Merge remote-tracking branch 'upstream/master'

[nominatim.git] / nominatim / tools / database_import.py

diff --git a/nominatim/tools/database_import.py b/nominatim/tools/database_import.py
index 069157000568fea2ecc59452f86c605ab9459041..0dd93490d7b4c8c316bdd5635a02a3b1c9a6c6da 100644 (file)
--- a/nominatim/tools/database_import.py
+++ b/nominatim/tools/database_import.py
@@ -5,19 +5,19 @@ import logging
 import os
 import selectors
 import subprocess
 import os
 import selectors
 import subprocess

-import shutil

 from pathlib import Path
 
 import psutil
 from pathlib import Path
 
 import psutil

-import psycopg2
+import psycopg2.extras
+from psycopg2 import sql as pysql

 
 

-from ..db.connection import connect, get_pg_env
-from ..db import utils as db_utils
-from ..db.async_connection import DBConnection
-from ..db.sql_preprocessor import SQLPreprocessor
-from .exec_utils import run_osm2pgsql
-from ..errors import UsageError
-from ..version import POSTGRESQL_REQUIRED_VERSION, POSTGIS_REQUIRED_VERSION
+from nominatim.db.connection import connect, get_pg_env
+from nominatim.db import utils as db_utils
+from nominatim.db.async_connection import DBConnection
+from nominatim.db.sql_preprocessor import SQLPreprocessor
+from nominatim.tools.exec_utils import run_osm2pgsql
+from nominatim.errors import UsageError
+from nominatim.version import POSTGRESQL_REQUIRED_VERSION, POSTGIS_REQUIRED_VERSION

 
 LOG = logging.getLogger()
 
 
 LOG = logging.getLogger()
 


@@ -89,49 +89,6 @@ def setup_extensions(conn):
         raise UsageError('PostGIS version is too old.')
 
 
         raise UsageError('PostGIS version is too old.')
 
 

-def install_module(src_dir, project_dir, module_dir, conn=None):
-    """ Copy the normalization module from src_dir into the project
-        directory under the '/module' directory. If 'module_dir' is set, then
-        use the module from there instead and check that it is accessible
-        for Postgresql.
-
-        The function detects when the installation is run from the
-        build directory. It doesn't touch the module in that case.
-
-        If 'conn' is given, then the function also tests if the module
-        can be access via the given database.
-    """
-    if not module_dir:
-        module_dir = project_dir / 'module'
-
-        if not module_dir.exists() or not src_dir.samefile(module_dir):
-
-            if not module_dir.exists():
-                module_dir.mkdir()
-
-            destfile = module_dir / 'nominatim.so'
-            shutil.copy(str(src_dir / 'nominatim.so'), str(destfile))
-            destfile.chmod(0o755)
-
-            LOG.info('Database module installed at %s', str(destfile))
-        else:
-            LOG.info('Running from build directory. Leaving database module as is.')
-    else:
-        LOG.info("Using custom path for database module at '%s'", module_dir)
-
-    if conn is not None:
-        with conn.cursor() as cur:
-            try:
-                cur.execute("""CREATE FUNCTION nominatim_test_import_func(text)
-                               RETURNS text AS '{}/nominatim.so', 'transliteration'
-                               LANGUAGE c IMMUTABLE STRICT;
-                               DROP FUNCTION nominatim_test_import_func(text)
-                            """.format(module_dir))
-            except psycopg2.DatabaseError as err:
-                LOG.fatal("Error accessing database module: %s", err)
-                raise UsageError("Database module cannot be accessed.") from err
-
-

 def import_base_data(dsn, sql_dir, ignore_partitions=False):
     """ Create and populate the tables with basic static data that provides
         the background for geocoding. Data is assumed to not yet exist.
 def import_base_data(dsn, sql_dir, ignore_partitions=False):
     """ Create and populate the tables with basic static data that provides
         the background for geocoding. Data is assumed to not yet exist.


@@ -146,11 +103,11 @@ def import_base_data(dsn, sql_dir, ignore_partitions=False):
             conn.commit()
 
 
             conn.commit()
 
 

-def import_osm_data(osm_file, options, drop=False, ignore_errors=False):
-    """ Import the given OSM file. 'options' contains the list of
+def import_osm_data(osm_files, options, drop=False, ignore_errors=False):
+    """ Import the given OSM files. 'options' contains the list of

         default settings for osm2pgsql.
     """
         default settings for osm2pgsql.
     """

-    options['import_file'] = osm_file
+    options['import_file'] = osm_files

     options['append'] = False
     options['threads'] = 1
 
     options['append'] = False
     options['threads'] = 1
 


@@ -158,7 +115,12 @@ def import_osm_data(osm_file, options, drop=False, ignore_errors=False):
         # Make some educated guesses about cache size based on the size
         # of the import file and the available memory.
         mem = psutil.virtual_memory()
         # Make some educated guesses about cache size based on the size
         # of the import file and the available memory.
         mem = psutil.virtual_memory()

