-8') as outfd, \
+            infile.open(encoding='utf-8') as infd:
         print("#", title, file=outfd)
         has_empty = False
         for line in infd:

lib-lua/themes/nominatim/init.lua

@@ -65,7 +65,19 @@ local table_definitions = {
             { column = 'geometry', type = 'geometry', projection = 'WGS84', not_null = true }
         },
         indexes = {}
-    }
+    },
+    place_postcode = {
+        ids = { type = 'any', id_column = 'osm_id', type_column = 'osm_type' },
+        columns = {
+            { column = 'postcode', type = 'text', not_null = true },
+            { column = 'country_code', type = 'text' },
+            { column = 'centroid', type = 'point', projection = 'WGS84', not_null = true },
+            { column = 'geometry', type = 'geometry', projection = 'WGS84' }
+        },
+        indexes = {
+            { column = 'postcode', method = 'btree' }
+        }
+     }
 }
 
 local insert_row = {}
@@ -113,6 +125,7 @@ local PlaceTransform = {}
 
 -- Special transform meanings which are interpreted elsewhere
 PlaceTransform.fallback = 'fallback'
+PlaceTransform.postcode_area = 'postcode_area'
 PlaceTransform.delete = 'delete'
 PlaceTransform.extra = 'extra'
 
@@ -419,11 +432,25 @@ function Place:write_place(k, v, mfunc)
     return 0
 end
 
-function Place:write_row(k, v)
+
+function Place:geometry_is_valid()
     if self.geometry == nil then
         self.geometry = self.geom_func(self.object)
+
+        if self.geometry == nil or self.geometry:is_null() then
+            self.geometry = false
+            return false
+        end
+
+        return true
     end
-    if self.geometry == nil or self.geometry:is_null() then
+
+    return self.geometry ~= false
+end
+
+
+function Place:write_row(k, v)
+    if not self:geometry_is_valid() then
         return 0
     end
 
@@ -675,9 +702,6 @@ function module.process_tags(o)
     if o.address.country ~= nil and #o.address.country ~= 2 then
         o.address['country'] = nil
     end
-    if POSTCODE_FALLBACK and fallback == nil and o.address.postcode ~= nil then
-        fallback = {'place', 'postcode', PlaceTransform.always}
-    end
 
     if o.address.interpolation ~= nil then
         o:write_place('place', 'houses', PlaceTransform.always)
@@ -685,20 +709,41 @@ function module.process_tags(o)
     end
 
     -- collect main keys
+    local postcode_collect = false
     for k, v in pairs(o.intags) do
         local ktable = MAIN_KEYS[k]
         if ktable then
             local ktype = ktable[v] or ktable[1]
             if type(ktype) == 'function' then
                 o:write_place(k, v, ktype)
+            elseif ktype == 'postcode_area' then
+                postcode_collect = true
+                if o.object.type == 'relation'
+                        and o.address.postcode ~= nil
+                        and o:geometry_is_valid() then
+                    insert_row.place_postcode{
+                        postcode = o.address.postcode,
+                        centroid = o.geometry:centroid(),
+                        geometry = o.geometry
+                    }
+                end
             elseif ktype == 'fallback' and o.has_name then
                 fallback = {k, v, PlaceTransform.named}
             end
         end
     end
 
-    if fallback ~= nil and o.num_entries == 0 then
-        o:write_place(fallback[1], fallback[2], fallback[3])
+    if o.num_entries == 0 then
+        if fallback ~= nil then
+            o:write_place(fallback[1], fallback[2], fallback[3])
+        elseif POSTCODE_FALLBACK and not postcode_collect
+                and o.address.postcode ~= nil
+                and o:geometry_is_valid() then
+            insert_row.place_postcode{
+                postcode = o.address.postcode,
+                centroid = o.geometry:centroid()
+            }
+        end
     end
 end
 

lib-lua/themes/nominatim/presets.lua

@@ -117,7 +117,8 @@ module.MAIN_TAGS.all_boundaries = {
     boundary = {'named',
                 place = 'delete',
                 land_area = 'delete',
-                postal_code = 'always'},
+                protected_area = 'fallback',
+                postal_code = 'postcode_area'},
     landuse = 'fallback',
     place = 'always'
 }
@@ -198,7 +199,7 @@ module.MAIN_TAGS_POIS = function (group)
                 no = group},
     landuse = {cemetery = 'always'},
     leisure = {'always',
-               nature_reserve = 'fallback',
+               nature_reserve = 'named',
                swimming_pool = 'named',
                garden = 'named',
                common = 'named',
@@ -321,7 +322,6 @@ module.NAME_TAGS = {}
 
 module.NAME_TAGS.core = {main = {'name', 'name:*',
                                  'int_name', 'int_name:*',
-                                 'nat_name', 'nat_name:*',
                                  'reg_name', 'reg_name:*',
                                  'loc_name', 'loc_name:*',
                                  'old_name', 'old_name:*',

lib-sql/functions.sql

@@ -2,7 +2,7 @@
 --
 -- This file is part of Nominatim. (https://nominatim.org)
 --
--- Copyright (C) 2022 by the Nominatim developer community.
+-- Copyright (C) 2025 by the Nominatim developer community.
 -- For a full list of authors see the git log.
 
 {% include('functions/utils.sql') %}
@@ -18,7 +18,7 @@
     {% include 'functions/placex_triggers.sql' %}
 {% endif %}
 
-{% if 'location_postcode' in db.tables %}
+{% if 'location_postcodes' in db.tables %}
     {% include 'functions/postcode_triggers.sql' %}
 {% endif %}
 

lib-sql/functions/importance.sql

@@ -2,7 +2,7 @@
 --
 -- This file is part of Nominatim. (https://nominatim.org)
 --
--- Copyright (C) 2022 by the Nominatim developer community.
+-- Copyright (C) 2026 by the Nominatim developer community.
 -- For a full list of authors see the git log.
 
 -- Functions for interpreting wkipedia/wikidata tags and computing importance.
@@ -166,7 +166,7 @@ BEGIN
   END LOOP;
 
   -- Nothing? Then try with the wikidata tag.
-  IF result.importance is null AND extratags ? 'wikidata' THEN
+  IF extratags ? 'wikidata' THEN
     FOR match IN
 {% if 'wikimedia_importance' in db.tables %}
       SELECT * FROM wikimedia_importance
@@ -185,18 +185,18 @@ BEGIN
   END IF;
 
   -- Still nothing? Fall back to a default.
-  IF result.importance is null THEN
-    result.importance := 0.40001 - (rank_search::float / 75);
-  END IF;
+  result.importance := 0.40001 - (rank_search::float / 75);
 
 {% if 'secondary_importance' in db.tables %}
   FOR match IN
     SELECT ST_Value(rast, centroid) as importance
-    FROM secondary_importance
-    WHERE ST_Intersects(ST_ConvexHull(rast), centroid) LIMIT 1
+      FROM secondary_importance
+      WHERE ST_Intersects(ST_ConvexHull(rast), centroid) LIMIT 1
   LOOP
-    -- Secondary importance as tie breaker with 0.0001 weight.
-    result.importance := result.importance + match.importance::float / 655350000;
+    IF match.importance is not NULL THEN
+      -- Secondary importance as tie breaker with 0.0001 weight.
+      result.importance := result.importance + match.importance::float / 655350000;
+    END IF;
   END LOOP;
 {% endif %}
 

lib-sql/functions/partition-functions.sql

@@ -2,7 +2,7 @@
 --
 -- This file is part of Nominatim. (https://nominatim.org)
 --
--- Copyright (C) 2022 by the Nominatim developer community.
+-- Copyright (C) 2026 by the Nominatim developer community.
 -- For a full list of authors see the git log.
 
 DROP TYPE IF EXISTS nearfeaturecentr CASCADE;
@@ -123,10 +123,12 @@ BEGIN
     RETURN TRUE;
   END IF;
 
-  IF in_rank_search <= 4 and not in_estimate THEN
-    INSERT INTO location_area_country (place_id, country_code, geometry)
-      (SELECT in_place_id, in_country_code, geom
-       FROM split_geometry(in_geometry) as geom);
+  IF in_rank_search <= 4 THEN
+    IF not in_estimate and in_country_code is not NULL THEN
+      INSERT INTO location_area_country (place_id, country_code, geometry)
+        (SELECT in_place_id, in_country_code, geom
+         FROM split_geometry(in_geometry) as geom);
+    END IF;
     RETURN TRUE;
   END IF;
 
@@ -212,7 +214,6 @@ DECLARE
 BEGIN
 {% for partition in db.partitions %}
   IF in_partition = {{ partition }} THEN
-    DELETE FROM search_name_{{ partition }} values WHERE place_id = in_place_id;
     IF in_rank_address > 0 THEN
       INSERT INTO search_name_{{ partition }} (place_id, address_rank, name_vector, centroid)
         values (in_place_id, in_rank_address, in_name_vector, in_geometry);
@@ -251,7 +252,6 @@ BEGIN
 
 {% for partition in db.partitions %}
   IF in_partition = {{ partition }} THEN
-    DELETE FROM location_road_{{ partition }} where place_id = in_place_id;
     INSERT INTO location_road_{{ partition }} (partition, place_id, country_code, geometry)
       values (in_partition, in_place_id, in_country_code, in_geometry);
     RETURN TRUE;

lib-sql/functions/place_triggers.sql

@@ -2,7 +2,7 @@
 --
 -- This file is part of Nominatim. (https://nominatim.org)
 --
--- Copyright (C) 2022 by the Nominatim developer community.
+-- Copyright (C) 2026 by the Nominatim developer community.
 -- For a full list of authors see the git log.
 
 CREATE OR REPLACE FUNCTION place_insert()
@@ -66,7 +66,8 @@ BEGIN
     -- They get their parent from the interpolation.
     UPDATE placex p SET indexed_status = 2
       FROM planet_osm_ways w
-      WHERE w.id = NEW.osm_id and p.osm_type = 'N' and p.osm_id = any(w.nodes);
+      WHERE w.id = NEW.osm_id and p.osm_type = 'N' and p.osm_id = any(w.nodes)
+            and indexed_status = 0;
 
     -- If there is already an entry in place, just update that, if necessary.
     IF existing.osm_type is not null THEN
@@ -89,35 +90,6 @@ BEGIN
     RETURN NEW;
   END IF;
 
-  -- ---- Postcode points.
-
-  IF NEW.class = 'place' AND NEW.type = 'postcode' THEN
-    -- Pure postcodes are never queried from placex so we don't add them.
-    -- location_postcodes is filled from the place table directly.
-
-    -- Remove any old placex entry.
-    DELETE FROM placex WHERE osm_type = NEW.osm_type and osm_id = NEW.osm_id;
-
-    IF existing.osm_type IS NOT NULL THEN
-      IF coalesce(existing.address, ''::hstore) != coalesce(NEW.address, ''::hstore)
-         OR existing.geometry::text != NEW.geometry::text
-      THEN
-        UPDATE place
-          SET name = NEW.name,
-              address = NEW.address,
-              extratags = NEW.extratags,
-              admin_level = NEW.admin_level,
-              geometry = NEW.geometry
-          WHERE osm_type = NEW.osm_type and osm_id = NEW.osm_id
-                and class = NEW.class and type = NEW.type;
-      END IF;
-
-      RETURN NULL;
-    END IF;
-
-    RETURN NEW;
-  END IF;
-
   -- ---- All other place types.
 
   -- When an area is changed from large to small: log and discard change
@@ -269,17 +241,6 @@ BEGIN
       WHERE osm_type = NEW.osm_type and osm_id = NEW.osm_id
             and class = NEW.class and type = NEW.type;
 
-    -- Postcode areas are only kept, when there is an actual postcode assigned.
-    IF NEW.class = 'boundary' AND NEW.type = 'postal_code' THEN
-      IF NEW.address is NULL OR NOT NEW.address ? 'postcode' THEN
-        -- postcode was deleted, no longer retain in placex
-        DELETE FROM placex where place_id = existingplacex.place_id;
-        RETURN NULL;
-      END IF;
-
-      NEW.name := hstore('ref', NEW.address->'postcode');
-    END IF;
-
     -- Boundaries must be areas.
     IF NEW.class in ('boundary')
        AND ST_GeometryType(NEW.geometry) not in ('ST_Polygon','ST_MultiPolygon')

lib-sql/functions/placex_triggers.sql

@@ -2,7 +2,7 @@
 --
 -- This file is part of Nominatim. (https://nominatim.org)
 --
--- Copyright (C) 2025 by the Nominatim developer community.
+-- Copyright (C) 2026 by the Nominatim developer community.
 -- For a full list of authors see the git log.
 
 -- Trigger functions for the placex table.
@@ -304,7 +304,6 @@ DECLARE
 BEGIN
   IF bnd.rank_search >= 26 or bnd.rank_address = 0
      or ST_GeometryType(bnd.geometry) NOT IN ('ST_Polygon','ST_MultiPolygon')
-     or bnd.type IN ('postcode', 'postal_code')
   THEN
     RETURN NULL;
   END IF;
@@ -341,26 +340,6 @@ BEGIN
     END IF;
   END IF;
 
-  -- If extratags has a place tag, look for linked nodes by their place type.
-  -- Area and node still have to have the same name.
-  IF bnd.extratags ? 'place' and bnd.extratags->'place' != 'postcode'
-     and bnd_name is not null
-  THEN
-    FOR linked_placex IN
-      SELECT * FROM placex
-      WHERE (position(lower(name->'name') in bnd_name) > 0
-             OR position(bnd_name in lower(name->'name')) > 0)
-        AND placex.class = 'place' AND placex.type = bnd.extratags->'place'
-        AND placex.osm_type = 'N'
-        AND (placex.linked_place_id is null or placex.linked_place_id = bnd.place_id)
-        AND placex.rank_search < 26 -- needed to select the right index
-        AND ST_Covers(bnd.geometry, placex.geometry)
-    LOOP
-      {% if debug %}RAISE WARNING 'Found type-matching place node %', linked_placex.osm_id;{% endif %}
-      RETURN linked_placex;
-    END LOOP;
-  END IF;
-
   IF bnd.extratags ? 'wikidata' THEN
     FOR linked_placex IN
       SELECT * FROM placex
@@ -377,6 +356,25 @@ BEGIN
     END LOOP;
   END IF;
 
+  -- If extratags has a place tag, look for linked nodes by their place type.
+  -- Area and node still have to have the same name.
+  IF bnd.extratags ? 'place' and bnd_name is not null
+  THEN
+    FOR linked_placex IN
+      SELECT * FROM placex
+      WHERE (position(lower(name->'name') in bnd_name) > 0
+             OR position(bnd_name in lower(name->'name')) > 0)
+        AND placex.class = 'place' AND placex.type = bnd.extratags->'place'
+        AND placex.osm_type = 'N'
+        AND (placex.linked_place_id is null or placex.linked_place_id = bnd.place_id)
+        AND placex.rank_search < 26 -- needed to select the right index
+        AND ST_Covers(bnd.geometry, placex.geometry)
+    LOOP
+      {% if debug %}RAISE WARNING 'Found type-matching place node %', linked_placex.osm_id;{% endif %}
+      RETURN linked_placex;
+    END LOOP;
+  END IF;
+
   -- Name searches can be done for ways as well as relations
   IF bnd_name is not null THEN
     {% if debug %}RAISE WARNING 'Looking for nodes with matching names';{% endif %}
@@ -393,7 +391,6 @@ BEGIN
         AND placex.class = 'place'
         AND (placex.linked_place_id is null or placex.linked_place_id = bnd.place_id)
         AND placex.rank_search < 26 -- needed to select the right index
-        AND placex.type != 'postcode'
         AND ST_Covers(bnd.geometry, placex.geometry)
     LOOP
       {% if debug %}RAISE WARNING 'Found matching place node %', linked_placex.osm_id;{% endif %}
@@ -468,7 +465,7 @@ BEGIN
     END IF;
   END LOOP;
 
-  name_vector := token_get_name_search_tokens(token_info);
+  name_vector := COALESCE(token_get_name_search_tokens(token_info), '{}'::INTEGER[]);
 
   -- Check if the parent covers all address terms.
   -- If not, create a search name entry with the house number as the name.
@@ -675,7 +672,7 @@ CREATE OR REPLACE FUNCTION placex_insert()
   AS $$
 DECLARE
   postcode TEXT;
-  result BOOLEAN;
+  result INT;
   is_area BOOLEAN;
   country_code VARCHAR(2);
   diameter FLOAT;
@@ -697,17 +694,7 @@ BEGIN
   ELSE
     is_area := ST_GeometryType(NEW.geometry) IN ('ST_Polygon','ST_MultiPolygon');
 
-    IF NEW.class in ('place','boundary')
-       AND NEW.type in ('postcode','postal_code')
-    THEN
-      IF NEW.address IS NULL OR NOT NEW.address ? 'postcode' THEN
-          -- most likely just a part of a multipolygon postcode boundary, throw it away
-          RETURN NULL;
-      END IF;
-
-      NEW.name := hstore('ref', NEW.address->'postcode');
-
-    ELSEIF NEW.class = 'highway' AND is_area AND NEW.name is null
+    IF NEW.class = 'highway' AND is_area AND NEW.name is null
            AND NEW.extratags ? 'area' AND NEW.extratags->'area' = 'yes'
     THEN
         RETURN NULL;
@@ -790,11 +777,12 @@ BEGIN
 
 
    -- add to tables for special search
-   -- Note: won't work on initial import because the classtype tables
-   -- do not yet exist. It won't hurt either.
   classtable := 'place_classtype_' || NEW.class || '_' || NEW.type;
-  SELECT count(*)>0 FROM pg_tables WHERE tablename = classtable and schemaname = current_schema() INTO result;
-  IF result THEN
+  SELECT count(*) INTO result
+    FROM pg_tables
+    WHERE classtable NOT SIMILAR TO '%\W%'
+          AND tablename = classtable and schemaname = current_schema();
+  IF result > 0 THEN
     EXECUTE 'INSERT INTO ' || classtable::regclass || ' (place_id, centroid) VALUES ($1,$2)' 
     USING NEW.place_id, NEW.centroid;
   END IF;
@@ -853,13 +841,15 @@ BEGIN
 
   NEW.indexed_date = now();
 
-  {% if 'search_name' in db.tables %}
-    DELETE from search_name WHERE place_id = NEW.place_id;
-  {% endif %}
-  result := deleteSearchName(NEW.partition, NEW.place_id);
-  DELETE FROM place_addressline WHERE place_id = NEW.place_id;
-  result := deleteRoad(NEW.partition, NEW.place_id);
-  result := deleteLocationArea(NEW.partition, NEW.place_id, NEW.rank_search);
+  IF OLD.indexed_status > 1 THEN
+    {% if 'search_name' in db.tables %}
+      DELETE from search_name WHERE place_id = NEW.place_id;
+    {% endif %}
+    result := deleteSearchName(NEW.partition, NEW.place_id);
+    DELETE FROM place_addressline WHERE place_id = NEW.place_id;
+    result := deleteRoad(NEW.partition, NEW.place_id);
+    result := deleteLocationArea(NEW.partition, NEW.place_id, NEW.rank_search);
+  END IF;
 
   NEW.extratags := NEW.extratags - 'linked_place'::TEXT;
   IF NEW.extratags = ''::hstore THEN
@@ -872,19 +862,12 @@ BEGIN
   NEW.linked_place_id := OLD.linked_place_id;
 
   -- Remove linkage, if we have computed a different new linkee.
-  UPDATE placex SET linked_place_id = null, indexed_status = 2
-    WHERE linked_place_id = NEW.place_id
-          and (linked_place is null or linked_place_id != linked_place);
-  -- update not necessary for osmline, cause linked_place_id does not exist
-
-  -- Postcodes are just here to compute the centroids. They are not searchable
-  -- unless they are a boundary=postal_code.
-  -- There was an error in the style so that boundary=postal_code used to be
-  -- imported as place=postcode. That's why relations are allowed to pass here.
-  -- This can go away in a couple of versions.
-  IF NEW.class = 'place'  and NEW.type = 'postcode' and NEW.osm_type != 'R' THEN
-    NEW.token_info := null;
-    RETURN NEW;
+  IF OLD.indexed_status > 1 THEN
+    UPDATE placex
+      SET linked_place_id = null,
+          indexed_status = CASE WHEN indexed_status = 0 THEN 2 ELSE indexed_status END
+      WHERE linked_place_id = NEW.place_id
+            and (linked_place is null or place_id != linked_place);
   END IF;
 
   -- Compute a preliminary centroid.
@@ -1055,7 +1038,9 @@ BEGIN
                 LOOP
                   UPDATE placex SET linked_place_id = NEW.place_id WHERE place_id = linked_node_id;
                   {% if 'search_name' in db.tables %}
-                    DELETE FROM search_name WHERE place_id = linked_node_id;
+                    IF OLD.indexed_status > 1 THEN
+                      DELETE FROM search_name WHERE place_id = linked_node_id;
+                    END IF;
                   {% endif %}
                 END LOOP;
               END IF;
@@ -1204,11 +1189,6 @@ BEGIN
     -- reset the address rank if necessary.
     UPDATE placex set linked_place_id = NEW.place_id, indexed_status = 2
       WHERE place_id = location.place_id;
-    -- ensure that those places are not found anymore
-    {% if 'search_name' in db.tables %}
-      DELETE FROM search_name WHERE place_id = location.place_id;
-    {% endif %}
-    PERFORM deleteLocationArea(NEW.partition, location.place_id, NEW.rank_search);
 
     SELECT wikipedia, importance
       FROM compute_importance(location.extratags, NEW.country_code,
@@ -1219,7 +1199,7 @@ BEGIN
     IF linked_importance is not null AND
        (NEW.importance is null or NEW.importance < linked_importance)
     THEN
-      NEW.importance = linked_importance;
+      NEW.importance := linked_importance;
     END IF;
   ELSE
     -- No linked place? As a last resort check if the boundary is tagged with
@@ -1261,7 +1241,7 @@ BEGIN
        LIMIT 1
     LOOP
       IF location.osm_id = NEW.osm_id THEN
-        {% if debug %}RAISE WARNING 'Updating names for country '%' with: %', NEW.country_code, NEW.name;{% endif %}
+        {% if debug %}RAISE WARNING 'Updating names for country ''%'' with: %', NEW.country_code, NEW.name;{% endif %}
         UPDATE country_name SET derived_name = NEW.name WHERE country_code = NEW.country_code;
       END IF;
     END LOOP;
@@ -1286,8 +1266,6 @@ BEGIN
     END IF;
   ELSEIF NEW.rank_address > 25 THEN
     max_rank := 25;
-  ELSEIF NEW.class in ('place','boundary') and NEW.type in ('postcode','postal_code') THEN
-    max_rank := NEW.rank_search;
   ELSE
     max_rank := NEW.rank_address;
   END IF;
@@ -1302,10 +1280,10 @@ BEGIN
   NEW.postcode := coalesce(token_get_postcode(NEW.token_info), NEW.postcode);
 
   -- if we have a name add this to the name search table
-  IF NEW.name IS NOT NULL THEN
+  name_vector := token_get_name_search_tokens(NEW.token_info);
+  IF array_length(name_vector, 1) is not NULL THEN
     -- Initialise the name vector using our name
     NEW.name := add_default_place_name(NEW.country_code, NEW.name);
-    name_vector := token_get_name_search_tokens(NEW.token_info);
 
     IF NEW.rank_search <= 25 and NEW.rank_address > 0 THEN
       result := add_location(NEW.place_id, NEW.country_code, NEW.partition,
@@ -1360,15 +1338,16 @@ CREATE OR REPLACE FUNCTION placex_delete()
   AS $$
 DECLARE
   b BOOLEAN;
+  result INT;
   classtable TEXT;
 BEGIN
   -- RAISE WARNING 'placex_delete % %',OLD.osm_type,OLD.osm_id;
 
   IF OLD.linked_place_id is null THEN
-    update placex set linked_place_id = null, indexed_status = 2 where linked_place_id = OLD.place_id and indexed_status = 0;
-    {% if debug %}RAISE WARNING 'placex_delete:01 % %',OLD.osm_type,OLD.osm_id;{% endif %}
-    update placex set linked_place_id = null where linked_place_id = OLD.place_id;
-    {% if debug %}RAISE WARNING 'placex_delete:02 % %',OLD.osm_type,OLD.osm_id;{% endif %}
+    UPDATE placex
+      SET linked_place_id = NULL,
+          indexed_status = CASE WHEN indexed_status = 0 THEN 2 ELSE indexed_status END
+      WHERE linked_place_id = OLD.place_id;
   ELSE
     update placex set indexed_status = 2 where place_id = OLD.linked_place_id and indexed_status = 0;
   END IF;
@@ -1392,6 +1371,7 @@ BEGIN
     -- reparenting also for OSM Interpolation Lines (and for Tiger?)
     update location_property_osmline set indexed_status = 2 where indexed_status = 0 and parent_place_id = OLD.place_id;
 
+    UPDATE location_postcodes SET indexed_status = 2 WHERE parent_place_id = OLD.place_id;
   END IF;
 
   {% if debug %}RAISE WARNING 'placex_delete:08 % %',OLD.osm_type,OLD.osm_id;{% endif %}
@@ -1417,15 +1397,16 @@ BEGIN
 
   -- remove from tables for special search
   classtable := 'place_classtype_' || OLD.class || '_' || OLD.type;
-  SELECT count(*)>0 FROM pg_tables WHERE tablename = classtable and schemaname = current_schema() INTO b;
-  IF b THEN
+  SELECT count(*) INTO result
+    FROM pg_tables
+    WHERE classtable NOT SIMILAR TO '%\W%'
+          AND tablename = classtable and schemaname = current_schema();
+
+  IF result > 0 THEN
     EXECUTE 'DELETE FROM ' || classtable::regclass || ' WHERE place_id = $1' USING OLD.place_id;
   END IF;
 
   {% if debug %}RAISE WARNING 'placex_delete:12 % %',OLD.osm_type,OLD.osm_id;{% endif %}
-
-  UPDATE location_postcode SET indexed_status = 2 WHERE parent_place_id = OLD.place_id;
-
   RETURN OLD;
 
 END;

lib-sql/functions/postcode_triggers.sql

@@ -2,10 +2,10 @@
 --
 -- This file is part of Nominatim. (https://nominatim.org)
 --
--- Copyright (C) 2022 by the Nominatim developer community.
+-- Copyright (C) 2025 by the Nominatim developer community.
 -- For a full list of authors see the git log.
 
