From: Sarah Hoffmann Date: Sun, 22 Sep 2024 08:39:10 +0000 (+0200) Subject: remove support for unindexed tokens X-Git-Url: https://git.openstreetmap.org./nominatim.git/commitdiff_plain/a690605a96191ed0ac230493bfb8ae3aa6504988 remove support for unindexed tokens This was a special feature of the legacy tokenizer who would not index very frequent tokens. --- diff --git a/src/nominatim_api/search/db_search_builder.py b/src/nominatim_api/search/db_search_builder.py index 6453509e..1ac6db2b 100644 --- a/src/nominatim_api/search/db_search_builder.py +++ b/src/nominatim_api/search/db_search_builder.py @@ -167,8 +167,7 @@ class SearchBuilder: expected_count = sum(t.count for t in hnrs) partials = {t.token: t.addr_count for trange in address - for t in self.query.get_partials_list(trange) - if t.is_indexed} + for t in self.query.get_partials_list(trange)} if not partials: # can happen when none of the partials is indexed @@ -219,11 +218,9 @@ class SearchBuilder: addr_partials = [t for r in address for t in self.query.get_partials_list(r)] addr_tokens = list({t.token for t in addr_partials}) - partials_indexed = all(t.is_indexed for t in name_partials.values()) \ - and all(t.is_indexed for t in addr_partials) exp_count = min(t.count for t in name_partials.values()) / (2**(len(name_partials) - 1)) - if (len(name_partials) > 3 or exp_count < 8000) and partials_indexed: + if (len(name_partials) > 3 or exp_count < 8000): yield penalty, exp_count, dbf.lookup_by_names(list(name_partials.keys()), addr_tokens) return @@ -232,8 +229,6 @@ class SearchBuilder: name_fulls = self.query.get_tokens(name, TokenType.WORD) if name_fulls: fulls_count = sum(t.count for t in name_fulls) - if partials_indexed: - penalty += 1.2 * sum(t.penalty for t in addr_partials if not t.is_indexed) if fulls_count < 50000 or addr_count < 30000: yield penalty,fulls_count / (2**len(addr_tokens)), \ @@ -243,8 +238,7 @@ class SearchBuilder: # To catch remaining results, lookup by name and address # We only do this if there is a reasonable number of results expected. exp_count = exp_count / (2**len(addr_tokens)) if addr_tokens else exp_count - if exp_count < 10000 and addr_count < 20000\ - and all(t.is_indexed for t in name_partials.values()): + if exp_count < 10000 and addr_count < 20000: penalty += 0.35 * max(1 if name_fulls else 0.1, 5 - len(name_partials) - len(addr_tokens)) yield penalty, exp_count,\ @@ -260,11 +254,10 @@ class SearchBuilder: addr_restrict_tokens = [] addr_lookup_tokens = [] for t in addr_partials: - if t.is_indexed: - if t.addr_count > 20000: - addr_restrict_tokens.append(t.token) - else: - addr_lookup_tokens.append(t.token) + if t.addr_count > 20000: + addr_restrict_tokens.append(t.token) + else: + addr_lookup_tokens.append(t.token) if addr_restrict_tokens: lookup.append(dbf.FieldLookup('nameaddress_vector', @@ -289,13 +282,12 @@ class SearchBuilder: addr_restrict_tokens = [] addr_lookup_tokens = [] for t in addr_partials: - if t.is_indexed: - if t.addr_count > 20000: - addr_restrict_tokens.append(t.token) - else: - addr_lookup_tokens.append(t.token) + if t.addr_count > 20000: + addr_restrict_tokens.append(t.token) + else: + addr_lookup_tokens.append(t.token) else: - addr_restrict_tokens = [t.token for t in addr_partials if t.is_indexed] + addr_restrict_tokens = [t.token for t in addr_partials] addr_lookup_tokens = [] return dbf.lookup_by_any_name([t.token for t in name_fulls], diff --git a/src/nominatim_api/search/icu_tokenizer.py b/src/nominatim_api/search/icu_tokenizer.py index 971e95be..1aadc97e 100644 --- a/src/nominatim_api/search/icu_tokenizer.py +++ b/src/nominatim_api/search/icu_tokenizer.py @@ -123,7 +123,7 @@ class ICUToken(qmod.Token): lookup_word = row.word_token return ICUToken(penalty=penalty, token=row.word_id, count=max(1, count), - lookup_word=lookup_word, is_indexed=True, + lookup_word=lookup_word, word_token=row.word_token, info=row.info, addr_count=max(1, addr_count)) @@ -259,7 +259,9 @@ class ICUQueryAnalyzer(AbstractQueryAnalyzer): if len(part.token) <= 4 and part[0].isdigit()\ and not node.has_tokens(i+1, qmod.TokenType.HOUSENUMBER): query.add_token(qmod.TokenRange(i, i+1), qmod.TokenType.HOUSENUMBER, - ICUToken(0.5, 0, 1, 1, part.token, True, part.token, None)) + ICUToken(penalty=0.5, token=0, + count=1, addr_count=1, lookup_word=part.token, + word_token=part.token, info=None)) def rerank_tokens(self, query: qmod.QueryStruct, parts: QueryParts) -> None: diff --git a/src/nominatim_api/search/query.py b/src/nominatim_api/search/query.py index 04b7f1b8..53482df8 100644 --- a/src/nominatim_api/search/query.py +++ b/src/nominatim_api/search/query.py @@ -101,7 +101,6 @@ class Token(ABC): count: int addr_count: int lookup_word: str - is_indexed: bool @abstractmethod diff --git a/test/python/api/search/test_api_search_query.py b/test/python/api/search/test_api_search_query.py index 7154ae08..71caf5b7 100644 --- a/test/python/api/search/test_api_search_query.py +++ b/test/python/api/search/test_api_search_query.py @@ -19,7 +19,7 @@ class MyToken(query.Token): def mktoken(tid: int): return MyToken(penalty=3.0, token=tid, count=1, addr_count=1, - lookup_word='foo', is_indexed=True) + lookup_word='foo') @pytest.mark.parametrize('ptype,ttype', [('NONE', 'WORD'), diff --git a/test/python/api/search/test_db_search_builder.py b/test/python/api/search/test_db_search_builder.py index 5d984014..371a6f02 100644 --- a/test/python/api/search/test_db_search_builder.py +++ b/test/python/api/search/test_db_search_builder.py @@ -33,7 +33,7 @@ def make_query(*args): q.add_token(TokenRange(start, end), ttype, MyToken(penalty=0.5 if ttype == TokenType.PARTIAL else 0.0, token=tid, count=1, addr_count=1, - lookup_word=word, is_indexed=True)) + lookup_word=word)) return q @@ -397,14 +397,14 @@ def make_counted_searches(name_part, name_full, address_part, address_full, q.add_node(BreakType.END, PhraseType.NONE) q.add_token(TokenRange(0, 1), TokenType.PARTIAL, - MyToken(0.5, 1, name_part, 1, 'name_part', True)) + MyToken(0.5, 1, name_part, 1, 'name_part')) q.add_token(TokenRange(0, 1), TokenType.WORD, - MyToken(0, 101, name_full, 1, 'name_full', True)) + MyToken(0, 101, name_full, 1, 'name_full')) for i in range(num_address_parts): q.add_token(TokenRange(i + 1, i + 2), TokenType.PARTIAL, - MyToken(0.5, 2, address_part, 1, 'address_part', True)) + MyToken(0.5, 2, address_part, 1, 'address_part')) q.add_token(TokenRange(i + 1, i + 2), TokenType.WORD, - MyToken(0, 102, address_full, 1, 'address_full', True)) + MyToken(0, 102, address_full, 1, 'address_full')) builder = SearchBuilder(q, SearchDetails()) diff --git a/test/python/api/search/test_token_assignment.py b/test/python/api/search/test_token_assignment.py index 884d2932..0d89ed5f 100644 --- a/test/python/api/search/test_token_assignment.py +++ b/test/python/api/search/test_token_assignment.py @@ -20,7 +20,7 @@ class MyToken(Token): def make_query(*args): q = QueryStruct([Phrase(args[0][1], '')]) dummy = MyToken(penalty=3.0, token=45, count=1, addr_count=1, - lookup_word='foo', is_indexed=True) + lookup_word='foo') for btype, ptype, _ in args[1:]: q.add_node(btype, ptype)