-        fsize = os.stat(str(osm_file)).st_size
+        fsize = 0
+        if isinstance(osm_files, list):
+            for fname in osm_files:
+                fsize += os.stat(str(fname)).st_size
+        else:
+            fsize = os.stat(str(osm_files)).st_size

         options['osm2pgsql_cache'] = int(min((mem.available + mem.cached) * 0.75,
                                              fsize * 2) / 1024 / 1024) + 1
 
         options['osm2pgsql_cache'] = int(min((mem.available + mem.cached) * 0.75,
                                              fsize * 2) / 1024 / 1024) + 1
 


@@ -174,47 +136,44 @@ def import_osm_data(osm_file, options, drop=False, ignore_errors=False):
         if drop:
             conn.drop_table('planet_osm_nodes')
 
         if drop:
             conn.drop_table('planet_osm_nodes')
 

-    if drop:
-        if options['flatnode_file']:
-            Path(options['flatnode_file']).unlink()
+    if drop and options['flatnode_file']:
+        Path(options['flatnode_file']).unlink()

 
 
 
 

-def create_tables(conn, config, sqllib_dir, reverse_only=False):
+def create_tables(conn, config, reverse_only=False):

     """ Create the set of basic tables.
         When `reverse_only` is True, then the main table for searching will
         be skipped and only reverse search is possible.
     """
     """ Create the set of basic tables.
         When `reverse_only` is True, then the main table for searching will
         be skipped and only reverse search is possible.
     """

-    sql = SQLPreprocessor(conn, config, sqllib_dir)
+    sql = SQLPreprocessor(conn, config)

     sql.env.globals['db']['reverse_only'] = reverse_only
 
     sql.run_sql_file(conn, 'tables.sql')
 
 
     sql.env.globals['db']['reverse_only'] = reverse_only
 
     sql.run_sql_file(conn, 'tables.sql')
 
 

-def create_table_triggers(conn, config, sqllib_dir):
+def create_table_triggers(conn, config):

     """ Create the triggers for the tables. The trigger functions must already
         have been imported with refresh.create_functions().
     """
     """ Create the triggers for the tables. The trigger functions must already
         have been imported with refresh.create_functions().
     """

-    sql = SQLPreprocessor(conn, config, sqllib_dir)
+    sql = SQLPreprocessor(conn, config)

     sql.run_sql_file(conn, 'table-triggers.sql')
 
 
     sql.run_sql_file(conn, 'table-triggers.sql')
 
 

-def create_partition_tables(conn, config, sqllib_dir):
-    """ Create tables that have explicit partioning.
+def create_partition_tables(conn, config):
+    """ Create tables that have explicit partitioning.

     """
     """

-    sql = SQLPreprocessor(conn, config, sqllib_dir)
+    sql = SQLPreprocessor(conn, config)

     sql.run_sql_file(conn, 'partition-tables.src.sql')
 
 
     sql.run_sql_file(conn, 'partition-tables.src.sql')
 
 

-def truncate_data_tables(conn, max_word_frequency=None):
+def truncate_data_tables(conn):

     """ Truncate all data tables to prepare for a fresh load.
     """
     with conn.cursor() as cur:
     """ Truncate all data tables to prepare for a fresh load.
     """
     with conn.cursor() as cur:

-        cur.execute('TRUNCATE word')

         cur.execute('TRUNCATE placex')
         cur.execute('TRUNCATE place_addressline')
         cur.execute('TRUNCATE location_area')
         cur.execute('TRUNCATE location_area_country')
         cur.execute('TRUNCATE placex')
         cur.execute('TRUNCATE place_addressline')
         cur.execute('TRUNCATE location_area')
         cur.execute('TRUNCATE location_area_country')

-        cur.execute('TRUNCATE location_property')

         cur.execute('TRUNCATE location_property_tiger')
         cur.execute('TRUNCATE location_property_osmline')
         cur.execute('TRUNCATE location_postcode')
         cur.execute('TRUNCATE location_property_tiger')
         cur.execute('TRUNCATE location_property_osmline')
         cur.execute('TRUNCATE location_postcode')


@@ -229,35 +188,33 @@ def truncate_data_tables(conn, max_word_frequency=None):
         for table in [r[0] for r in list(cur)]:
             cur.execute('TRUNCATE ' + table)
 
         for table in [r[0] for r in list(cur)]:
             cur.execute('TRUNCATE ' + table)
 