--- Trigger functions for location_postcode table.
+-- Trigger functions for location_postcodes table.
 
 
 -- Trigger for updates of location_postcode
@@ -13,7 +13,7 @@
 -- Computes the parent object the postcode most likely refers to.
 -- This will be the place that determines the address displayed when
 -- searching for this postcode.
-CREATE OR REPLACE FUNCTION postcode_update()
+CREATE OR REPLACE FUNCTION postcodes_update()
   RETURNS TRIGGER
   AS $$
 DECLARE
@@ -28,13 +28,10 @@ BEGIN
 
     partition := get_partition(NEW.country_code);
 
-    SELECT * FROM get_postcode_rank(NEW.country_code, NEW.postcode)
-      INTO NEW.rank_search, NEW.rank_address;
-
     NEW.parent_place_id = 0;
     FOR location IN
       SELECT place_id
-        FROM getNearFeatures(partition, NEW.geometry, NEW.geometry, NEW.rank_search)
+        FROM getNearFeatures(partition, NEW.centroid, NEW.centroid, NEW.rank_search)
         WHERE NOT isguess ORDER BY rank_address DESC, distance asc LIMIT 1
     LOOP
         NEW.parent_place_id = location.place_id;
@@ -45,3 +42,89 @@ END;
 $$
 LANGUAGE plpgsql;
 
+
+CREATE OR REPLACE FUNCTION postcodes_delete()
+  RETURNS TRIGGER
+  AS $$
+BEGIN
+{% if not disable_diff_updates %}
+  UPDATE placex p SET indexed_status = 2
+   WHERE p.postcode = OLD.postcode AND ST_Intersects(OLD.geometry, p.geometry)
+         AND indexed_status = 0;
+{% endif %}
+
+  RETURN OLD;
+END;
+$$
+LANGUAGE plpgsql;
+
+
+CREATE OR REPLACE FUNCTION postcodes_insert()
+  RETURNS TRIGGER
+  AS $$
+DECLARE
+  existing RECORD;
+BEGIN
+  IF NEW.osm_id is not NULL THEN
+    -- postcode area, remove existing from same OSM object
+    SELECT * INTO existing FROM location_postcodes p
+      WHERE p.osm_id = NEW.osm_id;
+
+    IF existing.place_id is not NULL THEN
+      IF existing.postcode != NEW.postcode or existing.country_code != NEW.country_code THEN
+        DELETE FROM location_postcodes p WHERE p.osm_id = NEW.osm_id;
+        existing := NULL;
+      END IF;
+    END IF;
+  END IF;
+
+  IF existing is NULL THEN
+    SELECT * INTO existing FROM location_postcodes p
+      WHERE p.country_code = NEW.country_code AND p.postcode = NEW.postcode;
+
+    IF existing.postcode is NULL THEN
+{% if not disable_diff_updates %}
+      UPDATE placex p SET indexed_status = 2
+       WHERE ST_Intersects(NEW.geometry, p.geometry)
+             AND indexed_status = 0
+             AND p.rank_address >= 22 AND not p.address ? 'postcode';
+{% endif %}
+
+      -- new entry, just insert
+      NEW.indexed_status := 1;
+      NEW.place_id := nextval('seq_place');
+      RETURN NEW;
+    END IF;
+  END IF;
+
+  -- update: only when there are changes
+  IF coalesce(NEW.osm_id, -1) != coalesce(existing.osm_id, -1)
+     OR (NEW.osm_id is not null AND NEW.geometry::text != existing.geometry::text)
+     OR (NEW.osm_id is null
+         AND (abs(ST_X(existing.centroid) - ST_X(NEW.centroid)) > 0.0000001
+              OR abs(ST_Y(existing.centroid) - ST_Y(NEW.centroid)) > 0.0000001))
+  THEN
+{% if not disable_diff_updates %}
+    UPDATE placex p SET indexed_status = 2
+      WHERE ST_Intersects(ST_Difference(NEW.geometry, existing.geometry), p.geometry)
+            AND indexed_status = 0
+            AND p.rank_address >= 22 AND not p.address ? 'postcode';
+
+    UPDATE placex p SET indexed_status = 2
+      WHERE ST_Intersects(ST_Difference(existing.geometry, NEW.geometry), p.geometry)
+            AND indexed_status = 0
+            AND p.postcode = OLD.postcode;
+{% endif %}
+
+    UPDATE location_postcodes p
+      SET osm_id = NEW.osm_id,
+          indexed_status = 2,
+          centroid = NEW.centroid,
+          geometry = NEW.geometry
+      WHERE p.country_code = NEW.country_code AND p.postcode = NEW.postcode;
+  END IF;
+
+  RETURN NULL;
+END;
+$$
+LANGUAGE plpgsql;

lib-sql/functions/ranking.sql

@@ -2,7 +2,7 @@
 --
 -- This file is part of Nominatim. (https://nominatim.org)
 --
--- Copyright (C) 2022 by the Nominatim developer community.
+-- Copyright (C) 2025 by the Nominatim developer community.
 -- For a full list of authors see the git log.
 
 -- Functions related to search and address ranks
@@ -114,66 +114,6 @@ $$
 LANGUAGE plpgsql IMMUTABLE PARALLEL SAFE;
 
 
--- Guess a ranking for postcodes from country and postcode format.
-CREATE OR REPLACE FUNCTION get_postcode_rank(country_code VARCHAR(2), postcode TEXT,
-                                             OUT rank_search SMALLINT,
-                                             OUT rank_address SMALLINT)
-AS $$
-DECLARE
-  part TEXT;
-BEGIN
-    rank_search := 30;
-    rank_address := 30;
-    postcode := upper(postcode);
-
-    IF country_code = 'gb' THEN
-        IF postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z]? [0-9][A-Z][A-Z])$' THEN
-            rank_search := 25;
-            rank_address := 5;
-        ELSEIF postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z]? [0-9])$' THEN
-            rank_search := 23;
-            rank_address := 5;
-        ELSEIF postcode ~ '^([A-Z][A-Z]?[0-9][0-9A-Z])$' THEN
-            rank_search := 21;
-            rank_address := 5;
-        END IF;
-
-    ELSEIF country_code = 'sg' THEN
-        IF postcode ~ '^([0-9]{6})$' THEN
-            rank_search := 25;
-            rank_address := 11;
-        END IF;
-
-    ELSEIF country_code = 'de' THEN
-        IF postcode ~ '^([0-9]{5})$' THEN
-            rank_search := 21;
-            rank_address := 11;
-        END IF;
-
-    ELSE
-        -- Guess at the postcode format and coverage (!)
-        IF postcode ~ '^[A-Z0-9]{1,5}$' THEN -- Probably too short to be very local
-            rank_search := 21;
-            rank_address := 11;
-        ELSE
-            -- Does it look splitable into and area and local code?
-            part := substring(postcode from '^([- :A-Z0-9]+)([- :][A-Z0-9]+)$');
-
-            IF part IS NOT NULL THEN
-                rank_search := 25;
-                rank_address := 11;
-            ELSEIF postcode ~ '^[- :A-Z0-9]{6,}$' THEN
-                rank_search := 21;
-                rank_address := 11;
-            END IF;
-        END IF;
-    END IF;
-
-END;
-$$
-LANGUAGE plpgsql IMMUTABLE PARALLEL SAFE;
-
-
 -- Get standard search and address rank for an object.
 --
 -- \param country        Two-letter country code where the object is in.
@@ -198,12 +138,7 @@ AS $$
 DECLARE
   classtype TEXT;
 BEGIN
-  IF place_class in ('place','boundary')
-     and place_type in ('postcode','postal_code')
-  THEN
-    SELECT * INTO search_rank, address_rank
-      FROM get_postcode_rank(country, postcode);
-  ELSEIF extended_type = 'N' AND place_class = 'highway' THEN
+  IF extended_type = 'N' AND place_class = 'highway' THEN
     search_rank = 30;
     address_rank = 30;
   ELSEIF place_class = 'landuse' AND extended_type != 'A' THEN

lib-sql/functions/utils.sql

@@ -2,7 +2,7 @@
 --
 -- This file is part of Nominatim. (https://nominatim.org)
 --
--- Copyright (C) 2025 by the Nominatim developer community.
+-- Copyright (C) 2026 by the Nominatim developer community.
 -- For a full list of authors see the git log.
 
 -- Assorted helper functions for the triggers.
@@ -46,13 +46,13 @@ DECLARE
   r INTEGER[];
 BEGIN
   IF array_upper(a, 1) IS NULL THEN
-    RETURN b;
+    RETURN COALESCE(b, '{}'::INTEGER[]);
   END IF;
   IF array_upper(b, 1) IS NULL THEN
-    RETURN a;
+    RETURN COALESCE(a, '{}'::INTEGER[]);
   END IF;
   r := a;
-  FOR i IN 1..array_upper(b, 1) LOOP  
+  FOR i IN 1..array_upper(b, 1) LOOP
     IF NOT (ARRAY[b[i]] <@ r) THEN
       r := r || b[i];
     END IF;
@@ -139,37 +139,46 @@ $$
 LANGUAGE plpgsql IMMUTABLE PARALLEL SAFE;
 
 
--- Find the nearest artificial postcode for the given geometry.
--- TODO For areas there should not be more than two inside the geometry.
+-- Find the best-matching postcode for the given geometry
 CREATE OR REPLACE FUNCTION get_nearest_postcode(country VARCHAR(2), geom GEOMETRY)
   RETURNS TEXT
   AS $$
 DECLARE
   outcode TEXT;
   cnt INTEGER;
+  location RECORD;
 BEGIN
     -- If the geometry is an area then only one postcode must be within
     -- that area, otherwise consider the area as not having a postcode.
     IF ST_GeometryType(geom) in ('ST_Polygon','ST_MultiPolygon') THEN
-        SELECT min(postcode), count(*) FROM
-              (SELECT postcode FROM location_postcode
-                WHERE ST_Contains(geom, location_postcode.geometry) LIMIT 2) sub
-          INTO outcode, cnt;
+      SELECT min(postcode), count(*) FROM
+        (SELECT postcode FROM location_postcodes
+           WHERE geom && location_postcodes.geometry -- want to use the index
+                 AND ST_Contains(geom, location_postcodes.centroid)
+                 AND country_code = country
+           LIMIT 2) sub
+        INTO outcode, cnt;
 
         IF cnt = 1 THEN
             RETURN outcode;
-        ELSE
-            RETURN null;
         END IF;
+
+        RETURN null;
     END IF;
 
-    SELECT postcode FROM location_postcode
-     WHERE ST_DWithin(geom, location_postcode.geometry, 0.05)
-          AND location_postcode.country_code = country
-     ORDER BY ST_Distance(geom, location_postcode.geometry) LIMIT 1
-    INTO outcode;
+    -- Otherwise: be fully within the coverage area of a postcode
+    FOR location IN
+      SELECT postcode
+        FROM location_postcodes p
+       WHERE ST_Covers(p.geometry, geom)
+             AND p.country_code = country
+       ORDER BY osm_id is null, ST_Distance(p.centroid, geom)
+       LIMIT 1
+    LOOP
+        RETURN location.postcode;
+    END LOOP;
 
-    RETURN outcode;
+    RETURN NULL;
 END;
 $$
 LANGUAGE plpgsql STABLE PARALLEL SAFE;
@@ -314,6 +323,17 @@ END;
 $$
 LANGUAGE plpgsql;
 
+
+-- Return the bounding box of the geometry buffered by the given number
+-- of meters.
+CREATE OR REPLACE FUNCTION expand_by_meters(geom GEOMETRY, meters FLOAT)
+  RETURNS GEOMETRY
+  AS $$
+    SELECT ST_Envelope(ST_Buffer(geom::geography, meters, 1)::geometry)
+$$
+LANGUAGE sql IMMUTABLE PARALLEL SAFE;
+
+
 -- Create a bounding box with an extent computed from the radius (in meters)
 -- which in turn is derived from the given search rank.
 CREATE OR REPLACE FUNCTION place_node_fuzzy_area(geom GEOMETRY, rank_search INTEGER)
@@ -332,9 +352,7 @@ BEGIN
     radius := 1000;
   END IF;
 
-  RETURN ST_Envelope(ST_Collect(
-                     ST_Project(geom::geography, radius, 0.785398)::geometry,
-                     ST_Project(geom::geography, radius, 3.9269908)::geometry));
+  RETURN expand_by_meters(geom, radius);
 END;
 $$
 LANGUAGE plpgsql IMMUTABLE PARALLEL SAFE;
@@ -350,8 +368,6 @@ CREATE OR REPLACE FUNCTION add_location(place_id BIGINT, country_code varchar(2)
 DECLARE
   postcode TEXT;
 BEGIN
-  PERFORM deleteLocationArea(partition, place_id, rank_search);
-
   -- add postcode only if it contains a single entry, i.e. ignore postcode lists
   postcode := NULL;
   IF in_postcode is not null AND in_postcode not similar to '%(,|;)%' THEN

lib-sql/grants.sql

@@ -0,0 +1,47 @@
+-- SPDX-License-Identifier: GPL-2.0-only
+--
+-- This file is part of Nominatim. (https://nominatim.org)
+--
+-- Copyright (C) 2026 by the Nominatim developer community.
+-- For a full list of authors see the git log.
+--
+-- Grant read-only access to the web user for all Nominatim tables.
+
+-- Core tables
+GRANT SELECT ON import_status TO "{{config.DATABASE_WEBUSER}}";
+GRANT SELECT ON country_name TO "{{config.DATABASE_WEBUSER}}";
+GRANT SELECT ON nominatim_properties TO "{{config.DATABASE_WEBUSER}}";
+
+-- Location tables
+GRANT SELECT ON location_property_tiger TO "{{config.DATABASE_WEBUSER}}";
+GRANT SELECT ON location_property_osmline TO "{{config.DATABASE_WEBUSER}}";
+GRANT SELECT ON location_postcodes TO "{{config.DATABASE_WEBUSER}}";
+
+-- Search tables
+{% if not db.reverse_only %}
+GRANT SELECT ON search_name TO "{{config.DATABASE_WEBUSER}}";
+{% endif %}
+
+-- Main place tables
+GRANT SELECT ON placex TO "{{config.DATABASE_WEBUSER}}";
+GRANT SELECT ON place_addressline TO "{{config.DATABASE_WEBUSER}}";
+GRANT SELECT ON placex_entrance TO "{{config.DATABASE_WEBUSER}}";
+
+-- Error/delete tracking tables
+GRANT SELECT ON import_polygon_error TO "{{config.DATABASE_WEBUSER}}";
+GRANT SELECT ON import_polygon_delete TO "{{config.DATABASE_WEBUSER}}";
+
+-- Country grid
+GRANT SELECT ON country_osm_grid TO "{{config.DATABASE_WEBUSER}}";
+
+-- Tokenizer tables (word table)
+{% if 'word' in db.tables %}
+GRANT SELECT ON word TO "{{config.DATABASE_WEBUSER}}";
+{% endif %}
+
+-- Special phrase tables
+{% for table in db.tables %}
+{% if table.startswith('place_classtype_') %}
+GRANT SELECT ON {{ table }} TO "{{config.DATABASE_WEBUSER}}";
+{% endif %}
+{% endfor %}

lib-sql/indices.sql

@@ -2,7 +2,7 @@
 --
 -- This file is part of Nominatim. (https://nominatim.org)
 --
--- Copyright (C) 2022 by the Nominatim developer community.
+-- Copyright (C) 2025 by the Nominatim developer community.
 -- For a full list of authors see the git log.
 
 -- Indices used only during search and update.
@@ -21,30 +21,25 @@ CREATE INDEX IF NOT EXISTS idx_placex_parent_place_id
   ON placex USING BTREE (parent_place_id) {{db.tablespace.search_index}}
   WHERE parent_place_id IS NOT NULL;
 ---
--- Used to find postcode areas after a search in location_postcode.
-CREATE INDEX IF NOT EXISTS idx_placex_postcode_areas
-  ON placex USING BTREE (country_code, postcode) {{db.tablespace.search_index}}
-  WHERE osm_type = 'R' AND class = 'boundary' AND type = 'postal_code';
----
 CREATE INDEX IF NOT EXISTS idx_placex_geometry ON placex
   USING GIST (geometry) {{db.tablespace.search_index}};
+---
 -- Index is needed during import but can be dropped as soon as a full
 -- geometry index is in place. The partial index is almost as big as the full
 -- index.
----
 DROP INDEX IF EXISTS idx_placex_geometry_lower_rank_ways;
 ---
 CREATE INDEX IF NOT EXISTS idx_placex_geometry_reverse_lookupPolygon
   ON placex USING gist (geometry) {{db.tablespace.search_index}}
   WHERE St_GeometryType(geometry) in ('ST_Polygon', 'ST_MultiPolygon')
-    AND rank_address between 4 and 25 AND type != 'postcode'
+    AND rank_address between 4 and 25
     AND name is not null AND indexed_status = 0 AND linked_place_id is null;
 ---
 -- used in reverse large area lookup
 CREATE INDEX IF NOT EXISTS idx_placex_geometry_reverse_lookupPlaceNode
   ON placex USING gist (ST_Buffer(geometry, reverse_place_diameter(rank_search)))
   {{db.tablespace.search_index}}
-  WHERE rank_address between 4 and 25 AND type != 'postcode'
+  WHERE rank_address between 4 and 25
     AND name is not null AND linked_place_id is null AND osm_type = 'N';
 ---
 CREATE INDEX IF NOT EXISTS idx_osmline_parent_place_id
@@ -53,9 +48,6 @@ CREATE INDEX IF NOT EXISTS idx_osmline_parent_place_id
 ---
 CREATE INDEX IF NOT EXISTS idx_osmline_parent_osm_id
   ON location_property_osmline USING BTREE (osm_id) {{db.tablespace.search_index}};
----
-CREATE INDEX IF NOT EXISTS idx_postcode_postcode
-  ON location_postcode USING BTREE (postcode) {{db.tablespace.search_index}};
 
 {% if drop %}
 ---
@@ -82,8 +74,8 @@ CREATE INDEX IF NOT EXISTS idx_postcode_postcode
     deferred BOOLEAN
    );
 ---
-  CREATE INDEX IF NOT EXISTS idx_location_postcode_parent_place_id
-    ON location_postcode USING BTREE (parent_place_id) {{db.tablespace.address_index}};
+  CREATE INDEX IF NOT EXISTS idx_location_postcodes_parent_place_id
+    ON location_postcodes USING BTREE (parent_place_id) {{db.tablespace.address_index}};
 {% endif %}
 
 -- Indices only needed for search.

lib-sql/partition-tables.src.sql

@@ -2,36 +2,48 @@
 --
 -- This file is part of Nominatim. (https://nominatim.org)
 --
--- Copyright (C) 2022 by the Nominatim developer community.
+-- Copyright (C) 2026 by the Nominatim developer community.
 -- For a full list of authors see the git log.
 
 drop table IF EXISTS search_name_blank CASCADE;
 CREATE TABLE search_name_blank (
-  place_id BIGINT,
-  address_rank smallint,
-  name_vector integer[],
-  centroid GEOMETRY(Geometry, 4326)
+  place_id BIGINT NOT NULL,
+  address_rank smallint NOT NULL,
+  name_vector integer[] NOT NULL,
+  centroid GEOMETRY(Geometry, 4326) NOT NULL
   );
 
-
 {% for partition in db.partitions %}
   CREATE TABLE location_area_large_{{ partition }} () INHERITS (location_area_large) {{db.tablespace.address_data}};
-  CREATE INDEX idx_location_area_large_{{ partition }}_place_id ON location_area_large_{{ partition }} USING BTREE (place_id) {{db.tablespace.address_index}};
-  CREATE INDEX idx_location_area_large_{{ partition }}_geometry ON location_area_large_{{ partition }} USING GIST (geometry) {{db.tablespace.address_index}};
+  CREATE INDEX idx_location_area_large_{{ partition }}_place_id
+    ON location_area_large_{{ partition }}
+    USING BTREE (place_id) {{db.tablespace.address_index}};
+  CREATE INDEX idx_location_area_large_{{ partition }}_geometry
+    ON location_area_large_{{ partition }}
+    USING GIST (geometry) {{db.tablespace.address_index}};
 
   CREATE TABLE search_name_{{ partition }} () INHERITS (search_name_blank) {{db.tablespace.address_data}};
-  CREATE INDEX idx_search_name_{{ partition }}_place_id ON search_name_{{ partition }} USING BTREE (place_id) {{db.tablespace.address_index}};
-  CREATE INDEX idx_search_name_{{ partition }}_centroid_street ON search_name_{{ partition }} USING GIST (centroid) {{db.tablespace.address_index}} where address_rank between 26 and 27;
-  CREATE INDEX idx_search_name_{{ partition }}_centroid_place ON search_name_{{ partition }} USING GIST (centroid) {{db.tablespace.address_index}} where address_rank between 2 and 25;
+  CREATE UNIQUE INDEX idx_search_name_{{ partition }}_place_id
+    ON search_name_{{ partition }}
+    USING BTREE (place_id) {{db.tablespace.address_index}};
+  CREATE INDEX idx_search_name_{{ partition }}_centroid_street
+    ON search_name_{{ partition }} USING GIST (centroid) {{db.tablespace.address_index}}
+    WHERE address_rank between 26 and 27;
+  CREATE INDEX idx_search_name_{{ partition }}_centroid_place
+    ON search_name_{{ partition }} USING GIST (centroid) {{db.tablespace.address_index}}
+    WHERE address_rank between 2 and 25;
 
   DROP TABLE IF EXISTS location_road_{{ partition }};
   CREATE TABLE location_road_{{ partition }} (
-    place_id BIGINT,
-    partition SMALLINT,
+    place_id BIGINT NOT NULL,
+    partition SMALLINT NOT NULL,
     country_code VARCHAR(2),
-    geometry GEOMETRY(Geometry, 4326)
+    geometry GEOMETRY(Geometry, 4326) NOT NULL
     ) {{db.tablespace.address_data}};
-  CREATE INDEX idx_location_road_{{ partition }}_geometry ON location_road_{{ partition }} USING GIST (geometry) {{db.tablespace.address_index}};
-  CREATE INDEX idx_location_road_{{ partition }}_place_id ON location_road_{{ partition }} USING BTREE (place_id) {{db.tablespace.address_index}};
-
+  CREATE INDEX idx_location_road_{{ partition }}_geometry
+    ON location_road_{{ partition }}
+    USING GIST (geometry) {{db.tablespace.address_index}};
+  CREATE UNIQUE INDEX idx_location_road_{{ partition }}_place_id
+    ON location_road_{{ partition }}
+    USING BTREE (place_id) {{db.tablespace.address_index}};
 {% endfor %}

lib-sql/table-triggers.sql

@@ -2,7 +2,7 @@
 --
 -- This file is part of Nominatim. (https://nominatim.org)
 --
--- Copyright (C) 2022 by the Nominatim developer community.
+-- Copyright (C) 2025 by the Nominatim developer community.
 -- For a full list of authors see the git log.
 
