X-Git-Url: https://git.openstreetmap.org./nominatim.git/blobdiff_plain/25557e5f14734aa597dbe32293be25d2c539531c..7c7d15fe82f5516f6d65308bb6318618921fb2dc:/test/bdd/steps/steps_db_ops.py diff --git a/test/bdd/steps/steps_db_ops.py b/test/bdd/steps/steps_db_ops.py index 6bee88c1..8df5d617 100644 --- a/test/bdd/steps/steps_db_ops.py +++ b/test/bdd/steps/steps_db_ops.py @@ -1,43 +1,19 @@ -import re +# 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. +import logging +from itertools import chain + import psycopg2.extras -from check_functions import Almost from place_inserter import PlaceColumn +from table_compare import NominatimID, DBRow -class PlaceObjName(object): - - def __init__(self, placeid, conn): - self.pid = placeid - self.conn = conn - - def __str__(self): - if self.pid is None: - return "" - - if self.pid == 0: - return "place ID 0" - - cur = self.conn.cursor() - cur.execute("""SELECT osm_type, osm_id, class - FROM placex WHERE place_id = %s""", - (self.pid, )) - assert cur.rowcount == 1, "No entry found for place id %s" % self.pid - - return "%s%s:%s" % cur.fetchone() - -def compare_place_id(expected, result, column, context): - if expected == '0': - assert result == 0, \ - "Bad place id in column {}. Expected: 0, got: {!s}.".format( - column, PlaceObjName(result, context.db)) - elif expected == '-': - assert result is None, \ - "Bad place id in column {}: {!s}.".format( - column, PlaceObjName(result, context.db)) - else: - assert NominatimID(expected).get_place_id(context.db.cursor()) == result, \ - "Bad place id in column {}. Expected: {}, got: {!s}.".format( - column, expected, PlaceObjName(result, context.db)) +from nominatim.indexer import indexer +from nominatim.tokenizer import factory as tokenizer_factory def check_database_integrity(context): """ Check some generic constraints on the tables. @@ -51,88 +27,13 @@ def check_database_integrity(context): assert cur.fetchone()[0] == 0, "Duplicates found in place_addressline" -class NominatimID: - """ Splits a unique identifier for places into its components. - As place_ids cannot be used for testing, we use a unique - identifier instead that is of the form [:]. - """ - - id_regex = re.compile(r"(?P[NRW])(?P\d+)(:(?P\w+))?") - - def __init__(self, oid): - self.typ = self.oid = self.cls = None - - if oid is not None: - m = self.id_regex.fullmatch(oid) - assert m is not None, "ID '%s' not of form [:]" % oid - - self.typ = m.group('tp') - self.oid = m.group('id') - self.cls = m.group('cls') - - def __str__(self): - if self.cls is None: - return self.typ + self.oid - - return '%s%d:%s' % (self.typ, self.oid, self.cls) - - def table_select(self): - """ Return where clause and parameter list to select the object - from a Nominatim table. - """ - where = 'osm_type = %s and osm_id = %s' - params = [self.typ, self. oid] - - if self.cls is not None: - where += ' and class = %s' - params.append(self.cls) - - return where, params - - def get_place_id(self, cur): - where, params = self.table_select() - cur.execute("SELECT place_id FROM placex WHERE %s" % where, params) - assert cur.rowcount == 1, \ - "Expected exactly 1 entry in placex for %s found %s" % (str(self), cur.rowcount) - - return cur.fetchone()[0] - - -def assert_db_column(row, column, value, context): - if column == 'object': - return - - if column.startswith('centroid'): - if value == 'in geometry': - query = """SELECT ST_Within(ST_SetSRID(ST_Point({}, {}), 4326), - ST_SetSRID('{}'::geometry, 4326))""".format( - row['cx'], row['cy'], row['geomtxt']) - cur = context.db.cursor() - cur.execute(query) - assert cur.fetchone()[0], "(Row %s failed: %s)" % (column, query) - else: - fac = float(column[9:]) if column.startswith('centroid*') else 1.0 - x, y = value.split(' ') - assert Almost(float(x) * fac) == row['cx'], "Bad x coordinate" - assert Almost(float(y) * fac) == row['cy'], "Bad