Nominatim/test/python/test_tools_postcodes.py
2021-05-13 14:15:42 +02:00

186 lines
6.9 KiB
Python

"""
Tests for functions to maintain the artificial postcode table.
"""
import subprocess
import pytest
from nominatim.tools import postcodes
import dummy_tokenizer
class MockPostcodeTable:
""" A location_postcode table for testing.
"""
def __init__(self, conn):
self.conn = conn
with conn.cursor() as cur:
cur.execute("""CREATE TABLE location_postcode (
place_id BIGINT,
parent_place_id BIGINT,
rank_search SMALLINT,
rank_address SMALLINT,
indexed_status SMALLINT,
indexed_date TIMESTAMP,
country_code varchar(2),
postcode TEXT,
geometry GEOMETRY(Geometry, 4326))""")
cur.execute("""CREATE OR REPLACE FUNCTION token_normalized_postcode(postcode TEXT)
RETURNS TEXT AS $$ BEGIN RETURN postcode; END; $$ LANGUAGE plpgsql;
""")
conn.commit()
def add(self, country, postcode, x, y):
with self.conn.cursor() as cur:
cur.execute("""INSERT INTO location_postcode (place_id, indexed_status,
country_code, postcode,
geometry)
VALUES (nextval('seq_place'), 1, %s, %s,
'SRID=4326;POINT(%s %s)')""",
(country, postcode, x, y))
self.conn.commit()
@property
def row_set(self):
with self.conn.cursor() as cur:
cur.execute("""SELECT country_code, postcode,
ST_X(geometry), ST_Y(geometry)
FROM location_postcode""")
return set((tuple(row) for row in cur))
@pytest.fixture
def tokenizer():
return dummy_tokenizer.DummyTokenizer(None, None)
@pytest.fixture
def postcode_table(temp_db_conn, placex_table, word_table):
return MockPostcodeTable(temp_db_conn)
def test_import_postcodes_empty(dsn, postcode_table, tmp_path, tokenizer):
postcodes.update_postcodes(dsn, tmp_path, tokenizer)
assert not postcode_table.row_set
def test_import_postcodes_add_new(dsn, placex_table, postcode_table, tmp_path, tokenizer):
placex_table.add(country='xx', geom='POINT(10 12)',
address=dict(postcode='9486'))
postcode_table.add('yy', '9486', 99, 34)
postcodes.update_postcodes(dsn, tmp_path, tokenizer)
assert postcode_table.row_set == {('xx', '9486', 10, 12), }
def test_import_postcodes_replace_coordinates(dsn, placex_table, postcode_table, tmp_path, tokenizer):
placex_table.add(country='xx', geom='POINT(10 12)',
address=dict(postcode='AB 4511'))
postcode_table.add('xx', 'AB 4511', 99, 34)
postcodes.update_postcodes(dsn, tmp_path, tokenizer)
assert postcode_table.row_set == {('xx', 'AB 4511', 10, 12)}
def test_import_postcodes_replace_coordinates_close(dsn, placex_table, postcode_table, tmp_path, tokenizer):
placex_table.add(country='xx', geom='POINT(10 12)',
address=dict(postcode='AB 4511'))
postcode_table.add('xx', 'AB 4511', 10, 11.99999)
postcodes.update_postcodes(dsn, tmp_path, tokenizer)
assert postcode_table.row_set == {('xx', 'AB 4511', 10, 11.99999)}
def test_import_postcodes_remove(dsn, placex_table, postcode_table, tmp_path, tokenizer):
placex_table.add(country='xx', geom='POINT(10 12)',
address=dict(postcode='AB 4511'))
postcode_table.add('xx', 'badname', 10, 12)
postcodes.update_postcodes(dsn, tmp_path, tokenizer)
assert postcode_table.row_set == {('xx', 'AB 4511', 10, 12)}
def test_import_postcodes_ignore_empty_country(dsn, placex_table, postcode_table, tmp_path, tokenizer):
placex_table.add(country=None, geom='POINT(10 12)',
address=dict(postcode='AB 4511'))
postcodes.update_postcodes(dsn, tmp_path, tokenizer)
assert not postcode_table.row_set
def test_import_postcodes_remove_all(dsn, placex_table, postcode_table, tmp_path, tokenizer):
postcode_table.add('ch', '5613', 10, 12)
postcodes.update_postcodes(dsn, tmp_path, tokenizer)
assert not postcode_table.row_set
def test_import_postcodes_multi_country(dsn, placex_table, postcode_table, tmp_path, tokenizer):
placex_table.add(country='de', geom='POINT(10 12)',
address=dict(postcode='54451'))
placex_table.add(country='cc', geom='POINT(100 56)',
address=dict(postcode='DD23 T'))
placex_table.add(country='de', geom='POINT(10.3 11.0)',
address=dict(postcode='54452'))
placex_table.add(country='cc', geom='POINT(10.3 11.0)',
address=dict(postcode='54452'))
postcodes.update_postcodes(dsn, tmp_path, tokenizer)
assert postcode_table.row_set == {('de', '54451', 10, 12),
('de', '54452', 10.3, 11.0),
('cc', '54452', 10.3, 11.0),
('cc', 'DD23 T', 100, 56)}
@pytest.mark.parametrize("gzipped", [True, False])
def test_import_postcodes_extern(dsn, placex_table, postcode_table, tmp_path,
tokenizer, gzipped):
placex_table.add(country='xx', geom='POINT(10 12)',
address=dict(postcode='AB 4511'))
extfile = tmp_path / 'xx_postcodes.csv'
extfile.write_text("postcode,lat,lon\nAB 4511,-4,-1\nCD 4511,-5, -10")
if gzipped:
subprocess.run(['gzip', str(extfile)])
assert not extfile.is_file()
postcodes.update_postcodes(dsn, tmp_path, tokenizer)
assert postcode_table.row_set == {('xx', 'AB 4511', 10, 12),
('xx', 'CD 4511', -10, -5)}
def test_import_postcodes_extern_bad_column(dsn, placex_table, postcode_table,
tmp_path, tokenizer):
placex_table.add(country='xx', geom='POINT(10 12)',
address=dict(postcode='AB 4511'))
extfile = tmp_path / 'xx_postcodes.csv'
extfile.write_text("postode,lat,lon\nAB 4511,-4,-1\nCD 4511,-5, -10")
postcodes.update_postcodes(dsn, tmp_path, tokenizer)
assert postcode_table.row_set == {('xx', 'AB 4511', 10, 12)}
def test_import_postcodes_extern_bad_number(dsn, placex_table, postcode_table,
tmp_path, tokenizer):
placex_table.add(country='xx', geom='POINT(10 12)',
address=dict(postcode='AB 4511'))
extfile = tmp_path / 'xx_postcodes.csv'
extfile.write_text("postcode,lat,lon\nXX 4511,-4,NaN\nCD 4511,-5, -10\n34,200,0")
postcodes.update_postcodes(dsn, tmp_path, tokenizer)
assert postcode_table.row_set == {('xx', 'AB 4511', 10, 12),
('xx', 'CD 4511', -10, -5)}