 -- insert creates the location tables, creates location indexes if indexed == true
@@ -25,5 +25,9 @@ CREATE TRIGGER place_before_delete BEFORE DELETE ON place
 CREATE TRIGGER place_before_insert BEFORE INSERT ON place
     FOR EACH ROW EXECUTE PROCEDURE place_insert();
 
-CREATE TRIGGER location_postcode_before_update BEFORE UPDATE ON location_postcode
-    FOR EACH ROW EXECUTE PROCEDURE postcode_update();
+CREATE TRIGGER location_postcode_before_update BEFORE UPDATE ON location_postcodes
+    FOR EACH ROW EXECUTE PROCEDURE postcodes_update();
+CREATE TRIGGER location_postcodes_before_delete BEFORE DELETE ON location_postcodes
+    FOR EACH ROW EXECUTE PROCEDURE postcodes_delete();
+CREATE TRIGGER location_postcodes_before_insert BEFORE INSERT ON location_postcodes
+    FOR EACH ROW EXECUTE PROCEDURE postcodes_insert();

lib-sql/tables.sql

@@ -2,7 +2,7 @@
 --
 -- This file is part of Nominatim. (https://nominatim.org)
 --
--- Copyright (C) 2022 by the Nominatim developer community.
+-- Copyright (C) 2026 by the Nominatim developer community.
 -- For a full list of authors see the git log.
 
 drop table if exists import_status;
@@ -11,7 +11,6 @@ CREATE TABLE import_status (
   sequence_id integer,
   indexed boolean
   );
-GRANT SELECT ON import_status TO "{{config.DATABASE_WEBUSER}}" ;
 
 drop table if exists import_osmosis_log;
 CREATE TABLE import_osmosis_log (
@@ -23,82 +22,60 @@ CREATE TABLE import_osmosis_log (
   event text
   );
 
-CREATE TABLE new_query_log (
-  type text,
-  starttime timestamp,
-  ipaddress text,
-  useragent text,
-  language text,
-  query text,
-  searchterm text,
-  endtime timestamp,
-  results integer,
-  format text,
-  secret text
-  );
-CREATE INDEX idx_new_query_log_starttime ON new_query_log USING BTREE (starttime);
-GRANT INSERT ON new_query_log TO "{{config.DATABASE_WEBUSER}}" ;
-GRANT UPDATE ON new_query_log TO "{{config.DATABASE_WEBUSER}}" ;
-GRANT SELECT ON new_query_log TO "{{config.DATABASE_WEBUSER}}" ;
-
-GRANT SELECT ON TABLE country_name TO "{{config.DATABASE_WEBUSER}}";
-
 DROP TABLE IF EXISTS nominatim_properties;
 CREATE TABLE nominatim_properties (
     property TEXT NOT NULL,
     value TEXT
 );
-GRANT SELECT ON TABLE nominatim_properties TO "{{config.DATABASE_WEBUSER}}";
 
 drop table IF EXISTS location_area CASCADE;
 CREATE TABLE location_area (
-  place_id BIGINT,
-  keywords INTEGER[],
-  partition SMALLINT,
+  place_id BIGINT NOT NULL,
+  keywords INTEGER[] NOT NULL,
+  partition SMALLINT NOT NULL,
   rank_search SMALLINT NOT NULL,
   rank_address SMALLINT NOT NULL,
   country_code VARCHAR(2),
-  isguess BOOL,
+  isguess BOOL NOT NULL,
   postcode TEXT,
-  centroid GEOMETRY(Point, 4326),
-  geometry GEOMETRY(Geometry, 4326)
+  centroid GEOMETRY(Point, 4326) NOT NULL,
+  geometry GEOMETRY(Geometry, 4326) NOT NULL
   );
 
 CREATE TABLE location_area_large () INHERITS (location_area);
 
 DROP TABLE IF EXISTS location_area_country;
 CREATE TABLE location_area_country (
-  place_id BIGINT,
-  country_code varchar(2),
-  geometry GEOMETRY(Geometry, 4326)
+  place_id BIGINT NOT NULL,
+  country_code varchar(2) NOT NULL,
+  geometry GEOMETRY(Geometry, 4326) NOT NULL
   ) {{db.tablespace.address_data}};
 CREATE INDEX idx_location_area_country_geometry ON location_area_country USING GIST (geometry) {{db.tablespace.address_index}};
 
 
 CREATE TABLE location_property_tiger (
-  place_id BIGINT,
+  place_id BIGINT NOT NULL,
   parent_place_id BIGINT,
-  startnumber INTEGER,
-  endnumber INTEGER,
-  step SMALLINT,
-  partition SMALLINT,
-  linegeo GEOMETRY,
+  startnumber INTEGER NOT NULL,
+  endnumber INTEGER NOT NULL,
+  step SMALLINT NOT NULL,
+  partition SMALLINT NOT NULL,
+  linegeo GEOMETRY NOT NULL,
   postcode TEXT);
-GRANT SELECT ON location_property_tiger TO "{{config.DATABASE_WEBUSER}}";
 
 drop table if exists location_property_osmline;
 CREATE TABLE location_property_osmline (
     place_id BIGINT NOT NULL,
-    osm_id BIGINT,
+    osm_id BIGINT NOT NULL,
     parent_place_id BIGINT,
-    geometry_sector INTEGER,
+    geometry_sector INTEGER NOT NULL,
     indexed_date TIMESTAMP,
     startnumber INTEGER,
     endnumber INTEGER,
     step SMALLINT,
-    partition SMALLINT,
-    indexed_status SMALLINT,
-    linegeo GEOMETRY,
+    partition SMALLINT NOT NULL,
+    indexed_status SMALLINT NOT NULL,
+    linegeo GEOMETRY NOT NULL,
     address HSTORE,
     token_info JSONB, -- custom column for tokenizer use only
     postcode TEXT,
@@ -108,32 +85,31 @@ CREATE UNIQUE INDEX idx_osmline_place_id ON location_property_osmline USING BTRE
 CREATE INDEX idx_osmline_geometry_sector ON location_property_osmline USING BTREE (geometry_sector) {{db.tablespace.address_index}};
 CREATE INDEX idx_osmline_linegeo ON location_property_osmline USING GIST (linegeo) {{db.tablespace.search_index}}
   WHERE startnumber is not null;
-GRANT SELECT ON location_property_osmline TO "{{config.DATABASE_WEBUSER}}";
 
 drop table IF EXISTS search_name;
 {% if not db.reverse_only %}
 CREATE TABLE search_name (
-  place_id BIGINT,
-  importance FLOAT,
-  search_rank SMALLINT,
-  address_rank SMALLINT,
-  name_vector integer[],
-  nameaddress_vector integer[],
+  place_id BIGINT NOT NULL,
+  importance FLOAT NOT NULL,
+  search_rank SMALLINT NOT NULL,
+  address_rank SMALLINT NOT NULL,
+  name_vector integer[] NOT NULL,
+  nameaddress_vector integer[] NOT NULL,
   country_code varchar(2),
-  centroid GEOMETRY(Geometry, 4326)
+  centroid GEOMETRY(Geometry, 4326) NOT NULL
   ) {{db.tablespace.search_data}};
-CREATE INDEX idx_search_name_place_id ON search_name USING BTREE (place_id) {{db.tablespace.search_index}};
-GRANT SELECT ON search_name to "{{config.DATABASE_WEBUSER}}" ;
+CREATE UNIQUE INDEX idx_search_name_place_id
+  ON search_name USING BTREE (place_id) {{db.tablespace.search_index}};
 {% endif %}
 
 drop table IF EXISTS place_addressline;
 CREATE TABLE place_addressline (
-  place_id BIGINT,
-  address_place_id BIGINT,
-  distance FLOAT,
-  cached_rank_address SMALLINT,
-  fromarea boolean,
-  isaddress boolean
+  place_id BIGINT NOT NULL,
+  address_place_id BIGINT NOT NULL,
+  distance FLOAT NOT NULL,
+  cached_rank_address SMALLINT NOT NULL,
+  fromarea boolean NOT NULL,
+  isaddress boolean NOT NULL
   ) {{db.tablespace.search_data}};
 CREATE INDEX idx_place_addressline_place_id on place_addressline USING BTREE (place_id) {{db.tablespace.search_index}};
 
@@ -146,18 +122,18 @@ CREATE TABLE placex (
   linked_place_id BIGINT,
   importance FLOAT,
   indexed_date TIMESTAMP,
-  geometry_sector INTEGER,
-  rank_address SMALLINT,
-  rank_search SMALLINT,
-  partition SMALLINT,
-  indexed_status SMALLINT,
+  geometry_sector INTEGER NOT NULL,
+  rank_address SMALLINT NOT NULL,
+  rank_search SMALLINT NOT NULL,
+  partition SMALLINT NOT NULL,
+  indexed_status SMALLINT NOT NULL,
   LIKE place INCLUDING CONSTRAINTS,
   wikipedia TEXT, -- calculated wikipedia article name (language:title)
   token_info JSONB, -- custom column for tokenizer use only
   country_code varchar(2),
   housenumber TEXT,
   postcode TEXT,
-  centroid GEOMETRY(Geometry, 4326)
+  centroid GEOMETRY(Geometry, 4326) NOT NULL
   ) {{db.tablespace.search_data}};
 
 CREATE UNIQUE INDEX idx_place_id ON placex USING BTREE (place_id) {{db.tablespace.search_index}};
@@ -192,8 +168,7 @@ CREATE INDEX idx_placex_geometry_buildings ON placex
 --        - linking of place nodes with same type to boundaries
 CREATE INDEX idx_placex_geometry_placenode ON placex
   USING SPGIST (geometry) {{db.tablespace.address_index}}
-  WHERE osm_type = 'N' and rank_search < 26
-        and class = 'place' and type != 'postcode';
+  WHERE osm_type = 'N' and rank_search < 26 and class = 'place';
 
 -- Usage: - is node part of a way?
 --        - find parent of interpolation spatially
@@ -221,28 +196,30 @@ CREATE INDEX idx_placex_rank_boundaries_sector ON placex
 
 DROP SEQUENCE IF EXISTS seq_place;
 CREATE SEQUENCE seq_place start 1;
-GRANT SELECT on placex to "{{config.DATABASE_WEBUSER}}" ;
-GRANT SELECT on place_addressline to "{{config.DATABASE_WEBUSER}}" ;
-GRANT SELECT ON planet_osm_ways to "{{config.DATABASE_WEBUSER}}" ;
-GRANT SELECT ON planet_osm_rels to "{{config.DATABASE_WEBUSER}}" ;
-GRANT SELECT on location_area to "{{config.DATABASE_WEBUSER}}" ;
 
 -- Table for synthetic postcodes.
-DROP TABLE IF EXISTS location_postcode;
-CREATE TABLE location_postcode (
-  place_id BIGINT,
+DROP TABLE IF EXISTS location_postcodes;
+CREATE TABLE location_postcodes (
+  place_id BIGINT NOT NULL,
   parent_place_id BIGINT,
-  rank_search SMALLINT,
-  rank_address SMALLINT,
-  indexed_status SMALLINT,
+  osm_id BIGINT,
+  rank_search SMALLINT NOT NULL,
+  indexed_status SMALLINT NOT NULL,
   indexed_date TIMESTAMP,
-  country_code varchar(2),
-  postcode TEXT,
-  geometry GEOMETRY(Geometry, 4326)
+  country_code varchar(2) NOT NULL,
+  postcode TEXT NOT NULL,
+  centroid GEOMETRY(Geometry, 4326) NOT NULL,
+  geometry GEOMETRY(Geometry, 4326) NOT NULL
   );
-CREATE UNIQUE INDEX idx_postcode_id ON location_postcode USING BTREE (place_id) {{db.tablespace.search_index}};
-CREATE INDEX idx_postcode_geometry ON location_postcode USING GIST (geometry) {{db.tablespace.address_index}};
-GRANT SELECT ON location_postcode TO "{{config.DATABASE_WEBUSER}}" ;
+CREATE UNIQUE INDEX idx_location_postcodes_id ON location_postcodes
+  USING BTREE (place_id) {{db.tablespace.search_index}};
+CREATE INDEX idx_location_postcodes_geometry ON location_postcodes
+  USING GIST (geometry) {{db.tablespace.search_index}};
+CREATE INDEX IF NOT EXISTS idx_location_postcodes_postcode
+  ON location_postcodes USING BTREE (postcode, country_code)
+  {{db.tablespace.search_index}};
+CREATE INDEX IF NOT EXISTS idx_location_postcodes_osmid
+  ON location_postcodes USING BTREE (osm_id) {{db.tablespace.search_index}};
 
 -- Table to store location of entrance nodes
 DROP TABLE IF EXISTS placex_entrance;
@@ -255,7 +232,6 @@ CREATE TABLE placex_entrance (
   );
 CREATE UNIQUE INDEX idx_placex_entrance_place_id_osm_id ON placex_entrance
   USING BTREE (place_id, osm_id) {{db.tablespace.search_index}};
-GRANT SELECT ON placex_entrance TO "{{config.DATABASE_WEBUSER}}" ;
 
 -- Create an index on the place table for lookups to populate the entrance
 -- table
@@ -277,7 +253,6 @@ CREATE TABLE import_polygon_error (
   newgeometry GEOMETRY(Geometry, 4326)
   );
 CREATE INDEX idx_import_polygon_error_osmid ON import_polygon_error USING BTREE (osm_type, osm_id);
-GRANT SELECT ON import_polygon_error TO "{{config.DATABASE_WEBUSER}}";
 
 DROP TABLE IF EXISTS import_polygon_delete;
 CREATE TABLE import_polygon_delete (
@@ -287,7 +262,6 @@ CREATE TABLE import_polygon_delete (
   type TEXT NOT NULL
   );
 CREATE INDEX idx_import_polygon_delete_osmid ON import_polygon_delete USING BTREE (osm_type, osm_id);
-GRANT SELECT ON import_polygon_delete TO "{{config.DATABASE_WEBUSER}}";
 
 DROP SEQUENCE IF EXISTS file;
 CREATE SEQUENCE file start 1;
@@ -318,5 +292,3 @@ CREATE INDEX planet_osm_rels_relation_members_idx ON planet_osm_rels USING gin(p
 CREATE INDEX IF NOT EXISTS idx_place_interpolations
     ON place USING gist(geometry) {{db.tablespace.address_index}}
     WHERE osm_type = 'W' and address ? 'interpolation';
-
-GRANT SELECT ON table country_osm_grid to "{{config.DATABASE_WEBUSER}}";

lib-sql/tiger_import_start.sql

@@ -15,6 +15,99 @@ CREATE TABLE location_property_tiger_import (
   step SMALLINT,
   postcode TEXT);
 
+
+-- Lookup functions for tiger import when update 
+-- tables are dropped (see gh-issue #2463)
+CREATE OR REPLACE FUNCTION getNearestNamedRoadPlaceIdSlow(in_centroid GEOMETRY,
+                                                      in_token_info JSONB)
+  RETURNS BIGINT
+  AS $$
+DECLARE
+  out_place_id BIGINT;
+
+BEGIN
+  SELECT place_id INTO out_place_id 
+    FROM search_name
+    WHERE 
+        -- finds rows where name_vector shares elements with search tokens.
+        token_matches_street(in_token_info, name_vector)
+        -- limits search area
+        AND centroid && ST_Expand(in_centroid, 0.015) 
+        AND address_rank BETWEEN 26 AND 27
+    ORDER BY ST_Distance(centroid, in_centroid) ASC 
+    LIMIT 1;
+
+  RETURN out_place_id;
+END
+$$
+LANGUAGE plpgsql;
+
+
+CREATE OR REPLACE FUNCTION getNearestParallelRoadFeatureSlow(line GEOMETRY)
+  RETURNS BIGINT
+  AS $$
+DECLARE
+  r RECORD;
+  search_diameter FLOAT;
+  p1 GEOMETRY;
+  p2 GEOMETRY;
+  p3 GEOMETRY;
+
+BEGIN
+  IF ST_GeometryType(line) not in ('ST_LineString') THEN
+    RETURN NULL;
+  END IF;
+
+  p1 := ST_LineInterpolatePoint(line,0);
+  p2 := ST_LineInterpolatePoint(line,0.5);
+  p3 := ST_LineInterpolatePoint(line,1);
+
+    search_diameter := 0.0005;
+    WHILE search_diameter < 0.01 LOOP
+      FOR r IN
+        SELECT place_id FROM placex
+          WHERE ST_DWithin(line, geometry, search_diameter)
+          AND rank_address BETWEEN 26 AND 27
+          ORDER BY (ST_distance(geometry, p1)+
+                    ST_distance(geometry, p2)+
+                    ST_distance(geometry, p3)) ASC limit 1
+      LOOP
+        RETURN r.place_id;
+      END LOOP;
+      search_diameter := search_diameter * 2;
+    END LOOP;
+    RETURN NULL;
+END
+$$
+LANGUAGE plpgsql;
+
+
+CREATE OR REPLACE FUNCTION getNearestRoadPlaceIdSlow(point GEOMETRY)
+  RETURNS BIGINT
+  AS $$
+DECLARE
+  r RECORD;
+  search_diameter FLOAT;
+BEGIN
+    search_diameter := 0.00005;
+    WHILE search_diameter < 0.1 LOOP
+      FOR r IN
+        SELECT place_id FROM placex
+          WHERE ST_DWithin(geometry, point, search_diameter)
+          AND rank_address BETWEEN 26 AND 27
+          ORDER BY ST_Distance(geometry, point) ASC limit 1
+      LOOP
+        RETURN r.place_id;
+      END LOOP;
+      search_diameter := search_diameter * 2;
+    END LOOP;
+    RETURN NULL;
+END
+$$
+LANGUAGE plpgsql;
+
+
+-- Tiger import function
 CREATE OR REPLACE FUNCTION tiger_line_import(linegeo GEOMETRY, in_startnumber INTEGER,
                                              in_endnumber INTEGER, interpolationtype TEXT,
                                              token_info JSONB, in_postcode TEXT) RETURNS INTEGER
@@ -71,28 +164,51 @@ BEGIN
   place_centroid := ST_Centroid(linegeo);
   out_partition := get_partition('us');
 
-  out_parent_place_id := getNearestNamedRoadPlaceId(out_partition, place_centroid,
+  -- HYBRID LOOKUP LOGIC (see gh-issue #2463)
+  -- if partition tables exist, use them for fast spatial lookups
+  {% if 'location_road_0' in db.tables %}
+    out_parent_place_id := getNearestNamedRoadPlaceId(out_partition, place_centroid,
                                                     token_info);
 
-  IF out_parent_place_id IS NULL THEN
-    SELECT getNearestParallelRoadFeature(out_partition, linegeo)
-      INTO out_parent_place_id;
+    IF out_parent_place_id IS NULL THEN
+      SELECT getNearestParallelRoadFeature(out_partition, linegeo)
+        INTO out_parent_place_id;
+    END IF;
+
+    IF out_parent_place_id IS NULL THEN
+      SELECT getNearestRoadPlaceId(out_partition, place_centroid)
+        INTO out_parent_place_id;
+    END IF;
+
+  -- When updatable information has been dropped:
+  -- Partition tables no longer exist, but search_name still persists.
+  {% elif 'search_name' in db.tables %}
+    -- Fallback: Look up in 'search_name' table 
+    -- though spatial lookups here can be slower.
+    out_parent_place_id := getNearestNamedRoadPlaceIdSlow(place_centroid, token_info);
+
+    IF out_parent_place_id IS NULL THEN
+      out_parent_place_id := getNearestParallelRoadFeatureSlow(linegeo);
+    END IF;
+
+    IF out_parent_place_id IS NULL THEN
+      out_parent_place_id := getNearestRoadPlaceIdSlow(place_centroid);
+    END IF;
+  {% endif %}
+
+  -- If parent was found, insert street(line) into import table
+  IF out_parent_place_id IS NOT NULL THEN
+    INSERT INTO location_property_tiger_import (linegeo, place_id, partition,
+                                                parent_place_id, startnumber, endnumber,
+                                                step, postcode)
+    VALUES (linegeo, nextval('seq_place'), out_partition,
+            out_parent_place_id, startnumber, endnumber,
+            stepsize, in_postcode);
+
+    RETURN 1;
   END IF;
+  RETURN 0;
 
-  IF out_parent_place_id IS NULL THEN
-    SELECT getNearestRoadPlaceId(out_partition, place_centroid)
-      INTO out_parent_place_id;
-  END IF;
-
---insert street(line) into import table
-insert into location_property_tiger_import (linegeo, place_id, partition,
-                                            parent_place_id, startnumber, endnumber,
-                                            step, postcode)
-values (linegeo, nextval('seq_place'), out_partition,
-        out_parent_place_id, startnumber, endnumber,
-        stepsize, in_postcode);
-
-  RETURN 1;
 END;
 $$
 LANGUAGE plpgsql;

packaging/nominatim-api/README.md

@@ -23,7 +23,7 @@ an ASGI-capable server like uvicorn. To install them from pypi run:
 You need to have a Nominatim database imported with the 'nominatim-db'
 package. Go to the project directory, then run uvicorn as:
 
-    uvicorn --factory nominatim.server.falcon.server:run_wsgi
+    uvicorn --factory nominatim_api.server.falcon.server:run_wsgi
 
 ## Documentation
 

packaging/nominatim-db/pyproject.toml

@@ -15,12 +15,13 @@ classifiers = [
     "Operating System :: OS Independent",
 ]
 dependencies = [
-    "psycopg",
+    "psycopg != 3.3.0",
     "python-dotenv",
     "jinja2",
     "pyYAML>=5.1",
     "psutil",
-    "PyICU"
+    "PyICU",
+    "mwparserfromhell"
 ]
 dynamic = ["version"]
 

settings/country-names/ps.yaml

@@ -1,3 +1,4 @@
-name: 
-    default: Palestinian Territory
+name:
+    default: الأراضي الفلسطينية
+    en: Palestinian Territories
     "no": Det palestinske området

settings/country_settings.yaml

@@ -305,6 +305,7 @@ ch:
     names: !include country-names/ch.yaml
     postcode:
       pattern: "dddd"
+      extent: 3000
 
 
 # Côte d'Ivoire (Côte d’Ivoire)
@@ -472,6 +473,7 @@ ee:
     names: !include country-names/ee.yaml
     postcode:
       pattern: "ddddd"
+      extent: 3000
 
 
 # Egypt (مصر)
@@ -585,6 +587,7 @@ gb:
     postcode:
       pattern: "(l?ld[A-Z0-9]?) ?(dll)"
       output: \1 \2
+      extent: 700
 
 
 # Grenada (Grenada)
@@ -612,6 +615,7 @@ gg:
     postcode:
       pattern: "(GYdd?) ?(dll)"
       output: \1 \2
+      extent: 1000
 
 
 # Ghana (Ghana)
@@ -766,6 +770,7 @@ ie:
     postcode:
       pattern: "(ldd) ?([0123456789ACDEFHKNPRTVWXY]{4})"
       output: \1 \2
+      extent: 50
 
 
 # Israel (ישראל)
@@ -785,6 +790,7 @@ im:
     postcode:
       pattern: "(IMdd?) ?(dll)"
       output: \1 \2
+      extent: 700
 
 
 # India (India)
@@ -879,6 +885,7 @@ jp:
     postcode:
       pattern: "(ddd)-?(dddd)"
       output: \1-\2
+      extent: 3000
 
 
 # Kenya (Kenya)
@@ -1013,6 +1020,7 @@ li:
     names: !include country-names/li.yaml
     postcode:
       pattern: "dddd"
+      extent: 4000
 
 
 # Sri Lanka (ශ්‍රී ලංකාව இலங்கை)
@@ -1058,6 +1066,7 @@ lu:
     names: !include country-names/lu.yaml
     postcode:
       pattern: "dddd"
+      extent: 1000
 
 
 # Latvia (Latvija)
@@ -1290,6 +1299,7 @@ nl:
     postcode:
       pattern: "(dddd) ?(ll)"
       output: \1 \2
+      extent: 800
 
 
 # Norway (Norge)
@@ -1425,6 +1435,7 @@ pt:
     names: !include country-names/pt.yaml
     postcode:
       pattern: "dddd(?:-ddd)?"
+      extent: 1000
 
 
 # Palau (Belau)
@@ -1460,6 +1471,7 @@ ro:
     names: !include country-names/ro.yaml
     postcode:
       pattern: "dddddd"
+      extent: 2500
 
 
 # Serbia (Србија)

src/nominatim_api/localization.py

@@ -8,7 +8,6 @@
 Helper functions for localizing names of results.
 """
 from typing import Mapping, List, Optional
-from .config import Configuration
 from .results import AddressLines, BaseResultT
 
 import re
@@ -18,15 +17,15 @@ class Locales:
     """ Helper class for localization of names.
 
         It takes a list of language prefixes in their order of preferred
-        usage.
+        usage and comma separated name keys (Configuration.OUTPUT_NAMES).
     """
 
-    def __init__(self, langs: Optional[List[str]] = None):
-        self.config = Configuration(None)
+    def __init__(self, langs: Optional[List[str]] = None,
+                 names: str = 'name:XX,name') -> None:
         self.languages = langs or []
         self.name_tags: List[str] = []
 
-        parts = self.config.OUTPUT_NAMES.split(',')
+        parts = names.split(',') if names else []
 
         for part in parts:
             part = part.strip()
@@ -68,7 +67,7 @@ class Locales:
         return next(iter(names.values()))
 
     @staticmethod
-    def from_accept_languages(langstr: str) -> 'Locales':
+    def from_accept_languages(langstr: str, names: str = 'name:XX,name') -> 'Locales':
         """ Create a localization object from a language list in the
             format of HTTP accept-languages header.
 