y coordinate" - elif column == 'geometry': - geom = context.osm.parse_geometry(value, context.scene) - cur = context.db.cursor() - query = "SELECT ST_Equals(ST_SnapToGrid(%s, 0.00001, 0.00001), ST_SnapToGrid(ST_SetSRID('%s'::geometry, 4326), 0.00001, 0.00001))" % ( - geom, row['geomtxt'],) - cur.execute(query) - assert cur.fetchone()[0], "(Row %s failed: %s)" % (column, query) - elif value == '-': - assert row[column] is None, "Row %s" % column - else: - assert value == str(row[column]), \ - "Row '%s': expected: %s, got: %s" % (column, value, str(row[column])) - - ################################ GIVEN ################################## @given("the (?Pnamed )?places") def add_data_to_place_table(context, named): + """ Add entries into the place table. 'named places' makes sure that + the entries get a random name when none is explicitly given. + """ with context.db.cursor() as cur: cur.execute('ALTER TABLE place DISABLE TRIGGER place_before_insert') for row in context.table: @@ -141,6 +42,9 @@ def add_data_to_place_table(context, named): @given("the relations") def add_data_to_planet_relations(context): + """ Add entries into the osm2pgsql relation middle table. This is needed + for tests on data that looks up members. + """ with context.db.cursor() as cur: for r in context.table: last_node = 0 @@ -164,28 +68,24 @@ def add_data_to_planet_relations(context): else: members = None - tags = [] - for h in r.headings: - if h.startswith("tags+"): - tags.extend((h[5:], r[h])) + tags = chain.from_iterable([(h[5:], r[h]) for h in r.headings if h.startswith("tags+")]) cur.execute("""INSERT INTO planet_osm_rels (id, way_off, rel_off, parts, members, tags) VALUES (%s, %s, %s, %s, %s, %s)""", - (r['id'], last_node, last_way, parts, members, tags)) + (r['id'], last_node, last_way, parts, members, list(tags))) @given("the ways") def add_data_to_planet_ways(context): + """ Add entries into the osm2pgsql way middle table. This is necessary for + tests on that that looks up node ids in this table. + """ with context.db.cursor() as cur: for r in context.table: - tags = [] - for h in r.headings: - if h.startswith("tags+"): - tags.extend((h[5:], r[h])) - + tags = chain.from_iterable([(h[5:], r[h]) for h in r.headings if h.startswith("tags+")]) nodes = [ int(x.strip()) for x in r['nodes'].split(',') ] cur.execute("INSERT INTO planet_osm_ways (id, nodes, tags) VALUES (%s, %s, %s)", - (r['id'], nodes, tags)) + (r['id'], nodes, list(tags))) ################################ WHEN ################################## @@ -193,14 +93,17 @@ def add_data_to_planet_ways(context): def import_and_index_data_from_place_table(context): """ Import data previously set up in the place table. """ - context.nominatim.copy_from_place(context.db) - context.nominatim.run_setup_script('calculate-postcodes', 'index', 'index-noanalyse') + context.nominatim.run_nominatim('import', '--continue', 'load-data', + '--index-noanalyse', '-q') + check_database_integrity(context) @when("updating places") def update_place_table(context): - context.nominatim.run_setup_script( - 'create-functions', 'create-partition-functions', 'enable-diff-updates') + """ Update the place table with the given data. Also runs all triggers + related to updates and reindexes the new data. + """ + context.nominatim.run_nominatim('refresh', '--functions') with context.db.cursor() as cur: for row in context.table: PlaceColumn(context).add_row(row, False).db_insert(cur) @@ -210,185 +113,187 @@ def update_place_table(context): @when("updating postcodes") def update_postcodes(context): - context.nominatim.run_update_script('calculate-postcodes') + """ Rerun the calculation of postcodes. + """ + context.nominatim.run_nominatim('refresh', '--postcodes') @when("marking for delete (?P.