+
+@_migration(3, 6, 0, 0)
+def change_housenumber_transliteration(conn, **_):
+ """ Transliterate housenumbers.
+
+ The database schema switched from saving raw housenumbers in
+ placex.housenumber to saving transliterated ones.
+
+ Note: the function create_housenumber_id() has been dropped in later
+ versions.
+ """
+ with conn.cursor() as cur:
+ cur.execute("""CREATE OR REPLACE FUNCTION create_housenumber_id(housenumber TEXT)
+ RETURNS TEXT AS $$
+ DECLARE
+ normtext TEXT;
+ BEGIN
+ SELECT array_to_string(array_agg(trans), ';')
+ INTO normtext
+ FROM (SELECT lookup_word as trans,
+ getorcreate_housenumber_id(lookup_word)
+ FROM (SELECT make_standard_name(h) as lookup_word
+ FROM regexp_split_to_table(housenumber, '[,;]') h) x) y;
+ return normtext;
+ END;
+ $$ LANGUAGE plpgsql STABLE STRICT;""")
+ cur.execute("DELETE FROM word WHERE class = 'place' and type = 'house'")
+ cur.execute("""UPDATE placex
+ SET housenumber = create_housenumber_id(housenumber)
+ WHERE housenumber is not null""")
+
+
+@_migration(3, 7, 0, 0)
+def switch_placenode_geometry_index(conn, **_):
+ """ Replace idx_placex_geometry_reverse_placeNode index.
+
+ Make the index slightly more permissive, so that it can also be used
+ when matching up boundaries and place nodes. It makes the index
+ idx_placex_adminname index unnecessary.
+ """
+ with conn.cursor() as cur:
+ cur.execute(""" CREATE INDEX IF NOT EXISTS idx_placex_geometry_placenode ON placex
+ USING GIST (geometry)
+ WHERE osm_type = 'N' and rank_search < 26
+ and class = 'place' and type != 'postcode'
+ and linked_place_id is null""")
+ cur.execute(""" DROP INDEX IF EXISTS idx_placex_adminname """)
+
+
+@_migration(3, 7, 0, 1)
+def install_legacy_tokenizer(conn, config, **_):
+ """ Setup legacy tokenizer.
+
+ If no other tokenizer has been configured yet, then create the
+ configuration for the backwards-compatible legacy tokenizer
+ """
+ if properties.get_property(conn, 'tokenizer') is None:
+ with conn.cursor() as cur:
+ for table in ('placex', 'location_property_osmline'):
+ has_column = cur.scalar("""SELECT count(*) FROM information_schema.columns
+ WHERE table_name = %s
+ and column_name = 'token_info'""",
+ (table, ))
+ if has_column == 0:
+ cur.execute('ALTER TABLE {} ADD COLUMN token_info JSONB'.format(table))
+ tokenizer = tokenizer_factory.create_tokenizer(config, init_db=False,
+ module_name='legacy')
+
+ tokenizer.migrate_database(config)
+
+
+@_migration(4, 0, 99, 0)
+def create_tiger_housenumber_index(conn, **_):
+ """ Create idx_location_property_tiger_parent_place_id with included
+ house number.
+
+ The inclusion is needed for efficient lookup of housenumbers in
+ full address searches.
+ """
+ if conn.server_version_tuple() >= (11, 0, 0):
+ with conn.cursor() as cur:
+ cur.execute(""" CREATE INDEX IF NOT EXISTS
+ idx_location_property_tiger_housenumber_migrated
+ ON location_property_tiger
+ USING btree(parent_place_id)
+ INCLUDE (startnumber, endnumber) """)
+
+
+@_migration(4, 0, 99, 1)
+def create_interpolation_index_on_place(conn, **_):
+ """ Create idx_place_interpolations for lookup of interpolation lines
+ on updates.
+ """
+ with conn.cursor() as cur:
+ cur.execute("""CREATE INDEX IF NOT EXISTS idx_place_interpolations
+ ON place USING gist(geometry)
+ WHERE osm_type = 'W' and address ? 'interpolation'""")
+
+
+@_migration(4, 0, 99, 2)
+def add_step_column_for_interpolation(conn, **_):
+ """ Add a new column 'step' to the interpolations table.
+
+ Also convers the data into the stricter format which requires that
+ startnumbers comply with the odd/even requirements.
+ """
+ if conn.table_has_column('location_property_osmline', 'step'):
+ return
+
+ with conn.cursor() as cur:
+ # Mark invalid all interpolations with no intermediate numbers.
+ cur.execute("""UPDATE location_property_osmline SET startnumber = null
+ WHERE endnumber - startnumber <= 1 """)
+ # Align the start numbers where odd/even does not match.
+ cur.execute("""UPDATE location_property_osmline
+ SET startnumber = startnumber + 1,
+ linegeo = ST_LineSubString(linegeo,
+ 1.0 / (endnumber - startnumber)::float,
+ 1)
+ WHERE (interpolationtype = 'odd' and startnumber % 2 = 0)
+ or (interpolationtype = 'even' and startnumber % 2 = 1)
+ """)
+ # Mark invalid odd/even interpolations with no intermediate numbers.
+ cur.execute("""UPDATE location_property_osmline SET startnumber = null
+ WHERE interpolationtype in ('odd', 'even')
+ and endnumber - startnumber = 2""")
+ # Finally add the new column and populate it.
+ cur.execute("ALTER TABLE location_property_osmline ADD COLUMN step SMALLINT")
+ cur.execute("""UPDATE location_property_osmline
+ SET step = CASE WHEN interpolationtype = 'all'
+ THEN 1 ELSE 2 END
+ """)
+
+
+@_migration(4, 0, 99, 3)
+def add_step_column_for_tiger(conn, **_):
+ """ Add a new column 'step' to the tiger data table.
+ """
+ if conn.table_has_column('location_property_tiger', 'step'):
+ return
+
+ with conn.cursor() as cur:
+ cur.execute("ALTER TABLE location_property_tiger ADD COLUMN step SMALLINT")
+ cur.execute("""UPDATE location_property_tiger
+ SET step = CASE WHEN interpolationtype = 'all'
+ THEN 1 ELSE 2 END
+ """)
+
+
+@_migration(4, 0, 99, 4)
+def add_derived_name_column_for_country_names(conn, **_):
+ """ Add a new column 'derived_name' which in the future takes the
+ country names as imported from OSM data.
+ """
+ if not conn.table_has_column('country_name', 'derived_name'):
+ with conn.cursor() as cur:
+ cur.execute("ALTER TABLE country_name ADD COLUMN derived_name public.HSTORE")
+
+
+@_migration(4, 0, 99, 5)
+def mark_internal_country_names(conn, config, **_):
+ """ Names from the country table should be marked as internal to prevent
+ them from being deleted. Only necessary for ICU tokenizer.
+ """
+ import psycopg2.extras # pylint: disable=import-outside-toplevel
+
+ tokenizer = tokenizer_factory.get_tokenizer_for_db(config)
+ with tokenizer.name_analyzer() as analyzer:
+ with conn.cursor() as cur:
+ psycopg2.extras.register_hstore(cur)
+ cur.execute("SELECT country_code, name FROM country_name")
+
+ for country_code, names in cur:
+ if not names:
+ names = {}
+ names['countrycode'] = country_code
+ analyzer.add_country_names(country_code, names)