@@ -96,7 +95,7 @@ class Locales:
             if len(parts) > 1 and all(c[0] != parts[0] for c in candidates):
                 languages.append(parts[0])
 
-        return Locales(languages)
+        return Locales(languages, names)
 
     def localize(self, lines: AddressLines) -> None:
         """ Sets the local name of address parts according to the chosen

src/nominatim_api/logging.py

@@ -288,7 +288,7 @@ class TextLogger(BaseLogger):
             self._write(f"rank={res.rank_address}, ")
             self._write(f"osm={''.join(map(str, res.osm_object or []))}, ")
             self._write(f'cc={res.country_code}, ')
-            self._write(f'importance={res.importance or -1:.5f})\n')
+            self._write(f'importance={res.importance or float("NaN"):.5f})\n')
             total += 1
         self._write(f'TOTAL: {total}\n\n')
 

src/nominatim_api/lookup.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Implementation of place lookup by ID (doing many places at once).
@@ -291,12 +291,30 @@ async def find_in_postcode(conn: SearchConnection, collector: Collector) -> bool
                     .table_valued(sa.column('value', type_=sa.JSON))
         t = conn.t.postcode
         sql = sa.select(pid_tab.c.value['i'].as_integer().label('_idx'),
-                        t.c.place_id, t.c.parent_place_id,
-                        t.c.rank_search, t.c.rank_address,
+                        t.c.osm_id, t.c.place_id, t.c.parent_place_id,
+                        t.c.rank_search,
                         t.c.indexed_date, t.c.postcode, t.c.country_code,
-                        t.c.geometry.label('centroid'))\
+                        t.c.centroid)\
                 .where(t.c.place_id == pid_tab.c.value['id'].as_string().cast(sa.BigInteger))
 
+        if await collector.add_rows_from_sql(conn, sql, t.c.geometry,
+                                             nres.create_from_postcode_row):
+            return True
+
+    osm_ids = [{'i': i, 'oi': p.osm_id}
+               for i, p in collector.enumerate_free_osm_ids() if p.osm_type == 'R']
+
+    if osm_ids:
+        pid_tab = sa.func.JsonArrayEach(sa.type_coerce(osm_ids, sa.JSON))\
+                    .table_valued(sa.column('value', type_=sa.JSON))
+        t = conn.t.postcode
+        sql = sa.select(pid_tab.c.value['i'].as_integer().label('_idx'),
+                        t.c.osm_id, t.c.place_id, t.c.parent_place_id,
+                        t.c.rank_search,
+                        t.c.indexed_date, t.c.postcode, t.c.country_code,
+                        t.c.centroid)\
+                .where(t.c.osm_id == pid_tab.c.value['oi'].as_string().cast(sa.BigInteger))
+
         return await collector.add_rows_from_sql(conn, sql, t.c.geometry,
                                                  nres.create_from_postcode_row)
 

src/nominatim_api/result_formatting.py

@@ -10,7 +10,7 @@ Helper classes and functions for formatting results into API responses.
 from typing import Type, TypeVar, Dict, List, Callable, Any, Mapping, Optional, cast
 from collections import defaultdict
 from pathlib import Path
-import importlib
+import importlib.util
 
 from .server.content_types import CONTENT_JSON
 
@@ -43,7 +43,7 @@ class FormatDispatcher:
         return decorator
 
     def error_format_func(self, func: ErrorFormatFunc) -> ErrorFormatFunc:
-        """ Decorator for a function that formats error messges.
+        """ Decorator for a function that formats error messages.
             There is only one error formatter per dispatcher. Using
             the decorator repeatedly will overwrite previous functions.
         """
@@ -79,7 +79,7 @@ class FormatDispatcher:
     def set_content_type(self, fmt: str, content_type: str) -> None:
         """ Set the content type for the given format. This is the string
             that will be returned in the Content-Type header of the HTML
-            response, when the given format is choosen.
+            response, when the given format is chosen.
         """
         self.content_types[fmt] = content_type
 

src/nominatim_api/results.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Dataclasses for search results and helper functions to fill them.
@@ -407,11 +407,13 @@ def create_from_postcode_row(row: SaRow, class_type: Type[BaseResultT]) -> BaseR
     """
     return class_type(source_table=SourceTable.POSTCODE,
                       place_id=row.place_id,
+                      osm_object=None if row.osm_id is None else ('R', row.osm_id),
                       parent_place_id=row.parent_place_id,
-                      category=('place', 'postcode'),
+                      category=(('place', 'postcode') if row.osm_id is None
+                                else ('boundary', 'postal_code')),
                       names={'ref': row.postcode},
                       rank_search=row.rank_search,
-                      rank_address=row.rank_address,
+                      rank_address=5,
                       country_code=row.country_code,
                       centroid=Point.from_wkb(row.centroid),
                       geometry=_filter_geometries(row))
@@ -494,17 +496,15 @@ def _get_address_lookup_id(result: BaseResultT) -> int:
 
 async def _finalize_entry(conn: SearchConnection, result: BaseResultT) -> None:
     assert result.address_rows is not None
-    if result.category[0] not in ('boundary', 'place')\
-       or result.category[1] not in ('postal_code', 'postcode'):
-        postcode = result.postcode
-        if not postcode and result.address:
-            postcode = result.address.get('postcode')
-        if postcode and ',' not in postcode and ';' not in postcode:
-            result.address_rows.append(AddressLine(
-                category=('place', 'postcode'),
-                names={'ref': postcode},
-                fromarea=False, isaddress=True, rank_address=5,
-                distance=0.0))
+
+    postcode = result.postcode or (result.address and result.address.get('postcode'))
+    if postcode and ',' not in postcode and ';' not in postcode:
+        result.address_rows.append(AddressLine(
+            category=('place', 'postcode'),
+            names={'ref': postcode},
+            fromarea=False, isaddress=True, rank_address=5,
+            distance=0.0))
+
     if result.country_code:
         async def _get_country_names() -> Optional[Dict[str, str]]:
             t = conn.t.country_name
@@ -627,13 +627,6 @@ async def complete_address_details(conn: SearchConnection, results: List[BaseRes
         if current_result.country_code is None and row.country_code:
             current_result.country_code = row.country_code
 
-        if row.type in ('postcode', 'postal_code') and location_isaddress:
-            if not row.fromarea or \
-               (current_result.address and 'postcode' in current_result.address):
-                location_isaddress = False
-            else:
-                current_result.postcode = None
-
         assert current_result.address_rows is not None
         current_result.address_rows.append(_result_row_to_address_row(row, location_isaddress))
         current_rank_address = row.rank_address

src/nominatim_api/reverse.py

@@ -157,16 +157,19 @@ class ReverseGeocoder:
             include.extend(('natural', 'water', 'waterway'))
         return table.c.class_.in_(tuple(include))
 
-    async def _find_closest_street_or_poi(self, distance: float) -> Optional[SaRow]:
-        """ Look up the closest rank 26+ place in the database, which
-            is closer than the given distance.
+    async def _find_closest_street_or_pois(self, distance: float,
+                                           fuzziness: float) -> list[SaRow]:
+        """ Look up the closest rank 26+ place in the database.
+            The function finds the object that is closest to the reverse
+            search point as well as all objects within 'fuzziness' distance
+            to that best result.
         """
         t = self.conn.t.placex
 
         # PostgreSQL must not get the distance as a parameter because
         # there is a danger it won't be able to properly estimate index use
         # when used with prepared statements
-        diststr = sa.text(f"{distance}")
+        diststr = sa.text(f"{distance + fuzziness}")
 
         sql: SaLambdaSelect = sa.lambda_stmt(
             lambda: _select_from_placex(t)
@@ -174,9 +177,7 @@ class ReverseGeocoder:
             .where(t.c.indexed_status == 0)
             .where(t.c.linked_place_id == None)
             .where(sa.or_(sa.not_(t.c.geometry.is_area()),
-                          t.c.centroid.ST_Distance(WKT_PARAM) < diststr))
-            .order_by('distance')
-            .limit(2))
+                          t.c.centroid.ST_Distance(WKT_PARAM) < diststr)))
 
         if self.has_geometries():
             sql = self._add_geometry_columns(sql, t.c.geometry)
@@ -198,24 +199,44 @@ class ReverseGeocoder:
                                     self._filter_by_layer(t)))
 
         if not restrict:
-            return None
+            return []
 
-        sql = sql.where(sa.or_(*restrict))
+        inner = sql.where(sa.or_(*restrict)) \
+                   .add_columns(t.c.geometry.label('_geometry')) \
+                   .subquery()
 
-        # If the closest object is inside an area, then check if there is a
-        # POI node nearby and return that.
-        prev_row = None
-        for row in await self.conn.execute(sql, self.bind_params):
-            if prev_row is None:
-                if row.rank_search <= 27 or row.osm_type == 'N' or row.distance > 0:
-                    return row
-                prev_row = row
-            else:
-                if row.rank_search > 27 and row.osm_type == 'N'\
-                   and row.distance < 0.0001:
-                    return row
+        # Use a window function to get the closest results to the best result.
+        windowed = sa.select(inner,
+                             sa.func.first_value(inner.c.distance)
+                                    .over(order_by=inner.c.distance)
+                                    .label('_min_distance'),
+                             sa.func.first_value(
+                                        sa.case((inner.c.rank_search <= 27,
+                                                 inner.c._geometry.ST_ClosestPoint(WKT_PARAM)),
+                                                else_=None))
+                                    .over(order_by=inner.c.distance)
+                                    .label('_closest_point'),
+                             sa.func.first_value(sa.case((sa.or_(inner.c.rank_search <= 27,
+                                                                 inner.c.osm_type == 'N'), None),
+                                                         else_=inner.c._geometry))
+                                    .over(order_by=inner.c.distance)
+                                    .label('_best_geometry')) \
+                     .subquery()
 
-        return prev_row
+        outer = sa.select(*(c for c in windowed.c if not c.key.startswith('_')),
+                          sa.case((sa.or_(windowed.c._closest_point == None,
+                                          windowed.c.housenumber == None), None),
+                                  else_=windowed.c.centroid.ST_Distance(windowed.c._closest_point))
+                            .label('distance_from_best'),
+                          sa.case((sa.or_(windowed.c._best_geometry == None,
+                                          windowed.c.rank_search <= 27,
+                                          windowed.c.osm_type != 'N'), False),
+                                  else_=windowed.c.centroid.ST_CoveredBy(windowed.c._best_geometry))
+                            .label('best_inside')) \
+                  .where(windowed.c.distance < windowed.c._min_distance + fuzziness) \
+                  .order_by(windowed.c.distance)
+
+        return list(await self.conn.execute(outer, self.bind_params))
 
     async def _find_housenumber_for_street(self, parent_place_id: int) -> Optional[SaRow]:
         t = self.conn.t.placex
@@ -301,55 +322,69 @@ class ReverseGeocoder:
         """ Find a street or POI/address for the given WKT point.
         """
         log().section('Reverse lookup on street/address level')
-        distance = 0.006
-        parent_place_id = None
-
-        row = await self._find_closest_street_or_poi(distance)
         row_func: RowFunc = nres.create_from_placex_row
-        log().var_dump('Result (street/building)', row)
+        distance = 0.006
 
-        # If the closest result was a street, but an address was requested,
-        # check for a housenumber nearby which is part of the street.
-        if row is not None:
-            if self.max_rank > 27 \
-               and self.layer_enabled(DataLayer.ADDRESS) \
-               and row.rank_address <= 27:
-                distance = 0.001
-                parent_place_id = row.place_id
-                log().comment('Find housenumber for street')
-                addr_row = await self._find_housenumber_for_street(parent_place_id)
-                log().var_dump('Result (street housenumber)', addr_row)
-
-                if addr_row is not None:
-                    row = addr_row
-                    row_func = nres.create_from_placex_row
-                    distance = addr_row.distance
-                elif row.country_code == 'us' and parent_place_id is not None:
-                    log().comment('Find TIGER housenumber for street')
-                    addr_row = await self._find_tiger_number_for_street(parent_place_id)
-                    log().var_dump('Result (street Tiger housenumber)', addr_row)
-
-                    if addr_row is not None:
-                        row_func = cast(RowFunc,
-                                        functools.partial(nres.create_from_tiger_row,
-                                                          osm_type=row.osm_type,
-                                                          osm_id=row.osm_id))
-                        row = addr_row
-            else:
+        result = None
+        hnr_distance = None
+        parent_street = None
+        for row in await self._find_closest_street_or_pois(distance, 0.001):
+            if result is None:
+                log().var_dump('Closest result', row)
+                result = row
+                if self.max_rank > 27 \
+                        and self.layer_enabled(DataLayer.ADDRESS) \
+                        and result.rank_address <= 27:
+                    parent_street = result.place_id
+                    distance = 0.001
+                else:
+                    distance = row.distance
+            # If the closest result was a street but an address was requested,
+            # see if we can refine the result with a housenumber closeby.
+            elif parent_street is not None \
+                    and row.distance_from_best is not None \
+                    and row.distance_from_best < 0.001 \
+                    and (hnr_distance is None or hnr_distance > row.distance_from_best) \
+                    and row.parent_place_id == parent_street:
+                log().var_dump('Housenumber to closest result', row)
+                result = row
+                hnr_distance = row.distance_from_best
                 distance = row.distance
+            # If the closest object is inside an area, then check if there is
+            # a POI nearby and return that with preference.
+            elif result.osm_type != 'N' and result.rank_search > 27 \
+                    and result.distance == 0 \
+                    and row.best_inside:
+                log().var_dump('POI near closest result area', row)
+                result = row
+                break  # it can't get better than that, everything else is farther away
+
+        # For the US also check the TIGER data, when no housenumber/POI was found.
+        if result is not None and parent_street is not None and hnr_distance is None \
+                and result.country_code == 'us':
+            log().comment('Find TIGER housenumber for street')
+            addr_row = await self._find_tiger_number_for_street(parent_street)
+            log().var_dump('Result (street Tiger housenumber)', addr_row)
+
+            if addr_row is not None:
+                row_func = cast(RowFunc,
+                                functools.partial(nres.create_from_tiger_row,
+                                                  osm_type=row.osm_type,
+                                                  osm_id=row.osm_id))
+                result = addr_row
 
         # Check for an interpolation that is either closer than our result
         # or belongs to a close street found.
-        if self.max_rank > 27 and self.layer_enabled(DataLayer.ADDRESS):
+        # No point in doing this when the result is already inside a building,
+        # i.e. when the distance is already 0.
+        if self.max_rank > 27 and self.layer_enabled(DataLayer.ADDRESS) and distance > 0:
             log().comment('Find interpolation for street')
-            addr_row = await self._find_interpolation_for_street(parent_place_id,
-                                                                 distance)
+            addr_row = await self._find_interpolation_for_street(parent_street, distance)
             log().var_dump('Result (street interpolation)', addr_row)
             if addr_row is not None:
-                row = addr_row
-                row_func = nres.create_from_osmline_row
+                return addr_row, nres.create_from_osmline_row
 
-        return row, row_func
+        return result, row_func
 
     async def _lookup_area_address(self) -> Optional[SaRow]:
         """ Lookup large addressable areas for the given WKT point.

src/nominatim_api/search/db_search_builder.py

@@ -374,7 +374,7 @@ class SearchBuilder:
             tokens = self.get_country_tokens(assignment.country)
             if not tokens:
                 return None
-            sdata.set_strings('countries', tokens)
+            sdata.set_countries(tokens)
             sdata.penalty += self.query.get_in_word_penalty(assignment.country)
         elif self.details.countries:
             sdata.countries = dbf.WeightedStrings(self.details.countries,
@@ -413,7 +413,7 @@ class SearchBuilder:
         """
         tokens = self.query.get_tokens(trange, qmod.TOKEN_COUNTRY)
         if self.details.countries:
-            tokens = [t for t in tokens if t.lookup_word in self.details.countries]
+            tokens = [t for t in tokens if t.get_country() in self.details.countries]
 
         return tokens
 

src/nominatim_api/search/db_search_fields.py

@@ -22,7 +22,7 @@ class CountedTokenIDs:
     """ A list of token IDs with their respective counts, sorted
         from least frequent to most frequent.
 
-        If a token count is one, then statistics are likely to be unavaible
+        If a token count is one, then statistics are likely to be unavailable
         and a relatively high count is assumed instead.
     """
 
@@ -244,6 +244,21 @@ class SearchData:
 
             setattr(self, field, wstrs)
 
+    def set_countries(self, tokens: List[Token]) -> None:
+        """ Set the WeightedStrings properties for countries. Multiple
+            entries for the same country are deduplicated and the minimum
+            penalty is used. Adapts the global penalty, so that the
+            minimum penalty is 0.
+        """
+        if tokens:
+            min_penalty = min(t.penalty for t in tokens)
+            self.penalty += min_penalty
+            countries: dict[str, float] = {}
+            for t in tokens:
+                cc = t.get_country()
+                countries[cc] = min(t.penalty - min_penalty, countries.get(cc, 10000))
+            self.countries = WeightedStrings(list(countries.keys()), list(countries.values()))
+
     def set_qualifiers(self, tokens: List[Token]) -> None:
         """ Set the qulaifier field from the given tokens.
         """

src/nominatim_api/search/db_searches/address_search.py

@@ -175,7 +175,8 @@ class AddressSearch(base.AbstractSearch):
                 sql = sql.where(sa.select(tpc.c.postcode)
                                   .where(tpc.c.postcode.in_(self.postcodes.values))
                                   .where(tpc.c.country_code == t.c.country_code)
-                                  .where(t.c.centroid.within_distance(tpc.c.geometry, 0.4))
+                                  .where(t.c.centroid.intersects(tpc.c.geometry,
+                                                                 use_index=False))
                                   .exists())
 
         if details.viewbox is not None:
@@ -225,7 +226,7 @@ class AddressSearch(base.AbstractSearch):
             tpc = conn.t.postcode
             pcs = self.postcodes.values
 
-            pc_near = sa.select(sa.func.min(tpc.c.geometry.ST_Distance(t.c.centroid)
+            pc_near = sa.select(sa.func.min(tpc.c.centroid.ST_Distance(t.c.centroid)
                                             * (tpc.c.rank_search - 19)))\
                         .where(tpc.c.postcode.in_(pcs))\
                         .where(tpc.c.country_code == t.c.country_code)\

src/nominatim_api/search/db_searches/place_search.py

@@ -79,7 +79,8 @@ class PlaceSearch(base.AbstractSearch):
                 tpc = conn.t.postcode
                 sql = sql.where(sa.select(tpc.c.postcode)
                                   .where(tpc.c.postcode.in_(self.postcodes.values))
-                                  .where(t.c.centroid.within_distance(tpc.c.geometry, 0.4))
+                                  .where(t.c.centroid.intersects(tpc.c.geometry,
+                                                                 use_index=False))
                                   .exists())
 
         if details.viewbox is not None:
@@ -157,7 +158,7 @@ class PlaceSearch(base.AbstractSearch):
                 tpc = conn.t.postcode
                 pcs = self.postcodes.values
 
-                pc_near = sa.select(sa.func.min(tpc.c.geometry.ST_Distance(t.c.centroid)))\
+                pc_near = sa.select(sa.func.min(tpc.c.centroid.ST_Distance(t.c.centroid)))\
                             .where(tpc.c.postcode.in_(pcs))\
                             .scalar_subquery()
                 penalty += sa.case((t.c.postcode.in_(pcs), 0.0),

src/nominatim_api/search/db_searches/postcode_search.py

@@ -14,7 +14,7 @@ from . import base
 from ...typing import SaBind, SaExpression
 from ...sql.sqlalchemy_types import Geometry, IntArray
 from ...connection import SearchConnection
-from ...types import SearchDetails, Bbox
+from ...types import SearchDetails
 from ... import results as nres
 from ..db_search_fields import SearchData
 
@@ -42,10 +42,9 @@ class PostcodeSearch(base.AbstractSearch):
         t = conn.t.postcode
         pcs = self.postcodes.values
 
-        sql = sa.select(t.c.place_id, t.c.parent_place_id,
-                        t.c.rank_search, t.c.rank_address,
-                        t.c.postcode, t.c.country_code,
-                        t.c.geometry.label('centroid'))\
+        sql = sa.select(t.c.place_id, t.c.parent_place_id, t.c.osm_id,
+                        t.c.rank_search, t.c.postcode, t.c.country_code,
+                        t.c.centroid)\
                 .where(t.c.postcode.in_(pcs))
 
         if details.geometry_output:
@@ -59,7 +58,7 @@ class PostcodeSearch(base.AbstractSearch):
                                else_=1.0)
 
         if details.near is not None:
-            sql = sql.order_by(t.c.geometry.ST_Distance(NEAR_PARAM))
+            sql = sql.order_by(t.c.centroid.ST_Distance(NEAR_PARAM))
 
         sql = base.filter_by_area(sql, t, details)
 
@@ -100,29 +99,9 @@ class PostcodeSearch(base.AbstractSearch):
 
         results = nres.SearchResults()
         for row in await conn.execute(sql, bind_params):
-            p = conn.t.placex
-            placex_sql = base.select_placex(p)\
-                .add_columns(p.c.importance)\
-                .where(sa.text("""class = 'boundary'
-                                  AND type = 'postal_code'
-                                  AND osm_type = 'R'"""))\
-                .where(p.c.country_code == row.country_code)\
-                .where(p.c.postcode == row.postcode)\
-                .limit(1)
+            result = nres.create_from_postcode_row(row, nres.SearchResult)
 
-            if details.geometry_output:
-                placex_sql = base.add_geometry_columns(placex_sql, p.c.geometry, details)
-
-            for prow in await conn.execute(placex_sql, bind_params):
-                result = nres.create_from_placex_row(prow, nres.SearchResult)
-                if result is not None:
-                    result.bbox = Bbox.from_wkb(prow.bbox)
-                break
-            else:
-                result = nres.create_from_postcode_row(row, nres.SearchResult)
-
-            if result.place_id not in details.excluded:
-                result.accuracy = row.accuracy
-                results.append(result)
+            result.accuracy = row.accuracy
+            results.append(result)
 
         return results

src/nominatim_api/search/icu_tokenizer.py

@@ -59,12 +59,16 @@ class ICUToken(qmod.Token):
         assert self.info
         return self.info.get('class', ''), self.info.get('type', '')
 
-    def rematch(self, norm: str) -> None:
+    def get_country(self) -> str:
+        assert self.info
+        return cast(str, self.info.get('cc', ''))
+
+    def match_penalty(self, norm: str) -> float:
         """ Check how well the token matches the given normalized string
             and add a penalty, if necessary.
         """
         if not self.lookup_word:
-            return
+            return 0.0
 
         seq = difflib.SequenceMatcher(a=self.lookup_word, b=norm)
         distance = 0
@@ -75,7 +79,7 @@ class ICUToken(qmod.Token):
                 distance += max((ato-afrom), (bto-bfrom))
             elif tag != 'equal':
                 distance += abs((ato-afrom) - (bto-bfrom))
-        self.penalty += (distance/len(self.lookup_word))
+        return (distance/len(self.lookup_word))
 
     @staticmethod
     def from_db_row(row: SaRow) -> 'ICUToken':
@@ -330,9 +334,10 @@ class ICUQueryAnalyzer(AbstractQueryAnalyzer):
             norm = ''.join(f"{n.term_normalized}{'' if n.btype == qmod.BREAK_TOKEN else ' '}"
                            for n in query.nodes[start + 1:end + 1]).strip()
             for ttype, tokens in tlist.items():
-                if ttype != qmod.TOKEN_COUNTRY:
-                    for token in tokens:
-                        cast(ICUToken, token).rematch(norm)
+                for token in tokens:
+                    itok = cast(ICUToken, token)
+                    itok.penalty += itok.match_penalty(norm) * \
+                        (1 if ttype in (qmod.TOKEN_WORD, qmod.TOKEN_PARTIAL) else 2)
 
     def compute_break_penalties(self, query: qmod.QueryStruct) -> None:
         """ Set the break penalties for the nodes in the query.

src/nominatim_api/search/query.py

@@ -17,7 +17,7 @@ import dataclasses
 # The x value for the regression computation will be the position of the
 # token in the query. Thus we know the x values will be [0, query length).
 # As the denominator only depends on the x values, we can pre-compute here
-# the denominatior to use for a given query length.
+# the denominator to use for a given query length.
 # Note that query length of two or less is special cased and will not use
 # the values from this array. Thus it is not a problem that they are 0.
 LINFAC = [i * (sum(si * si for si in range(i)) - (i - 1) * i * (i - 1) / 4)
@@ -127,6 +127,12 @@ class Token(ABC):
             category objects.
         """
 
+    @abstractmethod
+    def get_country(self) -> str:
+        """ Return the country code this token is associated with
+            (currently for country tokens only).
+        """
+
 
 @dataclasses.dataclass
 class TokenRange:
@@ -225,7 +231,7 @@ class QueryNode:
         return max(0, -self.penalty)
 
     def name_address_ratio(self) -> float:
-        """ Return the propability that the partial token belonging to
+        """ Return the probability that the partial token belonging to
             this node forms part of a name (as opposed of part of the address).
         """
         if self.partial is None:
@@ -269,7 +275,7 @@ class QueryStruct:
         directed acyclic graph.
 
         A query also has a direction penalty 'dir_penalty'. This describes
-        the likelyhood if the query should be read from left-to-right or
+        the likelihood if the query should be read from left-to-right or
         vice versa. A negative 'dir_penalty' should be read as a penalty on
         right-to-left reading, while a positive value represents a penalty
         for left-to-right reading. The default value is 0, which is equivalent

src/nominatim_api/server/falcon/server.py

@@ -7,6 +7,8 @@
 """
 Server implementation using the falcon webserver framework.
 """
+from __future__ import annotations
+
 from typing import Optional, Mapping, Any, List, cast
 from pathlib import Path
 import asyncio
@@ -161,7 +163,7 @@ class APIMiddleware:
 
     def __init__(self, project_dir: Path, environ: Optional[Mapping[str, str]]) -> None:
         self.api = NominatimAPIAsync(project_dir, environ)
-        self.app: Optional[App] = None
+        self.app: Optional[App[Request, Response]] = None
 
     @property
     def config(self) -> Configuration:
@@ -169,7 +171,7 @@ class APIMiddleware:
         """
         return self.api.config
 
-    def set_app(self, app: App) -> None:
+    def set_app(self, app: App[Request, Response]) -> None:
         """ Set the Falcon application this middleware is connected to.
         """
         self.app = app
@@ -193,7 +195,7 @@ class APIMiddleware:
 
 
 def get_application(project_dir: Path,
-                    environ: Optional[Mapping[str, str]] = None) -> App:
+                    environ: Optional[Mapping[str, str]] = None) -> App[Request, Response]:
     """ Create a Nominatim Falcon ASGI application.
     """
     apimw = APIMiddleware(project_dir, environ)
@@ -215,7 +217,7 @@ def get_application(project_dir: Path,
     return app
 
 
-def run_wsgi() -> App:
+def run_wsgi() -> App[Request, Response]:
     """ Entry point for uvicorn.
 
