]> git.openstreetmap.org Git - nominatim.git/blobdiff - nominatim/tools/refresh.py
Integrated OSM views into importance computation
[nominatim.git] / nominatim / tools / refresh.py
index 859b5646ea188ec58badc636549dcaa09b9a9fd7..531de14db8d4911270cf603a4bc6fc0c27f29337 100644 (file)
+# SPDX-License-Identifier: GPL-2.0-only
+#
+# This file is part of Nominatim. (https://nominatim.org)
+#
+# Copyright (C) 2022 by the Nominatim developer community.
+# For a full list of authors see the git log.
 """
 Functions for bringing auxiliary data in the database up-to-date.
 """
-from ..db.utils import execute_file
+from typing import MutableSequence, Tuple, Any, Type, Mapping, Sequence, List, cast
+import logging
+import subprocess
+from textwrap import dedent
+from pathlib import Path
 
-def update_postcodes(conn, datadir):
-    """ Recalculate postcode centroids and add, remove and update entries in the
-        location_postcode table. `conn` is an opne connection to the database.
+from psycopg2 import sql as pysql
+
+from nominatim.config import Configuration
+from nominatim.db.connection import Connection, connect
+from nominatim.db.utils import execute_file
+from nominatim.db.sql_preprocessor import SQLPreprocessor
+from nominatim.version import version_str
+
+LOG = logging.getLogger()
+
+OSM_TYPE = {'N': 'node', 'W': 'way', 'R': 'relation'}
+
+def _add_address_level_rows_from_entry(rows: MutableSequence[Tuple[Any, ...]],
+                                       entry: Mapping[str, Any]) -> None:
+    """ Converts a single entry from the JSON format for address rank
+        descriptions into a flat format suitable for inserting into a
+        PostgreSQL table and adds these lines to `rows`.
     """
-    execute_file(conn, datadir / 'sql' / 'update-postcodes.sql')
+    countries = entry.get('countries') or (None, )
+    for key, values in entry['tags'].items():
+        for value, ranks in values.items():
+            if isinstance(ranks, list):
+                rank_search, rank_address = ranks
+            else:
+                rank_search = rank_address = ranks
+            if not value:
+                value = None
+            for country in countries:
+                rows.append((country, key, value, rank_search, rank_address))
+
 
+def load_address_levels(conn: Connection, table: str, levels: Sequence[Mapping[str, Any]]) -> None:
+    """ Replace the `address_levels` table with the contents of `levels'.
 
-def recompute_word_counts(conn, datadir):
-    """ Compute the frequency of full-word search terms.
+        A new table is created any previously existing table is dropped.
+        The table has the following columns:
+            country, class, type, rank_search, rank_address
     """
-    execute_file(conn, datadir / 'sql' / 'words_from_search_name.sql')
+    rows: List[Tuple[Any, ...]]  = []
+    for entry in levels:
+        _add_address_level_rows_from_entry(rows, entry)
+
+    with conn.cursor() as cur:
+        cur.drop_table(table)
+
+        cur.execute(pysql.SQL("""CREATE TABLE {} (
+                                        country_code varchar(2),
+                                        class TEXT,
+                                        type TEXT,
+                                        rank_search SMALLINT,
+                                        rank_address SMALLINT)
+                              """).format(pysql.Identifier(table)))
+
+        cur.execute_values(pysql.SQL("INSERT INTO {} VALUES %s")
+                           .format(pysql.Identifier(table)), rows)
+
+        cur.execute(pysql.SQL('CREATE UNIQUE INDEX ON {} (country_code, class, type)')
+                    .format(pysql.Identifier(table)))
+
+    conn.commit()
+
+
+def load_address_levels_from_config(conn: Connection, config: Configuration) -> None:
+    """ Replace the `address_levels` table with the content as
+        defined in the given configuration. Uses the parameter
+        NOMINATIM_ADDRESS_LEVEL_CONFIG to determine the location of the
+        configuration file.
+    """
+    cfg = config.load_sub_configuration('', config='ADDRESS_LEVEL_CONFIG')
+    load_address_levels(conn, 'address_levels', cfg)
+
+
+def create_functions(conn: Connection, config: Configuration,
+                     enable_diff_updates: bool = True,
+                     enable_debug: bool = False) -> None:
+    """ (Re)create the PL/pgSQL functions.
