From 60b03d506f60fdbaf6adc23242ca8367db8393b4 Mon Sep 17 00:00:00 2001 From: Sarah Hoffmann Date: Sun, 5 May 2024 09:39:52 +0200 Subject: [PATCH] add CSV format for importance import --- lib-sql/functions/importance.sql | 62 ++++++++++++++++++++-- lib-sql/tables.sql | 31 ++++------- nominatim/clicmd/refresh.py | 19 ++++--- nominatim/db/utils.py | 8 +++ nominatim/tools/refresh.py | 80 +++++++++++++++++++++++++++-- test/python/cli/test_cmd_refresh.py | 6 ++- 6 files changed, 167 insertions(+), 39 deletions(-) diff --git a/lib-sql/functions/importance.sql b/lib-sql/functions/importance.sql index 9b2fb773..22a87240 100644 --- a/lib-sql/functions/importance.sql +++ b/lib-sql/functions/importance.sql @@ -20,6 +20,54 @@ CREATE TYPE place_importance as ( wikipedia TEXT ); +{% if 'wikimedia_importance' in db.tables %} + +CREATE OR REPLACE FUNCTION get_wikipedia_match(extratags HSTORE, country_code varchar(2)) + RETURNS wikipedia_article_match + AS $$ +DECLARE + i INT; + wiki_article_title TEXT; + wiki_article_language TEXT; + result wikipedia_article_match; + entry RECORD; +BEGIN + IF extratags ? 'wikipedia' and strpos(extratags->'wikipedia', ':') IN (3,4) THEN + wiki_article_language := lower(trim(split_part(extratags->'wikipedia', ':', 1))); + wiki_article_title := trim(substr(extratags->'wikipedia', + strpos(extratags->'wikipedia', ':') + 1)); + + FOR result IN + SELECT language, title, importance FROM wikimedia_importance + WHERE language = wiki_article_language + and title = replace(wiki_article_title, ' ', '_') + LOOP + RETURN result; + END LOOP; + END IF; + + FOREACH wiki_article_language IN ARRAY ARRAY['ar','bg','ca','cs','da','de','en','es','eo','eu','fa','fr','ko','hi','hr','id','it','he','lt','hu','ms','nl','ja','no','pl','pt','kk','ro','ru','sk','sl','sr','fi','sv','tr','uk','vi','vo','war','zh'] + LOOP + IF extratags ? ('wikipedia:' || wiki_article_language) THEN + wiki_article_title := extratags->('wikipedia:' || wiki_article_language); + + FOR result IN + SELECT language, title, importance FROM wikimedia_importance + WHERE language = wiki_article_language + and title = replace(wiki_article_title, ' ', '_') + LOOP + RETURN result; + END LOOP; + END IF; + + END LOOP; + + RETURN NULL; +END; +$$ +LANGUAGE plpgsql IMMUTABLE STRICT; + +{% else %} -- See: http://stackoverflow.com/questions/6410088/how-can-i-mimic-the-php-urldecode-function-in-postgresql CREATE OR REPLACE FUNCTION decode_url_part(p varchar) @@ -93,6 +141,7 @@ END; $$ LANGUAGE plpgsql STABLE; +{% endif %} CREATE OR REPLACE FUNCTION compute_importance(extratags HSTORE, country_code varchar(2), @@ -118,9 +167,16 @@ BEGIN -- Nothing? Then try with the wikidata tag. IF result.importance is null AND extratags ? 'wikidata' THEN - FOR match IN SELECT * FROM wikipedia_article - WHERE wd_page_title = extratags->'wikidata' - ORDER BY language = 'en' DESC, langcount DESC LIMIT 1 + FOR match IN +{% if 'wikimedia_importance' in db.tables %} + SELECT * FROM wikimedia_importance + WHERE wikidata = extratags->'wikidata' + LIMIT 1 +{% else %} + SELECT * FROM wikipedia_article + WHERE wd_page_title = extratags->'wikidata' + ORDER BY language = 'en' DESC, langcount DESC LIMIT 1 +{% endif %} LOOP result.importance := match.importance; result.wikipedia := match.language || ':' || match.title; diff --git a/lib-sql/tables.sql b/lib-sql/tables.sql index eafed6d8..d3bc972a 100644 --- a/lib-sql/tables.sql +++ b/lib-sql/tables.sql @@ -273,28 +273,15 @@ GRANT SELECT ON import_polygon_delete TO "{{config.DATABASE_WEBUSER}}"; DROP SEQUENCE IF EXISTS file; CREATE SEQUENCE file start 1; --- null table so it won't error --- deliberately no drop - importing the table is expensive and static, if it is already there better to avoid removing it -CREATE TABLE IF NOT EXISTS wikipedia_article ( - language text NOT NULL, - title text NOT NULL, - langcount integer, - othercount integer, - totalcount integer, - lat double precision, - lon double precision, - importance double precision, - osm_type character(1), - osm_id bigint, - wd_page_title text, - instance_of text -); - -CREATE TABLE IF NOT EXISTS wikipedia_redirect ( - language text, - from_title text, - to_title text -); +{% if 'wikimedia_importance' not in db.tables and 'wikipedia_article' not in db.tables %} +-- create dummy tables here, if nothing was imported +CREATE TABLE wikimedia_importance ( + language TEXT NOT NULL, + title TEXT NOT NULL, + importance double precision NOT NULL, + wikidata TEXT +) {{db.tablespace.address_data}}; +{% endif %} -- osm2pgsql does not create indexes on the middle tables for Nominatim -- Add one for lookup of associated street relations. diff --git a/nominatim/clicmd/refresh.py b/nominatim/clicmd/refresh.py index 343fe48d..5eac53da 100644 --- a/nominatim/clicmd/refresh.py +++ b/nominatim/clicmd/refresh.py @@ -89,6 +89,7 @@ class UpdateRefresh: from ..tools import refresh, postcodes from ..indexer.indexer import Indexer + need_function_refresh = args.functions if args.postcodes: if postcodes.can_compute(args.config.get_libpq_dsn()): @@ -131,13 +132,7 @@ class UpdateRefresh: args.project_dir) > 0: LOG.fatal('FATAL: Cannot update secondary importance raster data') return 1 - - if args.functions: - LOG.warning('Create functions') - with connect(args.config.get_libpq_dsn()) as conn: - refresh.create_functions(conn, args.config, - args.diffs, args.enable_debug_statements) - self._get_tokenizer(args.config).update_sql_functions(args.config) + need_function_refresh = True if args.wiki_data: data_path = Path(args.config.WIKIPEDIA_DATA_PATH @@ -147,8 +142,16 @@ class UpdateRefresh: data_path) > 0: LOG.fatal('FATAL: Wikipedia importance file not found in %s', data_path) return 1 + need_function_refresh = True + + if need_function_refresh: + LOG.warning('Create functions') + with connect(args.config.get_libpq_dsn()) as conn: + refresh.create_functions(conn, args.config, + args.diffs, args.enable_debug_statements) + self._get_tokenizer(args.config).update_sql_functions(args.config) - # Attention: importance MUST come after wiki data import. + # Attention: importance MUST come after wiki data import and after functions. if args.importance: LOG.warning('Update importance values for database') with connect(args.config.get_libpq_dsn()) as conn: diff --git a/nominatim/db/utils.py b/nominatim/db/utils.py index e3f0712a..57048da3 100644 --- a/nominatim/db/utils.py +++ b/nominatim/db/utils.py @@ -92,6 +92,11 @@ class CopyBuffer: return self + def size(self) -> int: + """ Return the number of bytes the buffer currently contains. + """ + return self.buffer.tell() + def __exit__(self, exc_type: Any, exc_value: Any, traceback: Any) -> None: if self.buffer is not None: self.buffer.close() @@ -115,7 +120,10 @@ class CopyBuffer: def copy_out(self, cur: Cursor, table: str, columns: Optional[Iterable[str]] = None) -> None: """ Copy all collected data into the given table. + + The buffer is empty and reusable after this operation. """ if self.buffer.tell() > 0: self.buffer.seek(0) cur.copy_from(self.buffer, table, columns=columns) + self.buffer = io.StringIO() diff --git a/nominatim/tools/refresh.py b/nominatim/tools/refresh.py index 008fc714..a200ee13 100644 --- a/nominatim/tools/refresh.py +++ b/nominatim/tools/refresh.py @@ -8,6 +8,8 @@ Functions for bringing auxiliary data in the database up-to-date. """ from typing import MutableSequence, Tuple, Any, Type, Mapping, Sequence, List, cast +import csv +import gzip import logging from textwrap import dedent from pathlib import Path @@ -16,7 +18,7 @@ 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.utils import execute_file, CopyBuffer from nominatim.db.sql_preprocessor import SQLPreprocessor from nominatim.version import NOMINATIM_VERSION @@ -132,21 +134,89 @@ def import_wikipedia_articles(dsn: str, data_path: Path, ignore_errors: bool = F 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 import_importance_csv(dsn, data_path / 'wikimedia-importance.csv.gz') == 0 \ + or import_importance_sql(dsn, data_path / 'wikimedia-importance.sql.gz', + ignore_errors) == 0: + return 0 - if not datafile.exists(): + return 1 + + +def