         Make sure uvicorn is run from the project directory.

src/nominatim_api/server/starlette/server.py

@@ -50,7 +50,7 @@ class ParamWrapper(ASGIAdaptor):
                              headers={'content-type': self.content_type})
 
     def create_response(self, status: int, output: str, num_results: int) -> Response:
-        self.request.state.num_results = num_results
+        setattr(self.request.state, 'num_results', num_results)
         return Response(output, status_code=status, media_type=self.content_type)
 
     def base_uri(self) -> str:
@@ -95,7 +95,7 @@ class FileLoggingMiddleware(BaseHTTPMiddleware):
     async def dispatch(self, request: Request,
                        call_next: RequestResponseEndpoint) -> Response:
         qs = QueryStatistics()
-        request.state.query_stats = qs
+        setattr(request.state, 'query_stats', qs)
         response = await call_next(request)
 
         if response.status_code != 200 or 'start' not in qs:

src/nominatim_api/sql/sqlalchemy_functions.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Custom functions and expressions for SQLAlchemy.
@@ -32,7 +32,6 @@ def _default_intersects(element: PlacexGeometryReverseLookuppolygon,
                         compiler: 'sa.Compiled', **kw: Any) -> str:
     return ("(ST_GeometryType(placex.geometry) in ('ST_Polygon', 'ST_MultiPolygon')"
             " AND placex.rank_address between 4 and 25"
-            " AND placex.type != 'postcode'"
             " AND placex.name is not null"
             " AND placex.indexed_status = 0"
             " AND placex.linked_place_id is null)")
@@ -43,7 +42,6 @@ def _sqlite_intersects(element: PlacexGeometryReverseLookuppolygon,
                        compiler: 'sa.Compiled', **kw: Any) -> str:
     return ("(ST_GeometryType(placex.geometry) in ('POLYGON', 'MULTIPOLYGON')"
             " AND placex.rank_address between 4 and 25"
-            " AND placex.type != 'postcode'"
             " AND placex.name is not null"
             " AND placex.indexed_status = 0"
             " AND placex.linked_place_id is null)")
@@ -64,7 +62,6 @@ def default_reverse_place_diameter(element: IntersectsReverseDistance,
                                    compiler: 'sa.Compiled', **kw: Any) -> str:
     table = element.tablename
     return f"({table}.rank_address between 4 and 25"\
-           f" AND {table}.type != 'postcode'"\
            f" AND {table}.name is not null"\
            f" AND {table}.linked_place_id is null"\
            f" AND {table}.osm_type = 'N'" + \
@@ -79,7 +76,6 @@ def sqlite_reverse_place_diameter(element: IntersectsReverseDistance,
     table = element.tablename
 
     return (f"({table}.rank_address between 4 and 25"
-            f" AND {table}.type != 'postcode'"
             f" AND {table}.name is not null"
             f" AND {table}.linked_place_id is null"
             f" AND {table}.osm_type = 'N'"

src/nominatim_api/sql/sqlalchemy_schema.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 SQLAlchemy definitions for all tables used by the frontend.
@@ -67,15 +67,16 @@ class SearchTables:
             sa.Column('isaddress', sa.Boolean))
 
         self.postcode = sa.Table(
-            'location_postcode', meta,
+            'location_postcodes', meta,
             sa.Column('place_id', sa.BigInteger),
             sa.Column('parent_place_id', sa.BigInteger),
+            sa.Column('osm_id', sa.BigInteger),
             sa.Column('rank_search', sa.SmallInteger),
-            sa.Column('rank_address', sa.SmallInteger),
             sa.Column('indexed_status', sa.SmallInteger),
             sa.Column('indexed_date', sa.DateTime),
             sa.Column('country_code', sa.String(2)),
             sa.Column('postcode', sa.Text),
+            sa.Column('centroid', Geometry),
             sa.Column('geometry', Geometry))
 
         self.osmline = sa.Table(

src/nominatim_api/sql/sqlalchemy_types/geometry.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Custom types for SQLAlchemy.
@@ -178,6 +178,8 @@ class Geometry(types.UserDefinedType):  # type: ignore[type-arg]
 
     def bind_processor(self, dialect: 'sa.Dialect') -> Callable[[Any], str]:
         def process(value: Any) -> str:
+            if value is None:
+                return 'null'
             if isinstance(value, str):
                 return value
 

src/nominatim_api/v1/classtypes.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Hard-coded information about tag categories.
@@ -20,7 +20,9 @@ def get_label_tag(category: Tuple[str, str], extratags: Optional[Mapping[str, st
                   rank: int, country: Optional[str]) -> str:
     """ Create a label tag for the given place that can be used as an XML name.
     """
-    if rank < 26 and extratags and 'place' in extratags:
+    if category in (('place', 'postcode'), ('boundary', 'postal_code')):
+        label = 'postcode'
+    elif rank < 26 and extratags and 'place' in extratags:
         label = extratags['place']
     elif rank < 26 and extratags and 'linked_place' in extratags:
         label = extratags['linked_place']
@@ -28,8 +30,6 @@ def get_label_tag(category: Tuple[str, str], extratags: Optional[Mapping[str, st
         label = ADMIN_LABELS.get((country or '', rank // 2))\
                 or ADMIN_LABELS.get(('', rank // 2))\
                 or 'Administrative'
-    elif category[1] == 'postal_code':
-        label = 'postcode'
     elif rank < 26:
         label = category[1] if category[1] != 'yes' else category[0]
     elif rank < 28:

src/nominatim_api/v1/format.py

@@ -132,7 +132,7 @@ def _add_parent_rows_grouped(writer: JsonWriter, rows: AddressLines,
 
 @dispatch.format_func(DetailedResult, 'json')
 def _format_details_json(result: DetailedResult, options: Mapping[str, Any]) -> str:
-    locales = options.get('locales', Locales())
+    locales = options.get('locales') or Locales()
     geom = result.geometry.get('geojson')
     centroid = result.centroid.to_geojson()
 

src/nominatim_api/v1/helpers.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Helper function for parsing parameters and and outputting data
@@ -12,7 +12,7 @@ from typing import Tuple, Optional, Any, Dict, Iterable
 from itertools import chain
 import re
 
-from ..results import SearchResult, SearchResults, SourceTable
+from ..results import SearchResults, SourceTable
 from ..types import SearchDetails, GeometryFormat
 
 
@@ -106,10 +106,6 @@ def deduplicate_results(results: SearchResults, max_results: int) -> SearchResul
     classification_done = set()
     deduped = SearchResults()
     for result in results:
-        if result.source_table == SourceTable.POSTCODE:
-            assert result.names and 'ref' in result.names
-            if any(_is_postcode_relation_for(r, result.names['ref']) for r in results):
-                continue
         if result.source_table == SourceTable.PLACEX:
             classification = (result.osm_object[0] if result.osm_object else None,
                               result.category,
@@ -128,15 +124,6 @@ def deduplicate_results(results: SearchResults, max_results: int) -> SearchResul
     return deduped
 
 
-def _is_postcode_relation_for(result: SearchResult, postcode: str) -> bool:
-    return result.source_table == SourceTable.PLACEX \
-           and result.osm_object is not None \
-           and result.osm_object[0] == 'R' \
-           and result.category == ('boundary', 'postal_code') \
-           and result.names is not None \
-           and result.names.get('ref') == postcode
-
-
 def _deg(axis: str) -> str:
     return f"(?P<{axis}_deg>\\d+\\.\\d+)°?"
 

src/nominatim_api/v1/server_glue.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2025 by the Nominatim developer community.
+# Copyright (C) 2026 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Generic part of the server implementation of the v1 API.
@@ -174,7 +174,8 @@ async def details_endpoint(api: NominatimAPIAsync, params: ASGIAdaptor) -> Any:
     if result is None:
         params.raise_error('No place with that OSM ID found.', status=404)
 
-    locales = Locales.from_accept_languages(get_accepted_languages(params))
+    locales = Locales.from_accept_languages(get_accepted_languages(params),
+                                            params.config().OUTPUT_NAMES)
     locales.localize_results([result])
 
     output = params.formatting().format_result(
@@ -199,6 +200,7 @@ async def reverse_endpoint(api: NominatimAPIAsync, params: ASGIAdaptor) -> Any:
     details['max_rank'] = helpers.zoom_to_rank(params.get_int('zoom', 18))
     details['layers'] = get_layers(params)
     details['query_stats'] = params.query_stats()
+    details['entrances'] = params.get_bool('entrances', False)
 
     result = await api.reverse(coord, **details)
 
@@ -215,8 +217,8 @@ async def reverse_endpoint(api: NominatimAPIAsync, params: ASGIAdaptor) -> Any:
         query = ''
 
     if result:
-        Locales.from_accept_languages(get_accepted_languages(params)).localize_results(
-            [result])
+        Locales.from_accept_languages(get_accepted_languages(params),
+                                      params.config().OUTPUT_NAMES).localize_results([result])
 
     fmt_options = {'query': query,
                    'extratags': params.get_bool('extratags', False),
@@ -237,6 +239,7 @@ async def lookup_endpoint(api: NominatimAPIAsync, params: ASGIAdaptor) -> Any:
     debug = setup_debugging(params)
     details = parse_geometry_details(params, fmt)
     details['query_stats'] = params.query_stats()
+    details['entrances'] = params.get_bool('entrances', False)
 
     places = []
     for oid in (params.get('osm_ids') or '').split(','):
@@ -255,7 +258,8 @@ async def lookup_endpoint(api: NominatimAPIAsync, params: ASGIAdaptor) -> Any:
     if debug:
         return build_response(params, loglib.get_and_disable(), num_results=len(results))
 
-    Locales.from_accept_languages(get_accepted_languages(params)).localize_results(results)
+    Locales.from_accept_languages(get_accepted_languages(params),
+                                  params.config().OUTPUT_NAMES).localize_results(results)
 
     fmt_options = {'extratags': params.get_bool('extratags', False),
                    'namedetails': params.get_bool('namedetails', False),
@@ -348,7 +352,8 @@ async def search_endpoint(api: NominatimAPIAsync, params: ASGIAdaptor) -> Any:
     except UsageError as err:
         params.raise_error(str(err))
 
-    Locales.from_accept_languages(get_accepted_languages(params)).localize_results(results)
+    Locales.from_accept_languages(get_accepted_languages(params),
+                                  params.config().OUTPUT_NAMES).localize_results(results)
 
     if details['dedupe'] and len(results) > 1:
         results = helpers.deduplicate_results(results, max_results)

src/nominatim_db/clicmd/add_data.py

@@ -65,14 +65,14 @@ class UpdateAddData:
     def run(self, args: NominatimArgs) -> int:
         from ..tools import add_osm_data
 
+        if args.tiger_data:
+            return asyncio.run(self._add_tiger_data(args))
+
         with connect(args.config.get_libpq_dsn()) as conn:
             if is_frozen(conn):
                 print('Database is marked frozen. New data can\'t be added.')
                 return 1
 
-        if args.tiger_data:
-            return asyncio.run(self._add_tiger_data(args))
-
         osm2pgsql_params = args.osm2pgsql_options(default_cache=1000, default_threads=1)
         if args.file or args.diff:
             return add_osm_data.add_data_from_file(args.config.get_libpq_dsn(),

src/nominatim_db/clicmd/api.py

@@ -19,6 +19,7 @@ import nominatim_api as napi
 from nominatim_api.v1.helpers import zoom_to_rank, deduplicate_results
 from nominatim_api.server.content_types import CONTENT_JSON
 import nominatim_api.logging as loglib
+from ..config import Configuration
 from ..errors import UsageError
 from .args import NominatimArgs
 
@@ -91,18 +92,19 @@ def _get_geometry_output(args: NominatimArgs) -> napi.GeometryFormat:
         raise UsageError(f"Unknown polygon output format '{args.polygon_output}'.") from exp
 
 
-def _get_locales(args: NominatimArgs, default: Optional[str]) -> napi.Locales:
+def _get_locales(args: NominatimArgs, config: Configuration) -> napi.Locales:
     """ Get the locales from the language parameter.
     """
-    if args.lang:
-        return napi.Locales.from_accept_languages(args.lang)
-    if default:
-        return napi.Locales.from_accept_languages(default)
+    language = args.lang or config.DEFAULT_LANGUAGE
+    output_names = config.OUTPUT_NAMES
+
+    if language:
+        return napi.Locales.from_accept_languages(language, output_names)
 
     return napi.Locales()
 
 
-def _get_layers(args: NominatimArgs, default: napi.DataLayer) -> Optional[napi.DataLayer]:
+def _get_layers(args: NominatimArgs, default: Optional[napi.DataLayer]) -> Optional[napi.DataLayer]:
     """ Get the list of selected layers as a DataLayer enum.
     """
     if not args.layers:
@@ -134,7 +136,7 @@ def _print_output(formatter: napi.FormatDispatcher, result: Any,
                 json.dump(json.loads(output), sys.stdout, indent=4, ensure_ascii=False)
             except json.decoder.JSONDecodeError as err:
                 # Catch the error here, so that data can be debugged,
-                # when people are developping custom result formatters.
+                # when people are developing custom result formatters.
                 LOG.fatal("Parsing json failed: %s\nUnformatted output:\n%s", err, output)
         else:
             sys.stdout.write(output)
@@ -171,6 +173,10 @@ class APISearch:
                            help='Preferred area to find search results')
         group.add_argument('--bounded', action='store_true',
                            help='Strictly restrict results to viewbox area')
+        group.add_argument('--layer', metavar='LAYER',
+                           choices=[n.name.lower() for n in napi.DataLayer if n.name],
+                           action='append', required=False, dest='layers',
+                           help='Restrict results to one or more layers (may be repeated)')
         group.add_argument('--no-dedupe', action='store_false', dest='dedupe',
                            help='Do not remove duplicates from the result list')
         _add_list_format(parser)
@@ -187,6 +193,8 @@ class APISearch:
             raise UsageError(f"Unsupported format '{args.format}'. "
                              'Use --list-formats to see supported formats.')
 
+        layers = _get_layers(args, None)
+
         try:
             with napi.NominatimAPI(args.project_dir) as api:
                 params: Dict[str, Any] = {'max_results': args.limit + min(args.limit, 10),
@@ -197,6 +205,7 @@ class APISearch:
                                           'excluded': args.exclude_place_ids,
                                           'viewbox': args.viewbox,
                                           'bounded_viewbox': args.bounded,
+                                          'layers': layers,
                                           'entrances': args.entrances,
                                           }
 
@@ -214,7 +223,7 @@ class APISearch:
         except napi.UsageError as ex:
             raise UsageError(ex) from ex
 
-        locales = _get_locales(args, api.config.DEFAULT_LANGUAGE)
+        locales = _get_locales(args, api.config)
         locales.localize_results(results)
 
         if args.dedupe and len(results) > 1:
@@ -253,7 +262,7 @@ class APIReverse:
         group.add_argument('--layer', metavar='LAYER',
                            choices=[n.name.lower() for n in napi.DataLayer if n.name],
                            action='append', required=False, dest='layers',
-                           help='OSM id to lookup in format <NRW><id> (may be repeated)')
+                           help='Restrict results to one or more layers (may be repeated)')
 
         _add_api_output_arguments(parser)
         _add_list_format(parser)
@@ -287,7 +296,7 @@ class APIReverse:
             raise UsageError(ex) from ex
 
         if result is not None:
-            locales = _get_locales(args, api.config.DEFAULT_LANGUAGE)
+            locales = _get_locales(args, api.config)
             locales.localize_results([result])
 
         if args.format == 'debug':
@@ -352,7 +361,7 @@ class APILookup:
         except napi.UsageError as ex:
             raise UsageError(ex) from ex
 
-        locales = _get_locales(args, api.config.DEFAULT_LANGUAGE)
+        locales = _get_locales(args, api.config)
         locales.localize_results(results)
 
         if args.format == 'debug':
@@ -452,7 +461,7 @@ class APIDetails:
             raise UsageError(ex) from ex
 
         if result is not None:
-            locales = _get_locales(args, api.config.DEFAULT_LANGUAGE)
+            locales = _get_locales(args, api.config)
             locales.localize_results([result])
 
         if args.format == 'debug':

src/nominatim_db/clicmd/args.py

@@ -119,6 +119,7 @@ class NominatimArgs:
     enable_debug_statements: bool
     data_object: Sequence[Tuple[str, int]]
     data_area: Sequence[Tuple[str, int]]
+    ro_access: bool
 
     # Arguments to 'replication'
     init: bool

src/nominatim_db/clicmd/export.py

@@ -154,7 +154,7 @@ async def dump_results(conn: napi.SearchConnection,
     await add_result_details(conn, results,
                              LookupDetails(address_details=True))
 
-    locale = napi.Locales([lang] if lang else None)
+    locale = napi.Locales([lang] if lang else None, conn.config.OUTPUT_NAMES)
     locale.localize_results(results)
 
     for result in results:

src/nominatim_db/clicmd/refresh.py

@@ -65,6 +65,8 @@ class UpdateRefresh:
                            help='Update secondary importance raster data')
         group.add_argument('--importance', action='store_true',
                            help='Recompute place importances (expensive!)')
+        group.add_argument('--ro-access', action='store_true',
+                           help='Grant read-only access to web user for all tables')
         group.add_argument('--website', action='store_true',
                            help='DEPRECATED. This function has no function anymore'
                                 ' and will be removed in a future version.')
@@ -159,6 +161,11 @@ class UpdateRefresh:
             LOG.error('WARNING: Website setup is no longer required. '
                       'This function will be removed in future version of Nominatim.')
 
+        if args.ro_access:
+            from ..tools import admin
+            LOG.warning('Grant read-only access to web user')
+            admin.grant_ro_access(args.config.get_libpq_dsn(), args.config)
+
         if args.data_object or args.data_area:
             with connect(args.config.get_libpq_dsn()) as conn:
                 for obj in args.data_object or []:

src/nominatim_db/clicmd/setup.py

@@ -23,6 +23,7 @@ from ..tokenizer.base import AbstractTokenizer
 from ..version import NOMINATIM_VERSION
 from .args import NominatimArgs
 
+import time
 
 LOG = logging.getLogger()
 
@@ -86,6 +87,8 @@ class SetupAll:
         from ..tools import database_import, postcodes, freeze
         from ..indexer.indexer import Indexer
 
+        start_time = time.time()
+
         num_threads = args.threads or psutil.cpu_count() or 1
         country_info.setup_country_config(args.config)
 
@@ -138,6 +141,10 @@ class SetupAll:
         LOG.warning('Recompute word counts')
         tokenizer.update_statistics(args.config, threads=num_threads)
 
+        end_time = time.time()
+        elapsed = end_time - start_time
+        LOG.warning(f'Import completed successfully in {elapsed:.2f} seconds.')
+
         self._finalize_database(args.config.get_libpq_dsn(), args.offline)
 
         return 0

src/nominatim_db/config.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2025 by the Nominatim developer community.
+# Copyright (C) 2026 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Nominatim configuration accessor.
@@ -12,6 +12,7 @@ import importlib.util
 import logging
 import os
 import sys
+import re
 from pathlib import Path
 import json
 import yaml
@@ -80,6 +81,10 @@ class Configuration:
         self.lib_dir = _LibDirs()
         self._private_plugins: Dict[str, object] = {}
 
+        if re.fullmatch(r'[\w-]+', self.DATABASE_WEBUSER) is None:
+            raise UsageError("Misconfigured DATABASE_WEBUSER. "
+                             "Only alphnumberic characters, - and _ are allowed.")
+
     def set_libdirs(self, **kwargs: StrPath) -> None:
         """ Set paths to library functions and data.
         """
@@ -197,7 +202,7 @@ class Configuration:
         if dsn.startswith('pgsql:'):
             return dict((p.split('=', 1) for p in dsn[6:].split(';')))
 
-        return conninfo_to_dict(dsn)
+        return conninfo_to_dict(dsn)  # type: ignore
 
     def get_import_style_file(self) -> Path:
         """ Return the import style file as a path object. Translates the

src/nominatim_db/data/postcode_format.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Functions for formatting postcodes according to their country-specific
@@ -29,6 +29,9 @@ class CountryPostcodeMatcher:
         self.norm_pattern = re.compile(f'\\s*(?:{country_code.upper()}[ -]?)?({pc_pattern})\\s*')
         self.pattern = re.compile(pc_pattern)
 
+        # We want to exclude 0000, 00-000, 000 00 etc
+        self.zero_pattern = re.compile(r'^[0\- ]+$')
+
         self.output = config.get('output', r'\g<0>')
 
     def match(self, postcode: str) -> Optional[Match[str]]:
@@ -40,7 +43,10 @@ class CountryPostcodeMatcher:
         normalized = self.norm_pattern.fullmatch(postcode.upper())
 
         if normalized:
-            return self.pattern.fullmatch(normalized.group(1))
+            match = self.pattern.fullmatch(normalized.group(1))
+            if match and self.zero_pattern.match(match.string):
+                return None
+            return match
 
         return None
 
@@ -61,12 +67,15 @@ class PostcodeFormatter:
         self.country_without_postcode: Set[Optional[str]] = {None}
         self.country_matcher = {}
         self.default_matcher = CountryPostcodeMatcher('', {'pattern': '.*'})
+        self.postcode_extent: dict[Optional[str], int] = {}
 
         for ccode, prop in country_info.iterate('postcode'):
             if prop is False:
                 self.country_without_postcode.add(ccode)
             elif isinstance(prop, dict):
                 self.country_matcher[ccode] = CountryPostcodeMatcher(ccode, prop)
+                if 'extent' in prop:
+                    self.postcode_extent[ccode] = int(prop['extent'])
             else:
                 raise UsageError(f"Invalid entry 'postcode' for country '{ccode}'")
 
@@ -107,3 +116,9 @@ class PostcodeFormatter:
             `match()`
         """
         return self.country_matcher.get(country_code, self.default_matcher).normalize(match)
+
+    def get_postcode_extent(self, country_code: Optional[str]) -> int:
+        """ Return the extent (in m) to use for the given country. If no
+            specific extent is set, then the default of 5km will be returned.
+        """
+        return self.postcode_extent.get(country_code, 5000)

src/nominatim_db/db/query_pool.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2026 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 A connection pool that executes incoming queries in parallel.
@@ -27,20 +27,28 @@ class QueryPool:
         The results of the queries is discarded.
     """
     def __init__(self, dsn: str, pool_size: int = 1, **conn_args: Any) -> None:
+        self.is_cancelled = False
         self.wait_time = 0.0
         self.query_queue: 'asyncio.Queue[QueueItem]' = asyncio.Queue(maxsize=2 * pool_size)
 
-        self.pool = [asyncio.create_task(self._worker_loop(dsn, **conn_args))
+        self.pool = [asyncio.create_task(self._worker_loop_cancellable(dsn, **conn_args))
                      for _ in range(pool_size)]
 
     async def put_query(self, query: psycopg.abc.Query, params: Any) -> None:
         """ Schedule a query for execution.
         """
+        if self.is_cancelled:
+            await self.finish()
+            return
+
         tstart = time.time()
         await self.query_queue.put((query, params))
         self.wait_time += time.time() - tstart
         await asyncio.sleep(0)
 
+        if self.is_cancelled:
+            await self.finish()
+
     async def finish(self) -> None:
         """ Wait for all queries to finish and close the pool.
         """
@@ -56,6 +64,25 @@ class QueryPool:
             if excp is not None:
                 raise excp
 
+    def clear_queue(self) -> None:
+        """ Drop all items silently that might still be queued.
+        """
+        try:
+            while True:
+                self.query_queue.get_nowait()
+        except asyncio.QueueEmpty:
+            pass  # expected
+
+    async def _worker_loop_cancellable(self, dsn: str, **conn_args: Any) -> None:
+        try:
+            await self._worker_loop(dsn, **conn_args)
+        except Exception as e:
+            # Make sure the exception is forwarded to the main function
+            self.is_cancelled = True
+            # clear the queue here to ensure that any put() that may be blocked returns
+            self.clear_queue()
+            raise e
+
     async def _worker_loop(self, dsn: str, **conn_args: Any) -> None:
         conn_args['autocommit'] = True
         aconn = await psycopg.AsyncConnection.connect(dsn, **conn_args)

src/nominatim_db/db/sql_preprocessor.py

@@ -2,12 +2,13 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2026 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Preprocessing of SQL files.
 """
 from typing import Set, Dict, Any, cast
+import re
 
 import jinja2
 
@@ -34,7 +35,9 @@ def _get_tables(conn: Connection) -> Set[str]:
     with conn.cursor() as cur:
         cur.execute("SELECT tablename FROM pg_tables WHERE schemaname = 'public'")
 
-        return set((row[0] for row in list(cur)))
+        # paranoia check: make sure we don't get table names that cause
+        # an SQL injection later
+        return {row[0] for row in list(cur) if re.fullmatch(r'\w+', row[0])}
 
 
 def _get_middle_db_format(conn: Connection, tables: Set[str]) -> str:

src/nominatim_db/indexer/indexer.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Main work horse for indexing (computing addresses) the database.
@@ -59,7 +59,7 @@ class Indexer:
                 if await self.index_by_rank(0, 4) > 0:
                     _analyze()
 
-                if await self.index_boundaries(0, 30) > 100:
+                if await self.index_boundaries() > 100:
                     _analyze()
 
                 if await self.index_by_rank(5, 25) > 100:
@@ -74,7 +74,7 @@ class Indexer:
                 if not self.has_pending():
                     break
 
-    async def index_boundaries(self, minrank: int, maxrank: int) -> int:
+    async def index_boundaries(self, minrank: int = 0, maxrank: int = 30) -> int:
         """ Index only administrative boundaries within the given rank range.
         """
         total = 0
@@ -154,7 +154,7 @@ class Indexer:
         return total
 
     async def index_postcodes(self) -> int:
-        """Index the entries of the location_postcode table.
+        """Index the entries of the location_postcodes table.
         """
         LOG.warning("Starting indexing postcodes using %s threads", self.num_threads)
 
@@ -177,7 +177,7 @@ class Indexer:
 
             `total_tuples` may contain the total number of rows to process.
             When not supplied, the value will be computed using the
-            approriate runner function.
+            appropriate runner function.
         """
         LOG.warning("Starting %s (using batch size %s)", runner.name(), batch)
 

src/nominatim_db/indexer/runners.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Mix-ins that provide the actual commands for the indexer for various indexing
@@ -143,22 +143,22 @@ class InterpolationRunner:
 
 
 class PostcodeRunner(Runner):
-    """ Provides the SQL commands for indexing the location_postcode table.
+    """ Provides the SQL commands for indexing the location_postcodes table.
     """
 
     def name(self) -> str:
-        return "postcodes (location_postcode)"
+        return "postcodes (location_postcodes)"
 
     def sql_count_objects(self) -> Query:
-        return 'SELECT count(*) FROM location_postcode WHERE indexed_status > 0'
+        return 'SELECT count(*) FROM location_postcodes WHERE indexed_status > 0'
 
     def sql_get_objects(self) -> Query:
-        return """SELECT place_id FROM location_postcode
+        return """SELECT place_id FROM location_postcodes
                   WHERE indexed_status > 0
                   ORDER BY country_code, postcode"""
 
     def index_places_query(self, batch_size: int) -> Query:
-        return pysql.SQL("""UPDATE location_postcode SET indexed_status = 0
+        return pysql.SQL("""UPDATE location_postcodes SET indexed_status = 0
                                     WHERE place_id IN ({})""")\
                     .format(pysql.SQL(',').join((pysql.Placeholder() for _ in range(batch_size))))
 

src/nominatim_db/tokenizer/base.py

@@ -71,7 +71,7 @@ class AbstractAnalyzer(ABC):
     @abstractmethod
     def update_postcodes_from_db(self) -> None:
         """ Update the tokenizer's postcode tokens from the current content
-            of the `location_postcode` table.
+            of the `location_postcodes` table.
         """
 
     @abstractmethod

src/nominatim_db/tokenizer/icu_tokenizer.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2025 by the Nominatim developer community.
+# Copyright (C) 2026 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Tokenizer implementing normalisation as used before Nominatim 4 but using
@@ -294,13 +294,12 @@ class ICUTokenizer(AbstractTokenizer):
         with connect(self.dsn) as conn:
             drop_tables(conn, 'word')
             with conn.cursor() as cur:
-                cur.execute(f"ALTER TABLE {old} RENAME TO word")
-                for idx in ('word_token', 'word_id'):
-                    cur.execute(f"""ALTER INDEX idx_{old}_{idx}
-                                      RENAME TO idx_word_{idx}""")
-                for name, _ in WORD_TYPES:
-                    cur.execute(f"""ALTER INDEX idx_{old}_{name}
-                                    RENAME TO idx_word_{name}""")
+                cur.execute(pysql.SQL("ALTER TABLE {} RENAME TO word")
+                                 .format(pysql.Identifier(old)))
+                for idx in ['word_token', 'word_id'] + [n[0] for n in WORD_TYPES]:
+                    cur.execute(pysql.SQL("ALTER INDEX {} RENAME TO {}")
+                                     .format(pysql.Identifier(f"idx_{old}_{idx}"),
+                                             pysql.Identifier(f"idx_word_{idx}")))
             conn.commit()
 