*)") def delete_places(context, oids): - context.nominatim.run_setup_script( - 'create-functions', 'create-partition-functions', 'enable-diff-updates') + """ Remove entries from the place table. Multiple ids may be given + separated by commas. Also runs all triggers + related to updates and reindexes the new data. + """ + context.nominatim.run_nominatim('refresh', '--functions') with context.db.cursor() as cur: for oid in oids.split(','): - where, params = NominatimID(oid).table_select() - cur.execute("DELETE FROM place WHERE " + where, params) + NominatimID(oid).query_osm_id(cur, 'DELETE FROM place WHERE {}') context.nominatim.reindex_placex(context.db) ################################ THEN ################################## -@then("placex contains(?P exactly)?") -def check_placex_contents(context, exact): +@then("(?Pplacex|place) contains(?P exactly)?") +def check_place_contents(context, table, exact): + """ Check contents of place/placex tables. Each row represents a table row + and all data must match. Data not present in the expected table, may + be arbitry. The rows are identified via the 'object' column which must + have an identifier of the form '[:]'. When multiple + rows match (for example because 'class' was left out and there are + multiple entries for the given OSM object) then all must match. All + expected rows are expected to be present with at least one database row. + When 'exactly' is given, there must not be additional rows in the database. + """ with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: expected_content = set() for row in context.table: nid = NominatimID(row['object']) - where, params = nid.table_select() - cur.execute("""SELECT *, ST_AsText(geometry) as geomtxt, - ST_X(centroid) as cx, ST_Y(centroid) as cy - FROM placex where %s""" % where, - params) + query = 'SELECT *, ST_AsText(geometry) as geomtxt, ST_GeometryType(geometry) as geometrytype' + if table == 'placex': + query += ' ,ST_X(centroid) as cx, ST_Y(centroid) as cy' + query += " FROM %s WHERE {}" % (table, ) + nid.query_osm_id(cur, query) assert cur.rowcount > 0, "No rows found for " + row['object'] for res in cur: if exact: expected_content.add((res['osm_type'], res['osm_id'], res['class'])) - for h in row.headings: - if h in ('extratags', 'address'): - if row[h] == '-': - assert res[h] is None - else: - vdict = eval('{' + row[h] + '}') - assert vdict == res[h] - elif h.startswith('name'): - name = h[5:] if h.startswith('name+') else 'name' - assert name in res['name'] - assert res['name'][name] == row[h] - elif h.startswith('extratags+'): - assert res['extratags'][h[10:]] == row[h] - elif h.startswith('addr+'): - if row[h] == '-': - if res['address'] is not None: - assert h[5:] not in res['address'] - else: - assert h[5:] in res['address'], "column " + h - assert res['address'][h[5:]] == row[h], "column %s" % h - elif h in ('linked_place_id', 'parent_place_id'): - compare_place_id(row[h], res[h], h, context) - else: - assert_db_column(res, h, row[h], context) + + DBRow(nid, res, context).assert_row(row, ['object']) if exact: - cur.execute('SELECT osm_type, osm_id, class from placex') + cur.execute('SELECT osm_type, osm_id, class from {}'.format(table)) assert expected_content == set([(r[0], r[1], r[2]) for r in cur]) -@then("place contains(?P exactly)?") -def check_placex_contents(context, exact): - with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: - expected_content = set() - for row in context.table: - nid = NominatimID(row['object']) - where, params = nid.table_select() - cur.execute("""SELECT *, ST_AsText(geometry) as geomtxt, - ST_GeometryType(geometry) as geometrytype - FROM place where %s""" % where, - params) - assert cur.rowcount > 0, "No rows found for " + row['object'] - for res in cur: - if exact: - expected_content.add((res['osm_type'], res['osm_id'], res['class'])) - for h in row.headings: - msg = "%s: %s" % (row['object'], h) - if h in ('name', 'extratags', 'address'): - if row[h] == '-': - assert res[h] is None, msg - else: - vdict = eval('{' + row[h] + '}') - assert vdict == res[h], msg - elif h.startswith('name+'): - assert res['name'][h[5:]] == row[h], msg - elif h.startswith('extratags+'): - assert res['extratags'][h[10:]] == row[h], msg - elif h.startswith('addr+'): - if row[h] == '-': - if res['address'] is not None: - assert h[5:] not in res['address'] - else: - assert res['address'][h[5:]] == row[h], msg - elif h in ('linked_place_id', 'parent_place_id'): - compare_place_id(row[h], res[h], h, context) - else: - assert_db_column(res, h, row[h], context) +@then("(?P
placex|place) has no entry for (?P.*)") +def check_place_has_entry(context, table, oid): + """ Ensure that no database row for the given object exists. The ID + must be of the form '[:]'. + """ + with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: + NominatimID(oid).query_osm_id(cur, "SELECT * FROM %s where {}" % table) + assert cur.rowcount == 0, \ + "Found {} entries for ID {}".format(cur.rowcount, oid) - if exact: - cur.execute('SELECT osm_type, osm_id, class from place') - assert expected_content, set([(r[0], r[1], r[2]) for r in cur]) @then("search_name contains(?P not)?") def check_search_name_contents(context, exclude): - with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: - for row in context.table: - pid = NominatimID(row['object']).get_place_id(cur) - cur.execute("""SELECT *, ST_X(centroid) as cx, ST_Y(centroid) as cy - FROM search_name WHERE place_id = %s""", (pid, )) - assert cur.rowcount > 0, "No rows found for " + row['object'] + """ Check contents of place/placex tables. Each row represents a table row + and all data must match. Data not present in the expected table, may + be arbitry. The rows are identified via the 'object' column which must + have an identifier of the form '[:]'. All + expected rows are expected to be present with at least one database row. + """ + tokenizer = tokenizer_factory.get_tokenizer_for_db(context.nominatim.get_test_config()) + + with tokenizer.name_analyzer() as analyzer: + with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: + for row in context.table: + nid = NominatimID(row['object']) + nid.row_by_place_id(cur, 'search_name', + ['ST_X(centroid) as cx', 'ST_Y(centroid) as cy']) + assert cur.rowcount > 0, "No rows found for " + row['object'] + + for res in cur: + db_row = DBRow(nid, res, context) + for name, value in zip(row.headings, row.cells): + if name in ('name_vector', 'nameaddress_vector'): + items = [x.strip() for x in value.split(',')] + tokens = analyzer.get_word_token_info(items) - for res in cur: - for h in row.headings: - if h in ('name_vector', 'nameaddress_vector'): - terms = [x.strip() for x in row[h].split(',') if not x.strip().startswith('#')] - words = [x.strip()[1:] for x in row[h].split(',') if x.strip().startswith('#')] - with context.db.cursor() as subcur: - subcur.execute(""" SELECT word_id, word_token - FROM word, (SELECT unnest(%s::TEXT[]) as term) t - WHERE word_token = make_standard_name(t.term) - and class is null and country_code is null - and operator is null - UNION - SELECT word_id, word_token - FROM word, (SELECT unnest(%s::TEXT[]) as term) t - WHERE word_token = ' ' || make_standard_name(t.term) - and class is null and country_code is null - and operator is null - """, - (terms, words)) if not exclude: - assert subcur.rowcount >= len(terms) + len(words), \ - "No word entry found for " + row[h] + ". Entries found: " + str(subcur.rowcount) - for wid in subcur: + assert len(tokens) >= len(items), \ + "No word entry found for {}. Entries found: {!s}".format(value, len(tokens)) + for word, token, wid in tokens: if exclude: - assert wid[0] not in res[h], "Found term for %s/%s: %s" % (pid, h, wid[1]) + assert wid not in res[name], \ + "Found term for {}/{}: {}".format(nid, name, wid) else: - assert wid[0] in res[h], "Missing term for %s/%s: %s" % (pid, h, wid[1]) - else: - assert_db_column(res, h, row[h], context) + assert wid in res[name], \ + "Missing term for {}/{}: {}".format(nid, name, wid) + elif name != 'object': + assert db_row.contains(name, value), db_row.assert_msg(name, value) + +@then("search_name has no entry for (?P.