+    """
+    sql = SQLPreprocessor(conn, config)
+
+    sql.run_sql_file(conn, 'functions.sql',
+                     disable_diff_updates=not enable_diff_updates,
+                     debug=enable_debug)
+
+
+
+WEBSITE_SCRIPTS = (
+    'deletable.php',
+    'details.php',
+    'lookup.php',
+    'polygons.php',
+    'reverse.php',
+    'search.php',
+    'status.php'
+)
+
+# constants needed by PHP scripts: PHP name, config name, type
+PHP_CONST_DEFS = (
+    ('Database_DSN', 'DATABASE_DSN', str),
+    ('Default_Language', 'DEFAULT_LANGUAGE', str),
+    ('Log_DB', 'LOG_DB', bool),
+    ('Log_File', 'LOG_FILE', Path),
+    ('NoAccessControl', 'CORS_NOACCESSCONTROL', bool),
+    ('Places_Max_ID_count', 'LOOKUP_MAX_COUNT', int),
+    ('PolygonOutput_MaximumTypes', 'POLYGON_OUTPUT_MAX_TYPES', int),
+    ('Search_BatchMode', 'SEARCH_BATCH_MODE', bool),
+    ('Search_NameOnlySearchFrequencyThreshold', 'SEARCH_NAME_ONLY_THRESHOLD', str),
+    ('Use_US_Tiger_Data', 'USE_US_TIGER_DATA', bool),
+    ('MapIcon_URL', 'MAPICON_URL', str),
+)
+
+
+def import_wikipedia_articles(dsn: str, data_path: Path, ignore_errors: bool = False) -> int:
+    """ Replaces the wikipedia importance tables with new data.
+        The import is run in a single transaction so that the new data
+        is replace seamlessly.
+
+        Returns 0 if all was well and 1 if the importance file could not
+        be found. Throws an exception if there was an error reading the file.
+    """
+    datafile = data_path / 'wikimedia-importance.sql.gz'
+
+    if not datafile.exists():
+        return 1
+
+    pre_code = """BEGIN;
+                  DROP TABLE IF EXISTS "wikipedia_article";
+                  DROP TABLE IF EXISTS "wikipedia_redirect"
+               """
+    post_code = "COMMIT"
+    execute_file(dsn, datafile, ignore_errors=ignore_errors,
+                 pre_code=pre_code, post_code=post_code)
+
+    return 0
+
+def import_osm_views_geotiff(dsn: str, data_path: Path) -> int:
+    """ Replaces the OSM views table with new data.
+
+        Returns 0 if all was well and 1 if the OSM views GeoTIFF file could not
+        be found. Throws an exception if there was an error reading the file.
+    """
+    datafile = data_path / 'osmviews.tiff'
+    if not datafile.exists():
+        return 1
+    with connect(dsn) as conn:
+
+        postgis_version = conn.postgis_version_tuple()
+        if postgis_version[0] < 3:
+            return 2
+
+        with conn.cursor() as cur:
+            cur.drop_table("osm_views")
+            cur.drop_table("osm_views_stat")
+
+            # -ovr: 6 -> zoom 12, 5 -> zoom 13, 4 -> zoom 14, 3 -> zoom 15
+            reproject_geotiff = f"gdalwarp -q -multi -ovr 3 -overwrite \
+                -co COMPRESS=LZW -tr 0.01 0.01 -t_srs EPSG:4326 {datafile} raster2import.tiff"
+            subprocess.run(["/bin/bash", "-c" , reproject_geotiff], check=True)
+
+            tile_size = 256
+            import_geotiff = f"raster2pgsql -I -C -Y -t {tile_size}x{tile_size} raster2import.tiff \
+                public.osm_views | psql {dsn} > /dev/null"
+            subprocess.run(["/bin/bash", "-c" , import_geotiff], check=True)
+
+            cleanup = "rm raster2import.tiff"
+            subprocess.run(["/bin/bash", "-c" , cleanup], check=True)
+
+            # To normalize osm views data, the max view value is needed
+            cur.execute(f"""
+            CREATE TABLE osm_views_stat AS (
+                SELECT MAX(ST_Value(osm_views.rast, 1, x, y)) AS max_views_count
+                FROM osm_views CROSS JOIN
+                generate_series(1, {tile_size}) As x
+                CROSS JOIN generate_series(1, {tile_size}) As y
+                WHERE x <= ST_Width(rast) AND y <= ST_Height(rast));
+            """)
+            conn.commit()
+
+    return 0
+
+def recompute_importance(conn: Connection) -> None:
+    """ Recompute wikipedia links and importance for all entries in placex.