 
@@ -475,20 +474,23 @@ class ICUNameAnalyzer(AbstractAnalyzer):
         assert self.conn is not None
         word_tokens = set()
         for name in names:
-            norm_name = self._search_normalized(name.name)
-            if norm_name:
-                word_tokens.add(norm_name)
+            norm_name = self._normalized(name.name)
+            token_name = self._search_normalized(name.name)
+            if norm_name and token_name:
+                word_tokens.add((token_name, norm_name))
 
         with self.conn.cursor() as cur:
             # Get existing names
-            cur.execute("""SELECT word_token, coalesce(info ? 'internal', false) as is_internal
+            cur.execute("""SELECT word_token,
+                                  word as lookup,
+                                  coalesce(info ? 'internal', false) as is_internal
                              FROM word
-                             WHERE type = 'C' and word = %s""",
+                             WHERE type = 'C' and info->>'cc' = %s""",
                         (country_code, ))
             # internal/external names
-            existing_tokens: Dict[bool, Set[str]] = {True: set(), False: set()}
+            existing_tokens: Dict[bool, Set[Tuple[str, str]]] = {True: set(), False: set()}
             for word in cur:
-                existing_tokens[word[1]].add(word[0])
+                existing_tokens[word[2]].add((word[0], word[1]))
 
             # Delete names that no longer exist.
             gone_tokens = existing_tokens[internal] - word_tokens
@@ -496,10 +498,10 @@ class ICUNameAnalyzer(AbstractAnalyzer):
                 gone_tokens.update(existing_tokens[False] & word_tokens)
             if gone_tokens:
                 cur.execute("""DELETE FROM word
-                               USING unnest(%s::text[]) as token
-                               WHERE type = 'C' and word = %s
-                                     and word_token = token""",
-                            (list(gone_tokens), country_code))
+                               USING jsonb_array_elements(%s) as data
+                               WHERE type = 'C' and info->>'cc' = %s
+                                     and word_token = data->>0 and word = data->>1""",
+                            (Jsonb(list(gone_tokens)), country_code))
 
             # Only add those names that are not yet in the list.
             new_tokens = word_tokens - existing_tokens[True]
@@ -508,15 +510,17 @@ class ICUNameAnalyzer(AbstractAnalyzer):
             if new_tokens:
                 if internal:
                     sql = """INSERT INTO word (word_token, type, word, info)
-                               (SELECT token, 'C', %s, '{"internal": "yes"}'
-                                  FROM unnest(%s::text[]) as token)
+                               (SELECT data->>0, 'C', data->>1,
+                                       jsonb_build_object('internal', 'yes', 'cc', %s::text)
+                                  FROM jsonb_array_elements(%s) as data)
                            """
                 else:
-                    sql = """INSERT INTO word (word_token, type, word)
-                                   (SELECT token, 'C', %s
-                                    FROM unnest(%s::text[]) as token)
+                    sql = """INSERT INTO word (word_token, type, word, info)
+                                   (SELECT data->>0, 'C', data->>1,
+                                           jsonb_build_object('cc', %s::text)
+                                    FROM  jsonb_array_elements(%s) as data)
                           """
-                cur.execute(sql, (country_code, list(new_tokens)))
+                cur.execute(sql, (country_code, Jsonb(list(new_tokens))))
 
     def process_place(self, place: PlaceInfo) -> Mapping[str, Any]:
         """ Determine tokenizer information about the given place.

src/nominatim_db/tools/admin.py

@@ -16,6 +16,7 @@ from psycopg.types.json import Json
 from ..typing import DictCursorResult
 from ..config import Configuration
 from ..db.connection import connect, Cursor, register_hstore
+from ..db.sql_preprocessor import SQLPreprocessor
 from ..errors import UsageError
 from ..tokenizer import factory as tokenizer_factory
 from ..data.place_info import PlaceInfo
@@ -105,3 +106,12 @@ def clean_deleted_relations(config: Configuration, age: str) -> None:
             except psycopg.DataError as exc:
                 raise UsageError('Invalid PostgreSQL time interval format') from exc
         conn.commit()
+
+
+def grant_ro_access(dsn: str, config: Configuration) -> None:
+    """ Grant read-only access to the web user for all Nominatim tables.
+        This can be used to grant access to a different user after import.
+    """
+    with connect(dsn) as conn:
+        sql = SQLPreprocessor(conn, config)
+        sql.run_sql_file(conn, 'grants.sql')

src/nominatim_db/tools/check_database.py

@@ -113,8 +113,8 @@ def _get_indexes(conn: Connection) -> List[str]:
                'idx_placex_geometry_placenode',
                'idx_osmline_parent_place_id',
                'idx_osmline_parent_osm_id',
-               'idx_postcode_id',
-               'idx_postcode_postcode'
+               'idx_location_postcodes_id',
+               'idx_location_postcodes_postcode'
                ]
 
     # These won't exist if --reverse-only import was used

src/nominatim_db/tools/database_import.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Functions for setting up and importing a new Nominatim database.
@@ -157,6 +157,8 @@ def create_tables(conn: Connection, config: Configuration, reverse_only: bool =
 
     sql.run_sql_file(conn, 'tables.sql')
 
+    sql.run_sql_file(conn, 'grants.sql')
+
 
 def create_table_triggers(conn: Connection, config: Configuration) -> None:
     """ Create the triggers for the tables. The trigger functions must already
@@ -183,7 +185,7 @@ def truncate_data_tables(conn: Connection) -> None:
         cur.execute('TRUNCATE location_area_country')
         cur.execute('TRUNCATE location_property_tiger')
         cur.execute('TRUNCATE location_property_osmline')
-        cur.execute('TRUNCATE location_postcode')
+        cur.execute('TRUNCATE location_postcodes')
         if table_exists(conn, 'search_name'):
             cur.execute('TRUNCATE search_name')
         cur.execute('DROP SEQUENCE IF EXISTS seq_place')
@@ -193,7 +195,7 @@ def truncate_data_tables(conn: Connection) -> None:
                        WHERE tablename LIKE 'location_road_%'""")
 
         for table in [r[0] for r in list(cur)]:
-            cur.execute('TRUNCATE ' + table)
+            cur.execute(pysql.SQL('TRUNCATE {}').format(pysql.Identifier(table)))
 
     conn.commit()
 
@@ -225,7 +227,7 @@ async def load_data(dsn: str, threads: int) -> None:
                                       total=pysql.Literal(placex_threads),
                                       mod=pysql.Literal(imod)), None)
 
-        # Interpolations need to be copied seperately
+        # Interpolations need to be copied separately
         await pool.put_query("""
                 INSERT INTO location_property_osmline (osm_id, address, linegeo)
                   SELECT osm_id, address, geometry FROM place

src/nominatim_db/tools/freeze.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Functions for removing unnecessary data from the database.
@@ -18,10 +18,11 @@ UPDATE_TABLES = [
     'address_levels',
     'gb_postcode',
     'import_osmosis_log',
-    'import_polygon_%',
     'location_area%',
     'location_road%',
     'place',
+    'place_entrance',
+    'place_postcode',
     'planet_osm_%',
     'search_name_%',
     'us_postcode',

src/nominatim_db/tools/migration.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2024 by the Nominatim developer community.
+# Copyright (C) 2025 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Functions for database migration to newer software versions.
@@ -18,6 +18,8 @@ from ..db.connection import connect, Connection, \
 from ..db.sql_preprocessor import SQLPreprocessor
 from ..version import NominatimVersion, NOMINATIM_VERSION, parse_version
 from ..tokenizer import factory as tokenizer_factory
+from ..data.country_info import create_country_names, setup_country_config
+from .freeze import is_frozen
 from . import refresh
 
 LOG = logging.getLogger()
@@ -27,7 +29,7 @@ _MIGRATION_FUNCTIONS: List[Tuple[NominatimVersion, Callable[..., None]]] = []
 
 def migrate(config: Configuration, paths: Any) -> int:
     """ Check for the current database version and execute migrations,
-        if necesssary.
+        if necessary.
     """
     with connect(config.get_libpq_dsn()) as conn:
         register_hstore(conn)
@@ -141,7 +143,7 @@ def create_placex_entrance_table(conn: Connection, config: Configuration, **_: A
 
 @_migration(5, 1, 99, 1)
 def create_place_entrance_table(conn: Connection, config: Configuration, **_: Any) -> None:
-    """ Add the place_entrance table to store incomming entrance nodes
+    """ Add the place_entrance table to store incoming entrance nodes
     """
     if not table_exists(conn, 'place_entrance'):
         with conn.cursor() as cur:
@@ -156,3 +158,195 @@ def create_place_entrance_table(conn: Connection, config: Configuration, **_: An
             CREATE UNIQUE INDEX place_entrance_osm_id_idx ON place_entrance
               USING BTREE (osm_id);
               """)
+
+
+@_migration(5, 2, 99, 0)
+def convert_country_tokens(conn: Connection, config: Configuration, **_: Any) -> None:
+    """ Convert country word tokens
+
+        Country tokens now save the country in the info field instead of the
+        word. This migration removes all country tokens from the word table
+        and reimports the default country name. This means that custom names
+        are lost. If you need them back, invalidate the OSM objects containing
+        the names by setting indexed_status to 2 and then reindex the database.
+    """
+    tokenizer = tokenizer_factory.get_tokenizer_for_db(config)
+    # There is only one tokenizer at the time of migration, so we make
+    # some assumptions here about the structure of the database. This will
+    # fail if somebody has written a custom tokenizer.
+    with conn.cursor() as cur:
+        cur.execute("DELETE FROM word WHERE type = 'C'")
+    conn.commit()
+
+    setup_country_config(config)
+    create_country_names(conn, tokenizer, config.get_str_list('LANGUAGES'))
+
+
+@_migration(5, 2, 99, 1)
+def create_place_postcode_table(conn: Connection, config: Configuration, **_: Any) -> None:
+    """ Restructure postcode tables
+    """
+    sqlp = SQLPreprocessor(conn, config)
+    mutable = not is_frozen(conn)
+    has_place_table = table_exists(conn, 'place_postcode')
+    has_postcode_table = table_exists(conn, 'location_postcodes')
+    if mutable and not has_place_table:
+        with conn.cursor() as cur:
+            cur.execute(
+                """
+                CREATE TABLE place_postcode (
+                    osm_type CHAR(1) NOT NULL,
+                    osm_id BIGINT NOT NULL,
+                    postcode TEXT NOT NULL,
+                    country_code TEXT,
+                    centroid GEOMETRY(Point, 4326) NOT NULL,
+                    geometry GEOMETRY(Geometry, 4326)
+                )
+                """)
+            # Move postcode points into the new table
+            cur.execute("ALTER TABLE place DISABLE TRIGGER ALL")
+            cur.execute(
+                """
+                WITH deleted AS (
+                  DELETE FROM place
+                  WHERE (class = 'place' AND type = 'postcode')
+                        OR (osm_type = 'R'
+                            AND class = 'boundary' AND type = 'postal_code')
+                  RETURNING osm_type, osm_id, address->'postcode' as postcode,
+                            ST_Centroid(geometry) as centroid,
+                            (CASE WHEN class = 'place' THEN NULL ELSE geometry END) as geometry)
+                INSERT INTO place_postcode (osm_type, osm_id, postcode, centroid, geometry)
+                    (SELECT * FROM deleted
+                     WHERE deleted.postcode is not NULL AND deleted.centroid is not NULL)
+                """)
+            cur.execute(
+                """
+                CREATE INDEX place_postcode_osm_id_idx ON place_postcode
+                  USING BTREE (osm_type, osm_id)
+                """)
+            cur.execute(
+                """
+                CREATE INDEX place_postcode_postcode_idx ON place_postcode
+                  USING BTREE (postcode)
+                """)
+            cur.execute("ALTER TABLE place ENABLE TRIGGER ALL")
+    if not has_postcode_table:
+        sqlp.run_sql_file(conn, 'functions/postcode_triggers.sql')
+        with conn.cursor() as cur:
+            # create a new location_postcode table which will replace the
+            # old one atomically in the end
+            cur.execute(
+                """
+                CREATE TABLE location_postcodes (
+                    place_id BIGINT,
+                    osm_id BIGINT,
+                    rank_search SMALLINT,
+                    parent_place_id BIGINT,
+                    indexed_status SMALLINT,
+                    indexed_date TIMESTAMP,
+                    country_code VARCHAR(2),
+                    postcode TEXT,
+                    centroid Geometry(Point, 4326),
+                    geometry Geometry(Geometry, 4326) NOT NULL
+                )
+                """)
+            sqlp.run_string(conn,
+                            'GRANT SELECT ON location_postcodes TO "{{config.DATABASE_WEBUSER}}"')
+            # remove postcodes from the various auxiliary tables
+            cur.execute(
+                """
+                DELETE FROM place_addressline
+                  WHERE address_place_id = ANY(
+                    SELECT place_id FROM placex
+                      WHERE osm_type = 'R'
+                            AND class = 'boundary' AND type = 'postal_code')
+                """)
+            if mutable:
+                cur.execute(
+                    """
+                    SELECT deleteLocationArea(partition, place_id, rank_search),
+                           deleteSearchName(partition, place_id)
+                      FROM placex
+                      WHERE osm_type = 'R' AND class = 'boundary' AND type = 'postal_code'
+                    """)
+            if table_exists(conn, 'search_name'):
+                cur.execute(
+                    """
+                    DELETE FROM search_name
+                      WHERE place_id = ANY(
+                        SELECT place_id FROM placex
+                          WHERE osm_type = 'R'
+                                AND class = 'boundary' AND type = 'postal_code')
+                    """)
+            # move postcode areas from placex to location_postcodes
+            # avoiding automatic invalidation
+            cur.execute("ALTER TABLE placex DISABLE TRIGGER ALL")
+            cur.execute(
+                """
+                WITH deleted AS (
+                    DELETE FROM placex
+                      WHERE osm_type = 'R'
+                            AND class = 'boundary' AND type = 'postal_code'
+                      RETURNING place_id, osm_id, rank_search, parent_place_id,
+                                indexed_status, indexed_date,
+                                country_code, postcode, centroid, geometry)
+                INSERT INTO location_postcodes (SELECT * from deleted)
+                """)
+            cur.execute("ALTER TABLE placex ENABLE TRIGGER ALL")
+            # remove any old postcode centroid that would overlap with areas
+            cur.execute(
+                """
+                DELETE FROM location_postcode o USING location_postcodes n
+                   WHERE o.country_code = n.country_code
+                         AND o.postcode = n.postcode
+                """)
+            # copy over old postcodes
+            cur.execute(
+                """
+                INSERT INTO location_postcodes
+                    (SELECT place_id, NULL, rank_search, parent_place_id,
+                            indexed_status, indexed_date, country_code,
+                            postcode, geometry,
+                            ST_Expand(geometry, 0.05)
+                     FROM location_postcode)
+                """)
+            # add indexes and triggers
+            cur.execute("""CREATE INDEX idx_location_postcodes_geometry
+                           ON location_postcodes USING GIST(geometry)""")
+            cur.execute("""CREATE INDEX idx_location_postcodes_id
+                           ON location_postcodes USING BTREE(place_id)""")
+            cur.execute("""CREATE INDEX idx_location_postcodes_osmid
+                           ON location_postcodes USING BTREE(osm_id)""")
+            cur.execute("""CREATE INDEX idx_location_postcodes_postcode
+                           ON location_postcodes USING BTREE(postcode, country_code)""")
+            cur.execute("""CREATE INDEX idx_location_postcodes_parent_place_id
+                           ON location_postcodes USING BTREE(parent_place_id)""")
+            cur.execute("""CREATE TRIGGER location_postcodes_before_update
+                           BEFORE UPDATE ON location_postcodes
+                           FOR EACH ROW EXECUTE PROCEDURE postcodes_update()""")
+            cur.execute("""CREATE TRIGGER location_postcodes_before_delete
+                           BEFORE DELETE ON location_postcodes
+                           FOR EACH ROW EXECUTE PROCEDURE postcodes_delete()""")
+            cur.execute("""CREATE TRIGGER location_postcodes_before_insert
+                           BEFORE INSERT ON location_postcodes
+                           FOR EACH ROW EXECUTE PROCEDURE postcodes_insert()""")
+        sqlp.run_string(
+            conn,
+            """
+            CREATE INDEX IF NOT EXISTS idx_placex_geometry_reverse_lookupPolygon_nopostcode
+            ON placex USING gist (geometry) {{db.tablespace.search_index}}
+            WHERE St_GeometryType(geometry) in ('ST_Polygon', 'ST_MultiPolygon')
+            AND rank_address between 4 and 25
+            AND name is not null AND linked_place_id is null;
+
+            CREATE INDEX IF NOT EXISTS idx_placex_geometry_reverse_lookupPlaceNode_nopostcode
+              ON placex USING gist (ST_Buffer(geometry, reverse_place_diameter(rank_search)))
+              {{db.tablespace.search_index}}
+              WHERE rank_address between 4 and 25
+                AND name is not null AND linked_place_id is null AND osm_type = 'N';
+
+            CREATE INDEX idx_placex_geometry_placenode_nopostcode ON placex
+              USING SPGIST (geometry) {{db.tablespace.address_index}}
+              WHERE osm_type = 'N' and rank_search < 26 and class = 'place';
+            ANALYSE;
+            """)

src/nominatim_db/tools/postcodes.py

@@ -8,7 +8,7 @@
 Functions for importing, updating and otherwise maintaining the table
 of artificial postcode centroids.
 """
-from typing import Optional, Tuple, Dict, List, TextIO
+from typing import Optional, Tuple, Dict, TextIO
 from collections import defaultdict
 from pathlib import Path
 import csv
@@ -38,13 +38,26 @@ def _to_float(numstr: str, max_value: float) -> float:
     return num
 
 
+def _extent_to_rank(extent: int) -> int:
+    """ Guess a suitable search rank from the extent of a postcode.
+    """
+    if extent <= 100:
+        return 25
+    if extent <= 3000:
+        return 23
+    return 21
+
+
 class _PostcodeCollector:
     """ Collector for postcodes of a single country.
     """
 
-    def __init__(self, country: str, matcher: Optional[CountryPostcodeMatcher]):
+    def __init__(self, country: str, matcher: Optional[CountryPostcodeMatcher],
+                 default_extent: int, exclude: set[str] = set()):
         self.country = country
         self.matcher = matcher
+        self.extent = default_extent
+        self.exclude = exclude
         self.collected: Dict[str, PointsCentroid] = defaultdict(PointsCentroid)
         self.normalization_cache: Optional[Tuple[str, Optional[str]]] = None
 
@@ -61,7 +74,7 @@ class _PostcodeCollector:
                 normalized = self.matcher.normalize(match) if match else None
                 self.normalization_cache = (postcode, normalized)
 
-            if normalized:
+            if normalized and normalized not in self.exclude:
                 self.collected[normalized] += (x, y)
 
     def commit(self, conn: Connection, analyzer: AbstractAnalyzer,
@@ -73,61 +86,38 @@ class _PostcodeCollector:
         """
         if project_dir is not None:
             self._update_from_external(analyzer, project_dir)
-        to_add, to_delete, to_update = self._compute_changes(conn)
 
-        LOG.info("Processing country '%s' (%s added, %s deleted, %s updated).",
-                 self.country, len(to_add), len(to_delete), len(to_update))
+        with conn.cursor() as cur:
+            cur.execute("""SELECT postcode FROM location_postcodes
+                           WHERE country_code = %s AND osm_id is null""",
+                        (self.country, ))
+            to_delete = [row[0] for row in cur if row[0] not in self.collected]
+
+        to_add = [dict(zip(('pc', 'x', 'y'), (k, *v.centroid())))
+                  for k, v in self.collected.items()]
+        self.collected = defaultdict(PointsCentroid)
+
+        LOG.info("Processing country '%s' (%s added, %s deleted).",
+                 self.country, len(to_add), len(to_delete))
 
         with conn.cursor() as cur:
             if to_add:
                 cur.executemany(pysql.SQL(
-                    """INSERT INTO location_postcode
-                         (place_id, indexed_status, country_code,
-                          postcode, geometry)
-                       VALUES (nextval('seq_place'), 1, {}, %s,
-                               ST_SetSRID(ST_MakePoint(%s, %s), 4326))
-                    """).format(pysql.Literal(self.country)),
+                    """INSERT INTO location_postcodes
+                         (country_code, rank_search, postcode, centroid, geometry)
+                       VALUES ({}, {}, %(pc)s,
+                               ST_SetSRID(ST_MakePoint(%(x)s, %(y)s), 4326),
+                               expand_by_meters(ST_SetSRID(ST_MakePoint(%(x)s, %(y)s), 4326), {}))
+                    """).format(pysql.Literal(self.country),
+                                pysql.Literal(_extent_to_rank(self.extent)),
+                                pysql.Literal(self.extent)),
                     to_add)
             if to_delete:
-                cur.execute("""DELETE FROM location_postcode
+                cur.execute("""DELETE FROM location_postcodes
                                WHERE country_code = %s and postcode = any(%s)
+                                     AND osm_id is null
                             """, (self.country, to_delete))
-            if to_update:
-                cur.executemany(
-                    pysql.SQL("""UPDATE location_postcode
-                                 SET indexed_status = 2,
-                                     geometry = ST_SetSRID(ST_Point(%s, %s), 4326)
-                                 WHERE country_code = {} and postcode = %s
-                              """).format(pysql.Literal(self.country)),
-                    to_update)
-
-    def _compute_changes(
-            self, conn: Connection
-            ) -> Tuple[List[Tuple[str, float, float]], List[str], List[Tuple[float, float, str]]]:
-        """ Compute which postcodes from the collected postcodes have to be
-            added or modified and which from the location_postcode table
-            have to be deleted.
-        """
-        to_update = []
-        to_delete = []
-        with conn.cursor() as cur:
-            cur.execute("""SELECT postcode, ST_X(geometry), ST_Y(geometry)
-                           FROM location_postcode
-                           WHERE country_code = %s""",
-                        (self.country, ))
-            for postcode, x, y in cur:
-                pcobj = self.collected.pop(postcode, None)
-                if pcobj:
-                    newx, newy = pcobj.centroid()
-                    if abs(x - newx) > 0.0000001 or abs(y - newy) > 0.0000001:
-                        to_update.append((newx, newy, postcode))
-                else:
-                    to_delete.append(postcode)
-
-        to_add = [(k, *v.centroid()) for k, v in self.collected.items()]
-        self.collected = defaultdict(PointsCentroid)
-
-        return to_add, to_delete, to_update
+            cur.execute("ANALYSE location_postcodes")
 
     def _update_from_external(self, analyzer: AbstractAnalyzer, project_dir: Path) -> None:
         """ Look for an external postcode file for the active country in
@@ -152,7 +142,7 @@ class _PostcodeCollector:
                                     _to_float(row['lat'], 90))
                         self.collected[postcode] += centroid
                     except ValueError:
-                        LOG.warning("Bad coordinates %s, %s in %s country postcode file.",
+                        LOG.warning("Bad coordinates %s, %s in '%s' country postcode file.",
                                     row['lat'], row['lon'], self.country)
 
         finally:
@@ -169,69 +159,165 @@ class _PostcodeCollector:
 
         if fname.is_file():
             LOG.info("Using external postcode file '%s'.", fname)
-            return gzip.open(fname, 'rt')
+            return gzip.open(fname, 'rt', encoding='utf-8')
 
         return None
 
 
 def update_postcodes(dsn: str, project_dir: Optional[Path], tokenizer: AbstractTokenizer) -> None:
-    """ Update the table of artificial postcodes.
-
-        Computes artificial postcode centroids from the placex table,
-        potentially enhances it with external data and then updates the
-        postcodes in the table 'location_postcode'.
+    """ Update the table of postcodes from the input tables
+        placex and place_postcode.
     """
     matcher = PostcodeFormatter()
     with tokenizer.name_analyzer() as analyzer:
         with connect(dsn) as conn:
-            # First get the list of countries that currently have postcodes.
-            # (Doing this before starting to insert, so it is fast on import.)
-            with conn.cursor() as cur:
-                cur.execute("SELECT DISTINCT country_code FROM location_postcode")
-                todo_countries = set((row[0] for row in cur))
-
-            # Recompute the list of valid postcodes from placex.
-            with conn.cursor(name="placex_postcodes") as cur:
-                cur.execute("""
-                SELECT cc, pc, ST_X(centroid), ST_Y(centroid)
-                FROM (SELECT
-                        COALESCE(plx.country_code,
-                                 get_country_code(ST_Centroid(pl.geometry))) as cc,
-                        pl.address->'postcode' as pc,
-                        COALESCE(plx.centroid, ST_Centroid(pl.geometry)) as centroid
-                      FROM place AS pl LEFT OUTER JOIN placex AS plx
-                             ON pl.osm_id = plx.osm_id AND pl.osm_type = plx.osm_type
-                    WHERE pl.address ? 'postcode' AND pl.geometry IS NOT null) xx
-                WHERE pc IS NOT null AND cc IS NOT null
-                ORDER BY cc, pc""")
-
-                collector = None
-
-                for country, postcode, x, y in cur:
-                    if collector is None or country != collector.country:
-                        if collector is not None:
-                            collector.commit(conn, analyzer, project_dir)
-                        collector = _PostcodeCollector(country, matcher.get_matcher(country))
-                        todo_countries.discard(country)
-                    collector.add(postcode, x, y)
-
-                if collector is not None:
-                    collector.commit(conn, analyzer, project_dir)
-
-            # Now handle any countries that are only in the postcode table.
-            for country in todo_countries:
-                fmt = matcher.get_matcher(country)
-                _PostcodeCollector(country, fmt).commit(conn, analyzer, project_dir)
-
+            # Backfill country_code column where required
+            conn.execute("""UPDATE place_postcode
+                              SET country_code = get_country_code(centroid)
+                              WHERE country_code is null
+                         """)
+            # Now update first postcode areas
+            _update_postcode_areas(conn, analyzer, matcher)
+            # Then fill with estimated postcode centroids from other info
+            _update_guessed_postcode(conn, analyzer, matcher, project_dir)
             conn.commit()
 
         analyzer.update_postcodes_from_db()
 