import_importance_csv(dsn: str, data_file: Path) -> int: + """ Replace wikipedia importance table with data from a + single CSV file. + + The file must be a gzipped CSV and have the following columns: + language, title, importance, wikidata_id + + Other columns may be present but will be ignored. + """ + if not data_file.exists(): + return 1 + + # Only import the first occurance of a wikidata ID. + # This keeps indexes and table small. + wd_done = set() + + with connect(dsn) as conn: + with conn.cursor() as cur: + cur.drop_table('wikipedia_article') + cur.drop_table('wikipedia_redirect') + cur.drop_table('wikimedia_importance') + cur.execute("""CREATE TABLE wikimedia_importance ( + language TEXT NOT NULL, + title TEXT NOT NULL, + importance double precision NOT NULL, + wikidata TEXT + ) """) + + with gzip.open(str(data_file), 'rt') as fd, CopyBuffer() as buf: + for row in csv.DictReader(fd, delimiter='\t', quotechar='|'): + wd_id = int(row['wikidata_id'][1:]) + buf.add(row['language'], row['title'], row['importance'], + None if wd_id in wd_done else row['wikidata_id']) + wd_done.add(wd_id) + + if buf.size() > 10000000: + with conn.cursor() as cur: + buf.copy_out(cur, 'wikimedia_importance', + columns=['language', 'title', 'importance', + 'wikidata']) + + with conn.cursor() as cur: + buf.copy_out(cur, 'wikimedia_importance', + columns=['language', 'title', 'importance', 'wikidata']) + + with conn.cursor() as cur: + cur.execute("""CREATE INDEX IF NOT EXISTS idx_wikimedia_importance_title + ON wikimedia_importance (title)""") + cur.execute("""CREATE INDEX IF NOT EXISTS idx_wikimedia_importance_wikidata + ON wikimedia_importance (wikidata) + WHERE wikidata is not null""") + + conn.commit() + + return 0 + + +def import_importance_sql(dsn: str, data_file: Path, ignore_errors: bool) -> int: + """ Replace wikipedia importance table with data from an SQL file. + """ + if not data_file.exists(): return 1 pre_code = """BEGIN; DROP TABLE IF EXISTS "wikipedia_article"; - DROP TABLE IF EXISTS "wikipedia_redirect" + DROP TABLE IF EXISTS "wikipedia_redirect"; + DROP TABLE IF EXISTS "wikipedia_importance"; """ post_code = "COMMIT" - execute_file(dsn, datafile, ignore_errors=ignore_errors, + execute_file(dsn, data_file, ignore_errors=ignore_errors, pre_code=pre_code, post_code=post_code) return 0 + def import_secondary_importance(dsn: str, data_path: Path, ignore_errors: bool = False) -> int: """ Replaces the secondary importance raster data table with new data. diff --git a/test/python/cli/test_cmd_refresh.py b/test/python/cli/test_cmd_refresh.py index f3f93f0f..1179f22c 100644 --- a/test/python/cli/test_cmd_refresh.py +++ b/test/python/cli/test_cmd_refresh.py @@ -28,6 +28,7 @@ class TestRefresh: ('website', 'setup_website'), ]) def test_refresh_command(self, mock_func_factory, command, func): + mock_func_factory(nominatim.tools.refresh, 'create_functions') func_mock = mock_func_factory(nominatim.tools.refresh, func) assert self.call_nominatim('refresh', '--' + command) == 0 @@ -71,6 +72,7 @@ class TestRefresh: assert self.call_nominatim('refresh', '--wiki-data') == 1 + def test_refresh_secondary_importance_file_not_found(self): assert self.call_nominatim('refresh', '--secondary-importance') == 1 @@ -84,16 +86,18 @@ class TestRefresh: assert mocks[1].called == 1 - def test_refresh_importance_computed_after_wiki_import(self, monkeypatch): + def test_refresh_importance_computed_after_wiki_import(self, monkeypatch, mock_func_factory): calls = [] monkeypatch.setattr(nominatim.tools.refresh, 'import_wikipedia_articles', lambda *args, **kwargs: calls.append('import') or 0) monkeypatch.setattr(nominatim.tools.refresh, 'recompute_importance', lambda *args, **kwargs: calls.append('update')) + func_mock = mock_func_factory(nominatim.tools.refresh, 'create_functions') assert self.call_nominatim('refresh', '--importance', '--wiki-data') == 0 assert calls == ['import', 'update'] + assert func_mock.called == 1 @pytest.mark.parametrize('params', [('--data-object', 'w234'), ('--data-object', 'N23', '--data-object', 'N24'), -- 2.39.5