+        This is a long-running operations that must not be executed in
+        parallel with updates.
+    """
+    with conn.cursor() as cur:
+        cur.execute('ALTER TABLE placex DISABLE TRIGGER ALL')
+        cur.execute("""
+            UPDATE placex SET (wikipedia, importance) =
+               (SELECT wikipedia, importance
+                FROM compute_importance(extratags, country_code, osm_type, osm_id, centroid))
+            """)
+        cur.execute("""
+            UPDATE placex s SET wikipedia = d.wikipedia, importance = d.importance
+             FROM placex d
+             WHERE s.place_id = d.linked_place_id and d.wikipedia is not null
+                   and (s.wikipedia is null or s.importance < d.importance);
+            """)
+
+        cur.execute('ALTER TABLE placex ENABLE TRIGGER ALL')
+    conn.commit()
+
+
+def _quote_php_variable(var_type: Type[Any], config: Configuration,
+                        conf_name: str) -> str:
+    if var_type == bool:
+        return 'true' if config.get_bool(conf_name) else 'false'
+
+    if var_type == int:
+        return cast(str, getattr(config, conf_name))
+
+    if not getattr(config, conf_name):
+        return 'false'
+
+    if var_type == Path:
+        value = str(config.get_path(conf_name) or '')
+    else:
+        value = getattr(config, conf_name)
+
+    quoted = value.replace("'", "\\'")
+    return f"'{quoted}'"
+
+
+def setup_website(basedir: Path, config: Configuration, conn: Connection) -> None:
+    """ Create the website script stubs.
+    """
+    if not basedir.exists():
+        LOG.info('Creating website directory.')
+        basedir.mkdir()
+
+    template = dedent(f"""\
+                      <?php
+
+                      @define('CONST_Debug', $_GET['debug'] ?? false);
+                      @define('CONST_LibDir', '{config.lib_dir.php}');
+                      @define('CONST_TokenizerDir', '{config.project_dir / 'tokenizer'}');
+                      @define('CONST_NominatimVersion', '{version_str()}');
+
+                      """)
+
+    for php_name, conf_name, var_type in PHP_CONST_DEFS:
+        varout = _quote_php_variable(var_type, config, conf_name)
+
+        template += f"@define('CONST_{php_name}', {varout});\n"
+
+    template += f"\nrequire_once('{config.lib_dir.php}/website/{{}}');\n"
+
+    search_name_table_exists = bool(conn and conn.table_exists('search_name'))
+
+    for script in WEBSITE_SCRIPTS:
+        if not search_name_table_exists and script == 'search.php':
+            (basedir / script).write_text(template.format('reverse-only-search.php'), 'utf-8')
+        else:
+            (basedir / script).write_text(template.format(script), 'utf-8')
+
+
+def invalidate_osm_object(osm_type: str, osm_id: int, conn: Connection,
+                          recursive: bool = True) -> None:
+    """ Mark the given OSM object for reindexing. When 'recursive' is set
+        to True (the default), then all dependent objects are marked for
+        reindexing as well.
+
+        'osm_type' must be on of 'N' (node), 'W' (way) or 'R' (relation).
+        If the given object does not exist, then nothing happens.
+    """
+    assert osm_type in ('N', 'R', 'W')
+
+    LOG.warning("Invalidating OSM %s %s%s.",
+                OSM_TYPE[osm_type], osm_id,
+                ' and its dependent places' if recursive else '')
+
+    with conn.cursor() as cur:
+        if recursive:
+            sql = """SELECT place_force_update(place_id)
+                     FROM placex WHERE osm_type = %s and osm_id = %s"""
+        else:
+            sql = """UPDATE placex SET indexed_status = 2
+                     WHERE osm_type = %s and osm_id = %s"""
+
+        cur.execute(sql, (osm_type, osm_id))