-        if max_word_frequency is not None:
-            # Used by getorcreate_word_id to ignore frequent partial words.
-            cur.execute("""CREATE OR REPLACE FUNCTION get_maxwordfreq()
-                           RETURNS integer AS $$
-                             SELECT {} as maxwordfreq;
-                           $$ LANGUAGE SQL IMMUTABLE
-                        """.format(max_word_frequency))
-        conn.commit()
+    conn.commit()

 
 

-_COPY_COLUMNS = 'osm_type, osm_id, class, type, name, admin_level, address, extratags, geometry'

 
 

-def load_data(dsn, data_dir, threads):
+_COPY_COLUMNS = pysql.SQL(',').join(map(pysql.Identifier,
+                                        ('osm_type', 'osm_id', 'class', 'type',
+                                         'name', 'admin_level', 'address',
+                                         'extratags', 'geometry')))
+
+
+def load_data(dsn, threads):

     """ Copy data into the word and placex table.
     """
     """ Copy data into the word and placex table.
     """

-    # Pre-calculate the most important terms in the word list.
-    db_utils.execute_file(dsn, data_dir / 'words.sql')
-

     sel = selectors.DefaultSelector()
     # Then copy data from place to placex in <threads - 1> chunks.
     place_threads = max(1, threads - 1)
     for imod in range(place_threads):
         conn = DBConnection(dsn)
         conn.connect()
     sel = selectors.DefaultSelector()
     # Then copy data from place to placex in <threads - 1> chunks.
     place_threads = max(1, threads - 1)
     for imod in range(place_threads):
         conn = DBConnection(dsn)
         conn.connect()

-        conn.perform("""INSERT INTO placex ({0})
-                         SELECT {0} FROM place
-                         WHERE osm_id % {1} = {2}
-                           AND NOT (class='place' and type='houses')
-                           AND ST_IsValid(geometry)
-                     """.format(_COPY_COLUMNS, place_threads, imod))
+        conn.perform(
+            pysql.SQL("""INSERT INTO placex ({columns})
+                           SELECT {columns} FROM place
+                           WHERE osm_id % {total} = {mod}
+                             AND NOT (class='place' and (type='houses' or type='postcode'))
+                             AND ST_IsValid(geometry)
+                      """).format(columns=_COPY_COLUMNS,
+                                  total=pysql.Literal(place_threads),
+                                  mod=pysql.Literal(imod)))

         sel.register(conn, selectors.EVENT_READ, conn)
 
     # Address interpolations go into another table.
         sel.register(conn, selectors.EVENT_READ, conn)
 
     # Address interpolations go into another table.


@@ -285,3 +242,59 @@ def load_data(dsn, data_dir, threads):
     with connect(dsn) as conn:
         with conn.cursor() as cur:
             cur.execute('ANALYSE')
     with connect(dsn) as conn:
         with conn.cursor() as cur:
             cur.execute('ANALYSE')

+
+
+def create_search_indices(conn, config, drop=False):
+    """ Create tables that have explicit partitioning.
+    """
+
+    # If index creation failed and left an index invalid, they need to be
+    # cleaned out first, so that the script recreates them.
+    with conn.cursor() as cur:
+        cur.execute("""SELECT relname FROM pg_class, pg_index
+                       WHERE pg_index.indisvalid = false
+                             AND pg_index.indexrelid = pg_class.oid""")
+        bad_indices = [row[0] for row in list(cur)]
+        for idx in bad_indices:
+            LOG.info("Drop invalid index %s.", idx)
+            cur.execute('DROP INDEX "{}"'.format(idx))
+    conn.commit()
+
+    sql = SQLPreprocessor(conn, config)
+
+    sql.run_sql_file(conn, 'indices.sql', drop=drop)
+
+
+def create_country_names(conn, tokenizer, languages=None):
+    """ Add default country names to search index. `languages` is a comma-
+        separated list of language codes as used in OSM. If `languages` is not
+        empty then only name translations for the given languages are added
+        to the index.
+    """
+    if languages:
+        languages = languages.split(',')
+
+    def _include_key(key):
+        return key == 'name' or \
+               (key.startswith('name:') and (not languages or key[5:] in languages))
+
+    with conn.cursor() as cur:
+        psycopg2.extras.register_hstore(cur)
+        cur.execute("""SELECT country_code, name FROM country_name
+                       WHERE country_code is not null""")
+
+        with tokenizer.name_analyzer() as analyzer:
+            for code, name in cur:
+                names = {'countrycode': code}
+                if code == 'gb':
+                    names['short_name'] = 'UK'
+                if code == 'us':
+                    names['short_name'] = 'United States'
+
+                # country names (only in languages as provided)
+                if name:
+                    names.update(((k, v) for k, v in name.items() if _include_key(k)))
+
+                analyzer.add_country_names(code, names)
+
+    conn.commit()