hans/Nominatim
1
0
Fork 1
mirror of https://github.com/osm-search/Nominatim.git synced 2026-02-26 11:08:13 +00:00
Code Issues Packages Projects Releases Wiki Activity

do not lookup by address vector when only few tokens are available

Browse Source 
Names of countries and states are exceedingly rare in the word count
but are very frequent in the address. A short name has the danger
of producing too many results.
This commit is contained in:
Sarah Hoffmann
2023-07-31 14:27:39 +02:00
parent 8adeaa2c7e
commit afdbdb02a1
2 changed files with 11 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
nominatim/api/search/db_search_builder.py
View File

@@ -212,7 +212,7 @@ class SearchBuilder:
exp_count = min(exp_count, min(t.count for t in addr_partials)) \ exp_count = min(exp_count, min(t.count for t in addr_partials)) \
if addr_partials else exp_count if addr_partials else exp_count
if exp_count < 1000 and partials_indexed: if exp_count < 1000 and len(addr_tokens) > 3 and partials_indexed:
# Lookup by address partials and restrict results through name terms. # Lookup by address partials and restrict results through name terms.
# Give this a small penalty because lookups in the address index are # Give this a small penalty because lookups in the address index are
# more expensive # more expensive

18
test/python/api/search/test_db_search_builder.py
View File

@@ -332,9 +332,10 @@ def test_name_only_search_with_countries():
assert not search.housenumbers.values assert not search.housenumbers.values
def make_counted_searches(name_part, name_full, address_part, address_full): def make_counted_searches(name_part, name_full, address_part, address_full,
num_address_parts=1):
q = QueryStruct([Phrase(PhraseType.NONE, '')]) q = QueryStruct([Phrase(PhraseType.NONE, '')])
for i in range(2): for i in range(1 + num_address_parts):
q.add_node(BreakType.WORD, PhraseType.NONE) q.add_node(BreakType.WORD, PhraseType.NONE)
q.add_node(BreakType.END, PhraseType.NONE) q.add_node(BreakType.END, PhraseType.NONE)
@@ -342,15 +343,16 @@ def make_counted_searches(name_part, name_full, address_part, address_full):
MyToken(0.5, 1, name_part, 'name_part', True)) MyToken(0.5, 1, name_part, 'name_part', True))
q.add_token(TokenRange(0, 1), TokenType.WORD, q.add_token(TokenRange(0, 1), TokenType.WORD,
MyToken(0, 101, name_full, 'name_full', True)) MyToken(0, 101, name_full, 'name_full', True))
q.add_token(TokenRange(1, 2), TokenType.PARTIAL, for i in range(num_address_parts):
MyToken(0.5, 2, address_part, 'address_part', True)) q.add_token(TokenRange(i + 1, i + 2), TokenType.PARTIAL,
q.add_token(TokenRange(1, 2), TokenType.WORD, MyToken(0.5, 2, address_part, 'address_part', True))
MyToken(0, 102, address_full, 'address_full', True)) q.add_token(TokenRange(i + 1, i + 2), TokenType.WORD,
MyToken(0, 102, address_full, 'address_full', True))
builder = SearchBuilder(q, SearchDetails()) builder = SearchBuilder(q, SearchDetails())
return list(builder.build(TokenAssignment(name=TokenRange(0, 1), return list(builder.build(TokenAssignment(name=TokenRange(0, 1),
address=[TokenRange(1, 2)]))) address=[TokenRange(1, 1 + num_address_parts)])))
def test_infrequent_partials_in_name(): def test_infrequent_partials_in_name():
@@ -368,7 +370,7 @@ def test_infrequent_partials_in_name():
def test_frequent_partials_in_name_but_not_in_address(): def test_frequent_partials_in_name_but_not_in_address():
searches = make_counted_searches(10000, 1, 1, 1) searches = make_counted_searches(10000, 1, 1, 1, num_address_parts=4)
assert len(searches) == 1 assert len(searches) == 1
search = searches[0] search = searches[0]