 
-def can_compute(dsn: str) -> bool:
+def _insert_postcode_areas(conn: Connection, country_code: str,
+                           extent: int, pcs: list[dict[str, str]]) -> None:
+    if pcs:
+        with conn.cursor() as cur:
+            cur.executemany(
+                pysql.SQL(
+                    """ INSERT INTO location_postcodes
+                            (osm_id, country_code, rank_search, postcode, centroid, geometry)
+                            SELECT osm_id, country_code, {}, %(out)s, centroid, geometry
+                            FROM place_postcode
+                            WHERE osm_type = 'R'
+                                  and country_code = {} and postcode = %(in)s
+                                  and geometry is not null
+                    """).format(pysql.Literal(_extent_to_rank(extent)),
+                                pysql.Literal(country_code)),
+                pcs)
+
+
+def _update_postcode_areas(conn: Connection, analyzer: AbstractAnalyzer,
+                           matcher: PostcodeFormatter) -> None:
+    """ Update the postcode areas made from postcode boundaries.
     """
-        Check that the place table exists so that
-        postcodes can be computed.
+    # first delete all areas that have gone
+    conn.execute(""" DELETE FROM location_postcodes pc
+                     WHERE pc.osm_id is not null
+                       AND NOT EXISTS(
+                              SELECT * FROM place_postcode pp
+                              WHERE pp.osm_type = 'R' and pp.osm_id = pc.osm_id
+                                    and geometry is not null)
+                """)
+    # now insert all in country batches, triggers will ensure proper updates
+    with conn.cursor() as cur:
+        cur.execute(""" SELECT country_code, postcode FROM place_postcode
+                        WHERE geometry is not null and osm_type = 'R'
+                        ORDER BY country_code
+                    """)
+        country_code = None
+        fmt = None
+        pcs = []
+        for cc, postcode in cur:
+            if country_code is None:
+                country_code = cc
+                fmt = matcher.get_matcher(country_code)
+            elif country_code != cc:
+                _insert_postcode_areas(conn, country_code,
+                                       matcher.get_postcode_extent(country_code), pcs)
+                country_code = cc
+                fmt = matcher.get_matcher(country_code)
+                pcs = []
+
+            if fmt is not None:
+                if (m := fmt.match(postcode)):
+                    pcs.append({'out': fmt.normalize(m), 'in': postcode})
+
+        if country_code is not None and pcs:
+            _insert_postcode_areas(conn, country_code,
+                                   matcher.get_postcode_extent(country_code), pcs)
+
+
+def _update_guessed_postcode(conn: Connection, analyzer: AbstractAnalyzer,
+                             matcher: PostcodeFormatter, project_dir: Optional[Path]) -> None:
+    """ Computes artificial postcode centroids from the placex table,
+        potentially enhances it with external data and then updates the
+        postcodes in the table 'location_postcodes'.
+    """
+    # First get the list of countries that currently have postcodes.
+    # (Doing this before starting to insert, so it is fast on import.)
+    with conn.cursor() as cur:
+        cur.execute("""SELECT DISTINCT country_code FROM location_postcodes
+                        WHERE osm_id is null""")
+        todo_countries = {row[0] for row in cur}
+
+    # Next, get the list of postcodes that are already covered by areas.
+    area_pcs = defaultdict(set)
+    with conn.cursor() as cur:
+        cur.execute("""SELECT country_code, postcode
+                       FROM location_postcodes WHERE osm_id is not null
+                       ORDER BY country_code""")
+        for cc, pc in cur:
+            area_pcs[cc].add(pc)
+
+    # Create a temporary table which contains coverage of the postcode areas.
+    with conn.cursor() as cur:
+        cur.execute("DROP TABLE IF EXISTS _global_postcode_area")
+        cur.execute("""CREATE TABLE _global_postcode_area AS
+                       (SELECT ST_SubDivide(ST_SimplifyPreserveTopology(
+                                              ST_Union(geometry), 0.00001), 128) as geometry
+                        FROM place_postcode WHERE geometry is not null)
+                    """)
+        cur.execute("CREATE INDEX ON _global_postcode_area USING gist(geometry)")
+    # Recompute the list of valid postcodes from placex.
+    with conn.cursor(name="placex_postcodes") as cur:
+        cur.execute("""
+            SELECT country_code, postcode, ST_X(centroid), ST_Y(centroid)
+              FROM (
+                (SELECT country_code, address->'postcode' as postcode, centroid
+                  FROM placex WHERE address ? 'postcode')
+                UNION
+                (SELECT country_code, postcode, centroid
+                 FROM place_postcode WHERE geometry is null)
+              ) x
+              WHERE not postcode like '%,%' and not postcode like '%;%'
+                    AND NOT EXISTS(SELECT * FROM _global_postcode_area g
+                                   WHERE ST_Intersects(x.centroid, g.geometry))
+              ORDER BY country_code""")
+
+        collector = None
+
+        for country, postcode, x, y in cur:
+            if collector is None or country != collector.country:
+                if collector is not None:
+                    collector.commit(conn, analyzer, project_dir)
+                collector = _PostcodeCollector(country, matcher.get_matcher(country),
+                                               matcher.get_postcode_extent(country),
+                                               exclude=area_pcs[country])
+                todo_countries.discard(country)
+            collector.add(postcode, x, y)
+
+        if collector is not None:
+            collector.commit(conn, analyzer, project_dir)
+
+    # Now handle any countries that are only in the postcode table.
+    for country in todo_countries:
+        fmt = matcher.get_matcher(country)
+        ext = matcher.get_postcode_extent(country)
+        _PostcodeCollector(country, fmt, ext,
+                           exclude=area_pcs[country]).commit(conn, analyzer, project_dir)
+
+    conn.execute("DROP TABLE IF EXISTS _global_postcode_area")
+
+
+def can_compute(dsn: str) -> bool:
+    """ Check that the necessary tables exist so that postcodes can be computed.
     """
     with connect(dsn) as conn:
-        return table_exists(conn, 'place')
+        return table_exists(conn, 'place_postcode')

src/nominatim_db/tools/refresh.py

@@ -141,7 +141,9 @@ def import_importance_csv(dsn: str, data_file: Path) -> int:
 
             copy_cmd = """COPY wikimedia_importance(language, title, importance, wikidata)
                           FROM STDIN"""
-            with gzip.open(str(data_file), 'rt') as fd, cur.copy(copy_cmd) as copy:
+            with gzip.open(
+                    str(data_file), 'rt', encoding='utf-8') as fd, \
+                    cur.copy(copy_cmd) as copy:
                 for row in csv.DictReader(fd, delimiter='\t', quotechar='|'):
                     wd_id = int(row['wikidata_id'][1:])
                     copy.write_row((row['language'],

src/nominatim_db/tools/special_phrases/sp_wiki_loader.py

@@ -11,6 +11,8 @@ from typing import Iterable
 import re
 import logging
 
+import mwparserfromhell
+
 from ...config import Configuration
 from ...utils.url_utils import get_url
 from .special_phrase import SpecialPhrase
@@ -36,10 +38,6 @@ class SPWikiLoader:
     """
     def __init__(self, config: Configuration) -> None:
         self.config = config
-        # Compile the regex here to increase performances.
-        self.occurence_pattern = re.compile(
-            r'\| *([^\|]+) *\|\| *([^\|]+) *\|\| *([^\|]+) *\|\| *([^\|]+) *\|\| *([\-YN])'
-        )
         # Hack around a bug where building=yes was imported with quotes into the wiki
         self.type_fix_pattern = re.compile(r'\"|"')
 
@@ -58,11 +56,21 @@ class SPWikiLoader:
             LOG.warning('Importing phrases for lang: %s...', lang)
             loaded_xml = _get_wiki_content(lang)
 
-            # One match will be of format [label, class, type, operator, plural]
-            matches = self.occurence_pattern.findall(loaded_xml)
+            wikicode = mwparserfromhell.parse(loaded_xml)
 
-            for match in matches:
-                yield SpecialPhrase(match[0],
-                                    match[1],
-                                    self.type_fix_pattern.sub('', match[2]),
-                                    match[3])
+            for table in wikicode.filter_tags(matches=lambda t: t.tag == 'table'):
+                for row in table.contents.filter_tags(matches=lambda t: t.tag == 'tr'):
+                    cells = list(row.contents.filter_tags(matches=lambda t: t.tag == 'td'))
+
+                    if len(cells) < 5:
+                        continue
+
+                    label = cells[0].contents.strip_code().strip()
+                    cls = cells[1].contents.strip_code().strip()
+                    typ = cells[2].contents.strip_code().strip()
+                    operator = cells[3].contents.strip_code().strip()
+
+                    yield SpecialPhrase(label,
+                                        cls,
+                                        self.type_fix_pattern.sub('', typ),
+                                        operator)

src/nominatim_db/tools/tiger_data.py

@@ -17,13 +17,12 @@ import tarfile
 from psycopg.types.json import Json
 
 from ..config import Configuration
-from ..db.connection import connect
+from ..db.connection import connect, table_exists
 from ..db.sql_preprocessor import SQLPreprocessor
 from ..errors import UsageError
 from ..db.query_pool import QueryPool
 from ..data.place_info import PlaceInfo
 from ..tokenizer.base import AbstractTokenizer
-from . import freeze
 
 LOG = logging.getLogger()
 
@@ -90,16 +89,19 @@ async def add_tiger_data(data_dir: str, config: Configuration, threads: int,
     """
     dsn = config.get_libpq_dsn()
 
-    with connect(dsn) as conn:
-        if freeze.is_frozen(conn):
-            raise UsageError("Tiger cannot be imported when database frozen (Github issue #3048)")
-
     with TigerInput(data_dir) as tar:
         if not tar:
             return 1
 
         with connect(dsn) as conn:
             sql = SQLPreprocessor(conn, config)
+
+            if not table_exists(conn, 'search_name'):
+                raise UsageError(
+                    "Cannot perform tiger import: required tables are missing. "
+                    "See https://github.com/osm-search/Nominatim/issues/2463 for details."
+                )
+
             sql.run_sql_file(conn, 'tiger_import_start.sql')
 
         # Reading files and then for each file line handling

src/nominatim_db/version.py

@@ -55,7 +55,7 @@ def parse_version(version: str) -> NominatimVersion:
     return NominatimVersion(*[int(x) for x in parts[:2] + parts[2].split('-')])
 
 
-NOMINATIM_VERSION = parse_version('5.2.0-0')
+NOMINATIM_VERSION = parse_version('5.2.99-2')
 
 POSTGRESQL_REQUIRED_VERSION = (12, 0)
 POSTGIS_REQUIRED_VERSION = (3, 0)

test/bdd/conftest.py

@@ -9,6 +9,7 @@ Fixtures for BDD test steps
 """
 import sys
 import json
+import re
 from pathlib import Path
 
 import psycopg
@@ -20,7 +21,8 @@ sys.path.insert(0, str(SRC_DIR / 'src'))
 
 import pytest
 from pytest_bdd.parsers import re as step_parse
-from pytest_bdd import given, when, then
+from pytest_bdd import given, when, then, scenario
+from pytest_bdd.feature import get_features
 
 pytest.register_assert_rewrite('utils')
 
@@ -373,3 +375,57 @@ def check_place_missing_lines(db_conn, table, osm_type, osm_id, osm_class):
 
     with db_conn.cursor() as cur:
         assert cur.execute(sql, params).fetchone()[0] == 0
+
+
+if pytest.version_tuple >= (8, 0, 0):
+    def pytest_pycollect_makemodule(module_path, parent):
+        return BddTestCollector.from_parent(parent, path=module_path)
+
+
+class BddTestCollector(pytest.Module):
+
+    def __init__(self, **kwargs):
+        super().__init__(**kwargs)
+
+    def collect(self):
+        for item in super().collect():
+            yield item
+
+        if hasattr(self.obj, 'PYTEST_BDD_SCENARIOS'):
+            for path in self.obj.PYTEST_BDD_SCENARIOS:
+                for feature in get_features([str(Path(self.path.parent, path).resolve())]):
+                    yield FeatureFile.from_parent(self,
+                                                  name=str(Path(path, feature.rel_filename)),
+                                                  path=Path(feature.filename),
+                                                  feature=feature)
+
+
+# borrowed from pytest-bdd: src/pytest_bdd/scenario.py
+def make_python_name(string: str) -> str:
+    """Make python attribute name out of a given string."""
+    string = re.sub(r"\W", "", string.replace(" ", "_"))
+    return re.sub(r"^\d+_*", "", string).lower()
+
+
+class FeatureFile(pytest.File):
+    class obj:
+        pass
+
+    def __init__(self, feature, **kwargs):
+        self.feature = feature
+        super().__init__(**kwargs)
+
+    def collect(self):
+        for sname, sobject in self.feature.scenarios.items():
+            class_name = f"L{sobject.line_number}"
+            test_name = "test_" + make_python_name(sname)
+
+            @scenario(self.feature.filename, sname)
+            def _test():
+                pass
+
+            tclass = type(class_name, (),
+                          {test_name: staticmethod(_test)})
+            setattr(self.obj, class_name, tclass)
+
+            yield pytest.Class.from_parent(self, name=class_name, obj=tclass)

test/bdd/features/api/lookup/simple.feature

@@ -42,6 +42,22 @@ Feature: Tests for finding places by osm_type and osm_id
         | jsonv2      | json        |
         | geojson     | geojson     |
 
+    Scenario Outline: Lookup with entrances
+        When sending v1/lookup with format <format>
+          | osm_ids    | entrances |
+          | W429210603 | 1         |
+        Then a HTTP 200 is returned
+        And the result is valid <outformat>
+        And result 0 contains in field entrances+0
+          | osm_id     | type | lat        | lon       |
+          | 6580031131 | yes  | 47.2489382 | 9.5284033 |
+
+        Examples:
+          | format | outformat |
+          | json   | json |
+          | jsonv2 | json |
+          | geojson | geojson |
+
     Scenario: Linked places return information from the linkee
         When sending v1/lookup with format geocodejson
           | osm_ids |

test/bdd/features/api/reverse/v1_params.feature

@@ -167,3 +167,18 @@ Feature: v1/reverse Parameter Tests
           | json   | json |
           | jsonv2 | json |
           | xml    | xml |
+
+    Scenario Outline: Reverse with entrances
+        When sending v1/reverse with format <format>
+          | lat               | lon              | entrances | zoom |
+          | 47.24942041089678 | 9.52854573737568 | 1         | 18   |
+        Then a HTTP 200 is returned
+        And the result is valid <outformat>
+        And the result contains array field entrances where element 0 contains
+          | osm_id     | type | lat        | lon       |
+          | 6580031131 | yes  | 47.2489382 | 9.5284033 |
+
+        Examples:
+          | format | outformat |
+          | json   | json |
+          | jsonv2 | json |

test/bdd/features/db/import/addressing.feature

@@ -268,33 +268,6 @@ Feature: Address computation
             | W93    | R34      |
             | W93    | R4       |
 
-    Scenario: postcode boundaries do appear in the address of a way
-       Given the grid with origin DE
-            | 1 |   |   |   |   | 8 |   | 6 |   | 2 |
-            |   |10 |11 |   |   |   |   |   |   |   |
-            |   |13 |12 |   |   |   |   |   |   |   |
-            | 20|   |   | 21|   |   |   |   |   |   |
-            |   |   |   |   |   |   |   |   |   |   |
-            |   |   |   |   |   | 9 |   |   |   |   |
-            | 4 |   |   |   |   |   |   | 7 |   | 3 |
-        And the named places
-            | osm | class    | type           | admin | addr+postcode | geometry    |
-            | R1  | boundary | administrative | 6     | 10000         | (1,2,3,4,1) |
-            | R34 | boundary | administrative | 8     | 11000         | (1,6,7,4,1) |
-        And the places
-            | osm | class    | type        | addr+postcode | geometry |
-            | R4  | boundary | postal_code | 11200         | (1,8,9,4,1) |
-        And the named places
-            | osm | class    | type           | geometry |
-            | W93 | highway  | residential    | 20,21    |
-        And the places
-            | osm | class    | type        | addr+postcode | geometry |
-            | W22 | place    | postcode    | 11234         | (10,11,12,13,10) |
-        When importing
-        Then place_addressline contains
-            | object | address |
-            | W93    | R4      |
-
     Scenario: squares do not appear in the address of a street
         Given the grid
             |   | 1 |   | 2 |   |

test/bdd/features/db/import/linking.feature

@@ -16,21 +16,6 @@ Feature: Linking of places
          | R13     | - |
          | N256    | - |
 
-    Scenario: Postcode areas cannot be linked
-        Given the grid with origin US
-         | 1 |   | 2 |
-         |   | 9 |   |
-         | 4 |   | 3 |
-        And the named places
-         | osm | class    | type        | addr+postcode  | extra+wikidata | geometry    |
-         | R13 | boundary | postal_code | 12345          | Q87493         | (1,2,3,4,1) |
-         | N25 | place    | suburb      | 12345          | Q87493         | 9 |
-        When importing
-        Then placex contains
-         | object | linked_place_id |
-         | R13    | - |
-         | N25    | - |
-
      Scenario: Waterways are linked when in waterway relations
         Given the grid
          | 1 |  |   |  | 3 | 4  |  |   |  | 6 |
@@ -321,11 +306,17 @@ Feature: Linking of places
         Given the places
             | osm | class    | type           | name+name                | geometry    |
             | N9  | place    | city           | Popayán                  | 9           |
-            | R1  | boundary | administrative | Perímetro Urbano Popayán | (1,2,3,4,1) |
+        Given the places
+            | osm | class    | type           | name+name                | geometry    | admin |
+            | R1  | boundary | administrative | Perímetro Urbano Popayán | (1,2,3,4,1) | 8     |
         And the relations
             | id | members  |
             | 1  | N9:label |
         When importing
+        Then placex contains
+            | object      | linked_place_id |
+            | N9:place    | R1              |
+            | R1:boundary | -               |
         Then placex contains
             | object | name+_place_name | name+_place_name:es |
             | R1     | Popayán          | Popayán             |

test/bdd/features/db/import/placex.feature

@@ -46,23 +46,6 @@ Feature: Import into placex
           | object | admin_level |
           | N1     | 3           |
 
-    Scenario: postcode node without postcode is dropped
-        Given the places
-          | osm | class   | type     | name+ref |
-          | N1  | place   | postcode | 12334    |
-        When importing
-        Then placex has no entry for N1
-
-    Scenario: postcode boundary without postcode is dropped
-        Given the 0.01 grid
-          | 1 | 2 |
-          | 3 |   |
-        Given the places
-          | osm | class    | type        | name+ref | geometry  |
-          | R1  | boundary | postal_code | 554476   | (1,2,3,1) |
-        When importing
-        Then placex has no entry for R1
-
     Scenario: search and address ranks for boundaries are correctly assigned
         Given the named places
           | osm | class    | type |

test/bdd/features/db/import/postcodes.feature

@@ -121,8 +121,8 @@ Feature: Import of postcodes
             |    | 1 | 2 |   |    |
             |    | 4 | 3 |   |    |
         And the named places
-            | osm | class    | type           | geometry         |
-            | W93 | highway  | pedestriant    | (10,11,12,13,10) |
+            | osm | class    | type       | geometry         |
+            | W93 | highway  | pedestrian | (10,11,12,13,10) |
         And the named places
             | osm | class    | type        | addr+postcode | geometry    |
             | W22 | building | yes         | 45023         | (1,2,3,4,1) |
@@ -134,14 +134,13 @@ Feature: Import of postcodes
     Scenario: Roads get postcodes from nearby unnamed buildings without other info
         Given the grid with origin US
             | 10 |   |   |   | 11 |
-            |    | 1 | 2 |   |    |
-            |    | 4 | 3 |   |    |
+            |    | 1 |   |   |    |
         And the named places
             | osm | class    | type           | geometry |
             | W93 | highway  | residential    | 10,11    |
-        And the places
-            | osm | class    | type        | addr+postcode | geometry    |
-            | W22 | place    | postcode    | 45023         | (1,2,3,4,1) |
+        And the postcodes
+            | osm | postcode | centroid |
+            | W22 | 45023    | 1        |
         When importing
         Then placex contains
             | object | postcode |
@@ -172,26 +171,12 @@ Feature: Import of postcodes
     Scenario: Postcodes are added to the postcode
         Given the places
            | osm | class | type  | addr+postcode | addr+housenumber | geometry |
-           | N34 | place | house | 01982         | 111              |country:de |
+           | N34 | place | house | 01982         | 111              | country:de |
         When importing
-        Then location_postcode contains exactly
-           | country_code | postcode | geometry!wkt |
+        Then location_postcodes contains exactly
+           | country_code | postcode | centroid!wkt |
            | de           | 01982    | country:de |
 
-    @skip
-    Scenario: search and address ranks for GB post codes correctly assigned
-        Given the places
-         | osm  | class | type     | postcode | geometry |
-         | N1   | place | postcode | E45 2CD  | country:gb |
-         | N2   | place | postcode | E45 2    | country:gb |
-         | N3   | place | postcode | Y45      | country:gb |
-        When importing
-        Then location_postcode contains exactly
-         | postcode | country_code | rank_search | rank_address |
-         | E45 2CD  | gb           | 25          | 5 |
-         | E45 2    | gb           | 23          | 5 |
-         | Y45      | gb           | 21          | 5 |
-
     Scenario: Postcodes outside all countries are not added to the postcode table
         Given the places
             | osm | class | type  | addr+postcode | addr+housenumber | addr+place  | geometry  |
@@ -200,7 +185,7 @@ Feature: Import of postcodes
             | osm | class | type   | name        | geometry |
             | N1  | place | hamlet | Null Island | 0 0      |
         When importing
-        Then location_postcode contains exactly
+        Then location_postcodes contains exactly
             | place_id |
         When geocoding "111, 01982 Null Island"
         Then the result set contains

test/bdd/features/db/query/normalization.feature

@@ -154,19 +154,6 @@ Feature: Import and search of names
          | object |
          | R2 |
 
-    Scenario: Postcode boundaries without ref
-        Given the grid with origin FR
-          |   | 2 |   |
-          | 1 |   | 3 |
-        Given the places
-          | osm | class    | type        | postcode  | geometry |
-          | R1  | boundary | postal_code | 123-45    | (1,2,3,1) |
-        When importing
-        When geocoding "123-45"
-        Then result 0 contains
-         | object |
-         | R1 |
-
     Scenario Outline: Housenumbers with special characters are found
         Given the grid
             | 1 |  |   |  | 2 |

test/bdd/features/db/query/postcodes.feature

@@ -78,8 +78,8 @@ Feature: Querying fo postcode variants
            | N34 | place | house | EH4 7EA       | 111              | country:gb |
            | N35 | place | house | E4 7EA        | 111              | country:gb |
         When importing
-        Then location_postcode contains exactly
-           | country_code | postcode | geometry!wkt |
+        Then location_postcodes contains exactly
+           | country_code | postcode | centroid!wkt |
            | gb           | EH4 7EA  | country:gb |
            | gb           | E4 7EA   | country:gb |
         When geocoding "EH4 7EA"
@@ -90,20 +90,3 @@ Feature: Querying fo postcode variants
         Then result 0 contains
            | type     | display_name |
            | postcode | E4 7EA, United Kingdom |
-
-
-    Scenario: Postcode areas are preferred over postcode points
-        Given the grid with origin DE
-            | 1 | 2 |
-            | 4 | 3 |
-        Given the places
-            | osm | class    | type        | postcode | geometry    |
-            | R23 | boundary | postal_code | 12345    | (1,2,3,4,1) |
-        When importing
-        Then location_postcode contains exactly
-          | country_code | postcode |
-          | de           | 12345    |
-        When geocoding "12345, de"
-        Then result 0 contains
-          | object |
-          | R23 |

test/bdd/features/db/query/reverse.feature

@@ -9,13 +9,32 @@ Feature: Reverse searches
         And the places
           | osm | class   | type       | geometry    |
           | W1  | aeroway | terminal   | (1,2,3,4,1) |
-          | N1  | amenity | restaurant | 9           |
+          | N9  | amenity | restaurant | 9           |
         When importing
         And reverse geocoding 1.0001,1.0001
         Then the result contains
          | object |
-         | N1  |
+         | N9  |
         When reverse geocoding 1.0003,1.0001
         Then the result contains
          | object |
          | W1  |
+
+
+    Scenario: Find closest housenumber for street matches
+        Given the 0.0001 grid with origin 1,1
+          |    | 1 |   |    |
+          |    |   | 2 |    |
+          | 10 |   |   | 11 |
+        And the places
+          | osm | class   | type     | name        | geometry |
+          | W1  | highway | service  | Goose Drive | 10,11    |
+          | N2  | tourism | art_work | Beauty      | 2        |
+        And the places
+          | osm | class | type  | housenr | geometry |
+          | N1  | place | house | 23      | 1        |
+        When importing
+        When reverse geocoding 1.0002,1.0002
+        Then the result contains
+          | object |
+          | N1 |

test/bdd/features/db/update/naming.feature

@@ -1,20 +0,0 @@
-Feature: Update of names in place objects
-    Test all naming related issues in updates
-
-    Scenario: Delete postcode from postcode boundaries without ref
-        Given the grid with origin DE
-          | 1 | 2 |
-          | 4 | 3 |
-        Given the places
-          | osm | class    | type        | postcode | geometry |
-          | R1  | boundary | postal_code | 123-45    | (1,2,3,4,1) |
-        When importing
-        And geocoding "123-45"
-        Then result 0 contains
-         | object |
-         | R1 |
-        When updating places
-          | osm | class    | type        | geometry |
-          | R1  | boundary | postal_code | (1,2,3,4,1) |
-        Then placex has no entry for R1
-

test/bdd/features/db/update/postcode.feature

@@ -2,20 +2,22 @@ Feature: Update of postcode
     Tests for updating of data related to postcodes
 