*)") +def check_search_name_has_entry(context, oid): + """ Check that there is noentry in the search_name table for the given + objects. IDs are in format '[:]'. + """ + with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: + NominatimID(oid).row_by_place_id(cur, 'search_name') + + assert cur.rowcount == 0, \ + "Found {} entries for ID {}".format(cur.rowcount, oid) @then("location_postcode contains exactly") def check_location_postcode(context): + """ Check full contents for location_postcode table. Each row represents a table row + and all data must match. Data not present in the expected table, may + be arbitry. The rows are identified via 'country' and 'postcode' columns. + All rows must be present as excepted and there must not be additional + rows. + """ with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: cur.execute("SELECT *, ST_AsText(geometry) as geomtxt FROM location_postcode") assert cur.rowcount == len(list(context.table)), \ - "Postcode table has %d rows, expected %d rows." % (cur.rowcount, len(list(context.table))) + "Postcode table has {} rows, expected {}.".format(cur.rowcount, len(list(context.table))) + + results = {} + for row in cur: + key = (row['country_code'], row['postcode']) + assert key not in results, "Postcode table has duplicate entry: {}".format(row) + results[key] = DBRow((row['country_code'],row['postcode']), row, context) - table = list(cur) for row in context.table: - for i in range(len(table)): - if table[i]['country_code'] != row['country'] \ - or table[i]['postcode'] != row['postcode']: - continue - for h in row.headings: - if h not in ('country', 'postcode'): - assert_db_column(table[i], h, row[h], context) - -@then("word contains(?P not)?") -def check_word_table(context, exclude): + db_row = results.get((row['country'],row['postcode'])) + assert db_row is not None, \ + f"Missing row for country '{row['country']}' postcode '{row['postcode']}'." + + db_row.assert_row(row, ('country', 'postcode')) + +@then("there are(?P no)? word tokens for postcodes (?P.*)") +def check_word_table_for_postcodes(context, exclude, postcodes): + """ Check that the tokenizer produces postcode tokens for the given + postcodes. The postcodes are a comma-separated list of postcodes. + Whitespace matters. + """ + nctx = context.nominatim + tokenizer = tokenizer_factory.get_tokenizer_for_db(nctx.get_test_config()) + with tokenizer.name_analyzer() as ana: + plist = [ana.normalize_postcode(p) for p in postcodes.split(',')] + + plist.sort() + with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: - for row in context.table: - wheres = [] - values = [] - for h in row.headings: - wheres.append("%s = %%s" % h) - values.append(row[h]) - cur.execute("SELECT * from word WHERE %s" % ' AND '.join(wheres), values) - if exclude: - assert cur.rowcount == 0, "Row still in word table: %s" % '/'.join(values) - else: - assert cur.rowcount > 0, "Row not in word table: %s" % '/'.join(values) + if nctx.tokenizer == 'icu': + cur.execute("SELECT word FROM word WHERE type = 'P' and word = any(%s)", + (plist,)) + else: + cur.execute("""SELECT word FROM word WHERE word = any(%s) + and class = 'place' and type = 'postcode'""", + (plist,)) + + found = [row[0] for row in cur] + assert len(found) == len(set(found)), f"Duplicate rows for postcodes: {found}" + + if exclude: + assert len(found) == 0, f"Unexpected postcodes: {found}" + else: + assert set(found) == set(plist), \ + f"Missing postcodes {set(plist) - set(found)}. Found: {found}" @then("place_addressline contains") def check_place_addressline(context): + """ Check the contents of the place_addressline table. Each row represents + a table row and all data must match. Data not present in the expected + table, may be arbitry. The rows are identified via the 'object' column, + representing the addressee and the 'address' column, representing the + address item. + """ with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: for row in context.table: - pid = NominatimID(row['object']).get_place_id(cur) + nid = NominatimID(row['object']) + pid = nid.get_place_id(cur) apid = NominatimID(row['address']).get_place_id(cur) cur.execute(""" SELECT * FROM place_addressline WHERE place_id = %s AND address_place_id = %s""", @@ -397,36 +302,36 @@ def check_place_addressline(context): "No rows found for place %s and address %s" % (row['object'], row['address']) for res in cur: - for h in row.headings: - if h not in ('address', 'object'): - assert_db_column(res, h, row[h], context) + DBRow(nid, res, context).assert_row(row, ('address', 'object')) @then("place_addressline doesn't contain") def check_place_addressline_exclude(context): + """ Check that the place_addressline doesn't contain any entries for the + given addressee/address item pairs. + """ with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: for row in context.table: pid = NominatimID(row['object']).get_place_id(cur) - apid = NominatimID(row['address']).get_place_id(cur) - cur.execute(""" SELECT * FROM place_addressline - WHERE place_id = %s AND address_place_id = %s""", - (pid, apid)) - assert cur.rowcount == 0, \ - "Row found for place %s and address %s" % (row['object'], row['address']) - -@then("(?P\w+) expands to(?P no)? interpolation") + apid = NominatimID(row['address']).get_place_id(cur, allow_empty=True) + if apid is not None: + cur.execute(""" SELECT * FROM place_addressline + WHERE place_id = %s AND address_place_id = %s""", + (pid, apid)) + assert cur.rowcount == 0, \ + "Row found for place %s and address %s" % (row['object'], row['address']) + +@then("W(?P\d+) expands to(?P no)? interpolation") def check_location_property_osmline(context, oid, neg): - nid = NominatimID(oid) - - assert 'W' == nid.typ, "interpolation must be a way" - + """ Check that the given way is present in the interpolation table. + """ with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: cur.execute("""SELECT *, ST_AsText(linegeo) as geomtxt FROM location_property_osmline WHERE osm_id = %s AND startnumber IS NOT NULL""", - (nid.oid, )) + (oid, )) if neg: - assert cur.rowcount == 0 + assert cur.rowcount == 0, "Interpolation found for way {}.".format(oid) return todo = list(range(len(list(context.table)))) @@ -438,30 +343,10 @@ def check_location_property_osmline(context, oid, neg): todo.remove(i) break else: - assert False, "Unexpected row %s" % (str(res)) + assert False, "Unexpected row " + str(res) - for h in row.headings: - if h in ('start', 'end'): - continue - elif h == 'parent_place_id': - compare_place_id(row[h], res[h], h, context) - else: - assert_db_column(res, h, row[h], context) + DBRow(oid, res, context).assert_row(row, ('start', 'end')) assert not todo -@then("(?P
placex|place) has no entry for (?P.*)") -def check_placex_has_entry(context, table, oid): - with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: - nid = NominatimID(oid) - where, params = nid.table_select() - cur.execute("SELECT * FROM %s where %s" % (table, where), params) - assert cur.rowcount == 0 - -@then("search_name has no entry for (?P.*)") -def check_search_name_has_entry(context, oid): - with context.db.cursor(cursor_factory=psycopg2.extras.DictCursor) as cur: - pid = NominatimID(oid).get_place_id(cur) - cur.execute("SELECT * FROM search_name WHERE place_id = %s", (pid, )) - assert cur.rowcount == 0