      Scenario: Updating postcode in postcode boundaries without ref
-        Given the grid
-          | 1 | 2 |
-          | 4 | 3 |
-        Given the places
-          | osm | class    | type        | postcode | geometry |
-          | R1  | boundary | postal_code | 12345    | (1,2,3,4,1) |
+        Given the grid with origin FR
+          | 1 |   | 2 |
+          |   | 9 |   |
+          | 4 |   | 3 |
+        Given the postcodes
+          | osm | postcode | centroid | geometry |
+          | R1  | 12345    | 9        | (1,2,3,4,1) |
         When importing
         And geocoding "12345"
         Then result 0 contains
          | object |
          | R1 |
-        When updating places
-          | osm | class    | type        | postcode | geometry |
-          | R1  | boundary | postal_code | 54321    | (1,2,3,4,1) |
+        Given the postcodes
+          | osm | postcode | centroid | geometry |
+          | R1  | 54321    | 9        | (1,2,3,4,1) |
+        When refreshing postcodes
         And geocoding "12345"
         Then exactly 0 results are returned
         When geocoding "54321"
@@ -28,17 +30,21 @@ Feature: Update of postcode
            | osm | class | type  | addr+postcode | addr+housenumber | geometry |
            | N34 | place | house | 01982         | 111              | country:de |
         When importing
-        Then location_postcode contains exactly
-           | country_code | postcode | geometry!wkt |
+        Then location_postcodes contains exactly
+           | country_code | postcode | centroid!wkt |
            | de           | 01982    | country:de |
+        Given the postcodes
+           | osm | postcode | centroid   |
+           | N66 | 99201    | country:fr |
         When updating places
            | osm | class | type  | addr+postcode | addr+housenumber | geometry |
            | N35 | place | house | 4567          | 5                | country:ch |
-        And updating postcodes
-        Then location_postcode contains exactly
-           | country_code | postcode | geometry!wkt |
+        And refreshing postcodes
+        Then location_postcodes contains exactly
+           | country_code | postcode | centroid!wkt |
            | de           | 01982    | country:de |
            | ch           | 4567     | country:ch |
+           | fr           | 99201    | country:fr |
 
      Scenario: When the last postcode is deleted, it is deleted from postcode
         Given the places
@@ -47,9 +53,9 @@ Feature: Update of postcode
            | N35 | place | house | 4567          | 5                | country:ch |
         When importing
         And marking for delete N34
-        And updating postcodes
-        Then location_postcode contains exactly
-           | country_code | postcode | geometry!wkt |
+        And refreshing postcodes
+        Then location_postcodes contains exactly
+           | country_code | postcode | centroid!wkt |
            | ch           | 4567     | country:ch |
 
      Scenario: A postcode is not deleted from postcode when it exist in another country
@@ -59,64 +65,24 @@ Feature: Update of postcode
            | N35 | place | house | 01982         | 5                | country:fr |
         When importing
         And marking for delete N34
-        And updating postcodes
-        Then location_postcode contains exactly
-           | country_code | postcode | geometry!wkt|
+        And refreshing postcodes
+        Then location_postcodes contains exactly
+           | country_code | postcode | centroid!wkt|
            | fr           | 01982    | country:fr |
 
      Scenario: Updating a postcode is reflected in postcode table
         Given the places
-           | osm | class | type     | addr+postcode |  geometry |
+           | osm | class | type     | addr+postcode | geometry |
            | N34 | place | postcode | 01982         | country:de |
         When importing
         And updating places
            | osm | class | type     | addr+postcode |  geometry |
            | N34 | place | postcode | 20453         | country:de |
-        And updating postcodes
-        Then location_postcode contains exactly
-           | country_code | postcode | geometry!wkt |
+        And refreshing postcodes
+        Then location_postcodes contains exactly
+           | country_code | postcode | centroid!wkt |
            | de           | 20453    | country:de |
 
-     Scenario: When changing from a postcode type, the entry appears in placex
-        When importing
-        And updating places
-           | osm | class | type     | addr+postcode |  geometry |
-           | N34 | place | postcode | 01982         | country:de |
-        Then placex has no entry for N34
-        When updating places
-           | osm | class | type  | addr+postcode | housenr |  geometry |
-           | N34 | place | house | 20453         | 1       | country:de |
-        Then placex contains
-           | object | addr+housenumber | geometry!wkt |
-           | N34    | 1                | country:de |
-        And place contains exactly
-           | osm_type | osm_id | class | type  |
-           | N        | 34     | place | house |
-        When updating postcodes
-        Then location_postcode contains exactly
-           | country_code | postcode | geometry!wkt |
-           | de           | 20453    | country:de |
-
-     Scenario: When changing to a postcode type, the entry disappears from placex
-        When importing
-        And updating places
-           | osm | class | type  | addr+postcode | housenr |  geometry |
-           | N34 | place | house | 20453         | 1       | country:de |
-        Then placex contains
-           | object | addr+housenumber | geometry!wkt |
-           | N34    | 1                | country:de|
-        When updating places
-           | osm | class | type     | addr+postcode |  geometry |
-           | N34 | place | postcode | 01982         | country:de |
-        Then placex has no entry for N34
-        And place contains exactly
-           | osm_type | osm_id | class | type     |
-           | N        | 34     | place | postcode |
-        When updating postcodes
-        Then location_postcode contains exactly
-           | country_code | postcode | geometry!wkt |
-           | de           | 01982    | country:de |
-
     Scenario: When a parent is deleted, the postcode gets a new parent
         Given the grid with origin DE
            | 1 |   | 3 | 4 |
@@ -126,14 +92,59 @@ Feature: Update of postcode
            | osm | class    | type           | name  | admin | geometry    |
            | R1  | boundary | administrative | Big   | 6     | (1,4,6,2,1) |
            | R2  | boundary | administrative | Small | 6     | (1,3,5,2,1) |
-        Given the places
-           | osm | class | type     | addr+postcode | geometry |
-           | N9  | place | postcode | 12345         | 9        |
+        Given the postcodes
+           | osm | postcode | centroid |
+           | N9  | 12345    | 9        |
         When importing
-        Then location_postcode contains exactly
-           | postcode | geometry!wkt | parent_place_id |
+        Then location_postcodes contains exactly
+           | postcode | centroid!wkt | parent_place_id |
            | 12345    | 9            | R2              |
         When marking for delete R2
-        Then location_postcode contains exactly
-           | country_code | postcode | geometry!wkt | parent_place_id |
+        Then location_postcodes contains exactly
+           | country_code | postcode | centroid!wkt | parent_place_id |
            | de           | 12345    | 9            | R1              |
+
+    Scenario: When a postcode area appears, postcode points are shadowed
+        Given the grid with origin DE
+           | 1 |   | 3 |   |
+           |   | 9 |   | 8 |
+           | 2 |   | 5 |   |
+        Given the postcodes
+           | osm | postcode | centroid |
+           | N92 | 44321    | 9        |
+           | N4  | 00245    | 8        |
+        When importing
+        Then location_postcodes contains exactly
+           | country_code | postcode | osm_id | centroid!wkt |
+           | de           | 44321    | -      | 9            |
+           | de           | 00245    | -      | 8            |
+        Given the postcodes
+           | osm | postcode | centroid | geometry    |
+           | R45 | 00245    | 9        | (1,3,5,2,1) |
+        When refreshing postcodes
+        Then location_postcodes contains exactly
+           | country_code | postcode | osm_id | centroid!wkt |
+           | de           | 00245    | 45     | 9            |
+
+    Scenario: When a postcode area disappears, postcode points are unshadowed
+        Given the grid with origin DE
+           | 1 |   | 3 |   |
+           |   | 9 |   | 8 |
+           | 2 |   | 5 |   |
+        Given the postcodes
+           | osm | postcode | centroid | geometry    |
+           | R45 | 00245    | 9        | (1,3,5,2,1) |
+        Given the postcodes
+           | osm | postcode | centroid |
+           | N92 | 44321    | 9        |
+           | N4  | 00245    | 8        |
+        When importing
+        Then location_postcodes contains exactly
+           | country_code | postcode | osm_id | centroid!wkt |
+           | de           | 00245    | 45     | 9            |
+        When marking for delete R45
+        And refreshing postcodes
+        Then location_postcodes contains exactly
+           | country_code | postcode | osm_id | centroid!wkt |
+           | de           | 44321    | -      | 9            |
+           | de           | 00245    | -      | 8            |

test/bdd/features/osm2pgsql/import/tags.feature

@@ -92,12 +92,16 @@ Feature: Tag evaluation
             n6001 Tshop=bank,addr:postcode=12345
             n6002 Tshop=bank,tiger:zip_left=34343
             n6003 Tshop=bank,is_in:postcode=9009
+            n6004 Taddr:postcode=54322
             """
         Then place contains exactly
             | object | class | address!dict        |
             | N6001  | shop  | 'postcode': '12345' |
             | N6002  | shop  | 'postcode': '34343' |
             | N6003  | shop  | -                   |
+        And place_postcode contains exactly
+            | object | postcode | geometry |
+            | N6004  | 54322    | -        |
 
 
     Scenario: Postcode areas
@@ -107,11 +111,15 @@ Feature: Tag evaluation
             n2 x12.36853 y51.42362
             n3 x12.63666 y51.42362
             n4 x12.63666 y51.50618
-            w1 Tboundary=postal_code,ref=3456 Nn1,n2,n3,n4,n1
+            w1 Nn1,n2,n3,n4,n1
+            w2 Tboundary=postal_code,postal_code=443 Nn1,n2,n3,n4,n1
+            r1 Ttype=boundary,boundary=postal_code,postal_code=3456 Mw1@
             """
         Then place contains exactly
-            | object | class    | type        | name!dict     |
-            | W1     | boundary | postal_code | 'ref': '3456' |
+            | object |
+        And place_postcode contains exactly
+            | object | postcode | geometry!wkt |
+            | R1     | 3456     | (12.36853 51.50618, 12.36853 51.42362, 12.63666 51.42362, 12.63666 51.50618, 12.36853 51.50618) |
 
     Scenario: Main with extra
         When loading osm data
@@ -192,7 +200,9 @@ Feature: Tag evaluation
             | N12001 | tourism  | hotel    |
             | N12003 | building | shed     |
             | N12004 | building | yes      |
-            | N12005 | place    | postcode |
+        And place_postcode contains exactly
+            | object | postcode | geometry |
+            | N12005 | 12345    | -        |
 
 
     Scenario: Address interpolations

test/bdd/features/osm2pgsql/update/postcodes.feature

@@ -2,7 +2,6 @@ Feature: Update of postcode only objects
     Tests that changes to objects containing only a postcode are
     propagated correctly.
 
-
     Scenario: Adding a postcode-only node
         When loading osm data
             """
@@ -15,11 +14,10 @@ Feature: Update of postcode only objects
             """
             n34 Tpostcode=4456
             """
-        Then place contains exactly
-            | object | class | type     |
-            | N34    | place | postcode |
-        When indexing
-        Then placex contains exactly
+        Then place_postcode contains exactly
+            | object | postcode |
+            | N34    | 4456     |
+        And place contains exactly
             | object |
 
 
@@ -28,9 +26,11 @@ Feature: Update of postcode only objects
             """
             n34 Tpostcode=4456
             """
-        Then place contains exactly
-            | object | class | type     |
-            | N34    | place | postcode |
+        Then place_postcode contains exactly
+            | object | postcode |
+            | N34    | 4456     |
+        And place contains exactly
+            | object |
 
         When updating osm data
             """
@@ -38,8 +38,7 @@ Feature: Update of postcode only objects
             """
         Then place contains exactly
             | object |
-        When indexing
-        Then placex contains exactly
+        And place_postcode contains exactly
             | object |
 
 
@@ -57,8 +56,10 @@ Feature: Update of postcode only objects
             n34 Tpostcode=4456
             """
         Then place contains exactly
-            | object | class | type     |
-            | N34    | place | postcode |
+            | object |
+        And place_postcode contains exactly
+            | object | postcode |
+            | N34    | 4456     |
         When indexing
         Then placex contains exactly
             | object |
@@ -74,9 +75,9 @@ Feature: Update of postcode only objects
             """
             n34 Tpostcode=4456
             """
-        Then place contains exactly
-            | object | class | type     |
-            | N34    | place | postcode |
+        Then place_postcode contains exactly
+            | object | postcode |
+            | N34    | 4456     |
 
         When updating osm data
             """
@@ -85,6 +86,8 @@ Feature: Update of postcode only objects
         Then place contains exactly
             | object | class   | type   |
             | N34    | <class> | <type> |
+        And place_postcode contains exactly
+            | object |
         When indexing
         Then placex contains exactly
             | object | class   | type   |
@@ -96,7 +99,7 @@ Feature: Update of postcode only objects
             | place   | hamlet     |
 
 
-    Scenario: Converting na interpolation into a postcode-only node
+    Scenario: Converting an interpolation into a postcode-only node
         Given the grid
             | 1 | 2 |
         When loading osm data
@@ -119,14 +122,12 @@ Feature: Update of postcode only objects
             | object | class | type     |
             | N1     | place | house    |
             | N2     | place | house    |
-            | W34    | place | postcode |
+        Then place_postcode contains exactly
+            | object | postcode |
+            | W34    | 4456     |
         When indexing
         Then location_property_osmline contains exactly
             | osm_id |
-        And placex contains exactly
-            | object | class | type   |
-            | N1     | place | house  |
-            | N2     | place | house  |
 
 
     Scenario: Converting a postcode-only node into an interpolation
@@ -144,7 +145,9 @@ Feature: Update of postcode only objects
             | N1     | place   | house    |
             | N2     | place   | house    |
             | W33    | highway | residential |
-            | W34    | place   | postcode |
+        And place_postcode contains exactly
+            | object | postcode |
+            | W34    | 4456     |
 
         When updating osm data
             """
@@ -156,6 +159,8 @@ Feature: Update of postcode only objects
             | N2     | place   | house  |
             | W33    | highway | residential |
             | W34    | place   | houses |
+        And place_postcode contains exactly
+            | object |
         When indexing
         Then location_property_osmline contains exactly
             | osm_id | startnumber | endnumber |

test/bdd/test_api.py

@@ -15,7 +15,7 @@ import xml.etree.ElementTree as ET
 
 import pytest
 from pytest_bdd.parsers import re as step_parse
-from pytest_bdd import scenarios, when, given, then
+from pytest_bdd import when, given, then
 
 from nominatim_db import cli
 from nominatim_db.config import Configuration
@@ -150,4 +150,8 @@ def parse_api_json_response(api_response, fmt, num):
     return result
 
 
-scenarios('features/api')
+if pytest.version_tuple >= (8, 0, 0):
+    PYTEST_BDD_SCENARIOS = ['features/api']
+else:
+    from pytest_bdd import scenarios
+    scenarios('features/api')

test/bdd/test_db.py

@@ -11,15 +11,17 @@ These tests check the Nominatim import chain after the osm2pgsql import.
 """
 import asyncio
 import re
+from collections import defaultdict
 
 import psycopg
 
 import pytest
-from pytest_bdd import scenarios, when, then, given
+from pytest_bdd import when, then, given
 from pytest_bdd.parsers import re as step_parse
 
 from utils.place_inserter import PlaceColumn
 from utils.checks import check_table_content
+from utils.geometry_alias import ALIASES
 
 from nominatim_db.config import Configuration
 from nominatim_db import cli
@@ -97,6 +99,41 @@ def import_place_entrances(db_conn, datatable, node_grid):
                          data.columns.get('extratags')))
 
 
+@given(step_parse('the postcodes'), target_fixture=None)
+def import_place_postcode(db_conn, datatable, node_grid):
+    """ Insert todo rows into the place_postcode table. If a row for the
+        requested object already exists it is overwritten.
+    """
+    with db_conn.cursor() as cur:
+        for row in datatable[1:]:
+            data = defaultdict(lambda: None)
+            data.update((k, v) for k, v in zip(datatable[0], row))
+
+            if data['centroid'].startswith('country:'):
+                ccode = data['centroid'][8:].upper()
+                data['centroid'] = 'srid=4326;POINT({} {})'.format(*ALIASES[ccode])
+            else:
+                data['centroid'] = f"srid=4326;{node_grid.geometry_to_wkt(data['centroid'])}"
+
+            data['osm_type'] = data['osm'][0]
+            data['osm_id'] = data['osm'][1:]
+
+            if 'geometry' in data:
+                geom = f"'srid=4326;{node_grid.geometry_to_wkt(data['geometry'])}'::geometry"
+            else:
+                geom = 'null'
+
+            cur.execute(""" DELETE FROM place_postcode
+                            WHERE osm_type = %(osm_type)s and osm_id = %(osm_id)s""",
+                        data)
+            cur.execute(f"""INSERT INTO place_postcode
+                            (osm_type, osm_id, country_code, postcode, centroid, geometry)
+                            VALUES (%(osm_type)s, %(osm_id)s,
+                                    %(country)s, %(postcode)s,
+                                    %(centroid)s, {geom})""", data)
+    db_conn.commit()
+
+
 @given('the ways', target_fixture=None)
 def import_ways(db_conn, datatable):
     """ Import raw ways into the osm2pgsql way middle table.
@@ -168,7 +205,7 @@ def do_update(db_conn, update_config, node_grid, datatable):
 
 @when('updating entrances', target_fixture=None)
 def update_place_entrances(db_conn, datatable, node_grid):
-    """ Insert todo rows into the place_entrance table.
+    """ Update rows in the place_entrance table.
     """
     with db_conn.cursor() as cur:
         for row in datatable[1:]:
@@ -181,9 +218,10 @@ def update_place_entrances(db_conn, datatable, node_grid):
                            VALUES (%s, %s, %s, {})""".format(data.get_wkt()),
                         (data.columns['osm_id'], data.columns['type'],
                          data.columns.get('extratags')))
+    db_conn.commit()
 
 
-@when('updating postcodes')
+@when('refreshing postcodes')
 def do_postcode_update(update_config):
     """ Recompute the postcode centroids.
     """
@@ -203,6 +241,8 @@ def do_delete_place(db_conn, update_config, node_grid, otype, oid):
         if otype == 'N':
             cur.execute('DELETE FROM place_entrance WHERE osm_id = %s',
                         (oid, ))
+        cur.execute('DELETE FROM place_postcode WHERE osm_type = %s and osm_id = %s',
+                    (otype, oid))
     db_conn.commit()
 
     cli.nominatim(['index', '-q'], update_config.environ)
@@ -276,4 +316,8 @@ def then_check_interpolation_table_negative(db_conn, oid):
         assert cur.fetchone()[0] == 0
 
 
-scenarios('features/db')
+if pytest.version_tuple >= (8, 0, 0):
+    PYTEST_BDD_SCENARIOS = ['features/db']
+else:
+    from pytest_bdd import scenarios
+    scenarios('features/db')

test/bdd/test_osm2pgsql.py

@@ -11,7 +11,7 @@ import asyncio
 import random
 
 import pytest
-from pytest_bdd import scenarios, when, then, given
+from pytest_bdd import when, then, given
 from pytest_bdd.parsers import re as step_parse
 
 from nominatim_db import cli
@@ -43,7 +43,7 @@ def opl_writer(tmp_path, node_grid):
     def _write(data):
         fname = tmp_path / f"test_osm_{nr[0]}.opl"
         nr[0] += 1
-        with fname.open('wt') as fd:
+        with fname.open('wt', encoding='utf-8') as fd:
             for line in data.split('\n'):
                 if line.startswith('n') and ' x' not in line:
                     coord = node_grid.get(line[1:].split(' ')[0]) \
@@ -59,7 +59,7 @@ def opl_writer(tmp_path, node_grid):
 @given('the lua style file', target_fixture='osm2pgsql_options')
 def set_lua_style_file(osm2pgsql_options, docstring, tmp_path):
     style = tmp_path / 'custom.lua'
-    style.write_text(docstring)
+    style.write_text(docstring, encoding='utf-8')
     osm2pgsql_options['osm2pgsql_style'] = str(style)
 
     return osm2pgsql_options
@@ -106,4 +106,8 @@ def check_place_content(db_conn, datatable, node_grid, table, exact):
     check_table_content(db_conn, table, datatable, grid=node_grid, exact=bool(exact))
 
 
-scenarios('features/osm2pgsql')
+if pytest.version_tuple >= (8, 0, 0):
+    PYTEST_BDD_SCENARIOS = ['features/osm2pgsql']
+else:
+    from pytest_bdd import scenarios
+    scenarios('features/osm2pgsql')

test/bdd/utils/api_runner.py

@@ -2,7 +2,7 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2025 by the Nominatim developer community.
+# Copyright (C) 2026 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Various helper classes for running Nominatim commands.
@@ -54,15 +54,14 @@ class APIRunner:
     def create_engine_starlette(self, environ):
         import nominatim_api.server.starlette.server
         from asgi_lifespan import LifespanManager
-        import httpx
+        from starlette.testclient import TestClient
 
         async def _request(endpoint, params, http_headers):
             app = nominatim_api.server.starlette.server.get_application(None, environ)
 
             async with LifespanManager(app):
-                async with httpx.AsyncClient(app=app, base_url="http://nominatim.test") as client:
-                    response = await client.get("/" + endpoint, params=params,
-                                                headers=http_headers)
+                client = TestClient(app, base_url="http://nominatim.test")
+                response = client.get("/" + endpoint, params=params, headers=http_headers)
 
             return APIResponse(endpoint, response.status_code,
                                response.text, response.headers)

test/bdd/utils/checks.py

@@ -7,6 +7,7 @@
 """
 Helper functions to compare expected values.
 """
+import ast
 import collections.abc
 import json
 import re
@@ -58,7 +59,8 @@ COMPARISON_FUNCS = {
     None: lambda val, exp: str(val) == exp,
     'i': lambda val, exp: str(val).lower() == exp.lower(),
     'fm': lambda val, exp: re.fullmatch(exp, val) is not None,
-    'dict': lambda val, exp: val is None if exp == '-' else (val == eval('{' + exp + '}')),
+    'dict': lambda val, exp: (val is None if exp == '-'
+                              else (val == ast.literal_eval('{' + exp + '}'))),
     'in_box': within_box
 }
 

test/bdd/utils/grid.py

@@ -8,6 +8,7 @@
 A grid describing node placement in an area.
 Useful for visually describing geometries.
 """
+import re
 
 
 class Grid:
@@ -44,3 +45,28 @@ class Grid:
 
     def parse_line(self, value):
         return [self.parse_point(p) for p in value.split(',')]
+
+    def geometry_to_wkt(self, value):
+        """ Parses the given value into a geometry and returns the WKT.
+
+            The value can either be a WKT already or a geometry shortcut
+            with coordinates or grid points.
+        """
+        if re.fullmatch(r'([A-Z]+)\((.*)\)', value) is not None:
+            return value  # already a WKT
+
+        # points
+        if ',' not in value:
+            x, y = self.parse_point(value)
+            return f"POINT({x} {y})"
+
+        # linestring
+        if '(' not in value:
+            coords = ','.join(' '.join(f"{p:.7f}" for p in pt)
+                              for pt in self.parse_line(value))
+            return f"LINESTRING({coords})"
+
+        # simple polygons
+        coords = ','.join(' '.join(f"{p:.7f}" for p in pt)
+                          for pt in self.parse_line(value[1:-1]))
+        return f"POLYGON(({coords}))"

test/bdd/utils/place_inserter.py

@@ -2,15 +2,17 @@
 #
 # This file is part of Nominatim. (https://nominatim.org)
 #
-# Copyright (C) 2025 by the Nominatim developer community.
+# Copyright (C) 2026 by the Nominatim developer community.
 # For a full list of authors see the git log.
 """
 Helper classes for filling the place table.
 """
+import ast
 import random
 import string
 
 from .geometry_alias import ALIASES
+from .grid import Grid
 
 
 class PlaceColumn:
@@ -19,7 +21,7 @@ class PlaceColumn:
     """
     def __init__(self, grid=None):
         self.columns = {'admin_level': 15}
-        self.grid = grid
+        self.grid = grid or Grid()
         self.geometry = None
 
     def add_row(self, headings, row, force_name):
@@ -34,7 +36,8 @@ class PlaceColumn:
             self._add_hstore(
                 'name',
                 'name',
-                ''.join(random.choices(string.printable, k=random.randrange(30))),
+                ''.join(random.choices(string.ascii_uppercase)
+                        + random.choices(string.printable, k=random.randrange(30))),
             )
 
         return self
@@ -49,7 +52,7 @@ class PlaceColumn:
         elif key.startswith('addr+'):
             self._add_hstore('address', key[5:], value)
         elif key in ('name', 'address', 'extratags'):
-            self.columns[key] = eval('{' + value + '}')
+            self.columns[key] = ast.literal_eval('{' + value + '}')
         else:
             assert key in ('class', 'type'), "Unknown column '{}'.".format(key)
             self.columns[key] = None if value == '' else value
@@ -91,26 +94,9 @@ class PlaceColumn:
         if value.startswith('country:'):
             ccode = value[8:].upper()
             self.geometry = "ST_SetSRID(ST_Point({}, {}), 4326)".format(*ALIASES[ccode])
-        elif ',' not in value:
-            if self.grid:
-                pt = self.grid.parse_point(value)
-            else:
-                pt = value.split(' ')
-            self.geometry = f"ST_SetSRID(ST_Point({pt[0]}, {pt[1]}), 4326)"
-        elif '(' not in value:
-            if self.grid:
-                coords = ','.join(' '.join(f"{p:.7f}" for p in pt)
-                                  for pt in self.grid.parse_line(value))
-            else:
-                coords = value
-            self.geometry = f"'srid=4326;LINESTRING({coords})'::geometry"
         else:
-            if self.grid:
-                coords = ','.join(' '.join(f"{p:.7f}" for p in pt)
-                                  for pt in self.grid.parse_line(value[1:-1]))
-            else:
-                coords = value[1:-1]
-            self.geometry = f"'srid=4326;POLYGON(({coords}))'::geometry"
+            wkt = self.grid.geometry_to_wkt(value)
+            self.geometry = f"'srid=4326;{wkt}'::geometry"
 
     def _add_hstore(self, column, key, value):
         if column in self.columns:
@@ -120,7 +106,7 @@ class PlaceColumn:
 
     def get_wkt(self):
         if self.columns['osm_type'] == 'N' and self.geometry is None:
-            pt = self.grid.get(str(self.columns['osm_id'])) if self.grid else None
+            pt = self.grid.get(str(self.columns['osm_id']))
             if pt is None:
                 pt = (random.uniform(-180, 180), random.uniform(-90, 90))