Skip to content

armgit / wger

Public Imported from GitHub
0

Self hosted FLOSS fitness/workout, nutrition and weight tracker

Refine ingredient search ranking

Mikael Siidorow15fbcc24f482
diff --git a/wger/nutrition/api/filtersets.py b/wger/nutrition/api/filtersets.py
index 18599f729..34d41fbe1 100644
--- a/wger/nutrition/api/filtersets.py
+++ b/wger/nutrition/api/filtersets.py
@@ -17,7 +17,10 @@
import logging
# Django
-from django.contrib.postgres.search import TrigramSimilarity
+from django.contrib.postgres.search import (
+ TrigramSimilarity,
+ TrigramStrictWordSimilarity,
+)
from django.db.models import (
Case,
F,
@@ -39,6 +42,7 @@ from wger.nutrition.models import (
from wger.utils.db import (
PostgresILikeContains,
PostgresILikeExact,
+ PostgresILikeStartsWith,
is_postgres_db,
)
from wger.utils.language import load_language
@@ -113,27 +117,39 @@ class IngredientFilterSet(filters.FilterSet):
return barcode_qs
if is_postgres_db():
+ if not any(character.isalnum() for character in value):
+ return queryset.none()
+
# A trigram index cannot accelerate unrestricted one- or two-character
# substring patterns. Exact matching still supports valid short names.
if len(value) < 3:
exact = PostgresILikeExact(F('name'), value)
return queryset.filter(exact).order_by('name')
+ exact = PostgresILikeExact(F('name'), value)
+ starts_with = PostgresILikeStartsWith(F('name'), value)
+ contains = PostgresILikeContains(F('name'), value)
+
candidates = Q(name__trigram_similar=value)
- if _has_literal_trigram(value):
- candidates |= PostgresILikeContains(F('name'), value)
+ # Whole-name similarity already retrieves specific multi-word queries;
+ # the extra substring scan is useful for single terms in long names.
+ if len(value.split(maxsplit=1)) == 1 and _has_literal_trigram(value):
+ candidates |= contains
return (
queryset.filter(candidates)
+ .annotate(
+ word_similarity=TrigramStrictWordSimilarity(value, 'name'),
+ similarity=TrigramSimilarity('name', value),
+ )
.annotate(
match_rank=Case(
- When(name__iexact=value, then=Value(3)),
- When(name__istartswith=value, then=Value(2)),
- When(name__icontains=value, then=Value(1)),
+ When(exact, then=Value(3)),
+ When(starts_with, then=Value(2)),
+ When(word_similarity=1, then=Value(1)),
default=Value(0),
output_field=IntegerField(),
),
- similarity=TrigramSimilarity('name', value),
)
.order_by('-match_rank', '-similarity', 'name')
)
diff --git a/wger/nutrition/tests/test_search_api.py b/wger/nutrition/tests/test_search_api.py
index 94e6c7bd0..f32e7b44c 100644
--- a/wger/nutrition/tests/test_search_api.py
+++ b/wger/nutrition/tests/test_search_api.py
@@ -278,10 +278,16 @@ class IngredientSearchRankingApiTestCase(BaseTestCase, ApiBaseTestCase):
'Rice',
'Rice Pudding',
'Salty Liquorice',
+ 'Brown rice, long grain, cooked without salt',
+ 'Licorice',
'Banana',
'Banana Bread',
'Salmon',
'Smoked Salmon',
+ 'İstanbul',
+ 'Istanbul spice',
+ 'Butter',
+ 'Apple Buttrr Spread',
'餅',
'牛乳',
)
@@ -320,6 +326,7 @@ class IngredientSearchRankingApiTestCase(BaseTestCase, ApiBaseTestCase):
def test_search_finds_words_in_long_ingredient_names(self):
cases = (
('chicken', 'Chicken, Boiled Without Salt'),
+ ('chicken boiled', 'Chicken, Boiled Without Salt'),
('chicken', "'Nduja Chicken, Mozzarella, Slow Roast Tomato Sandwich"),
('lunch', 'Lunch out (placeholder)'),
('pasta', 'Pasta Carbonara, Bacon'),
@@ -364,10 +371,8 @@ class IngredientSearchRankingApiTestCase(BaseTestCase, ApiBaseTestCase):
self.assertEqual(self.search_names('餅'), ['餅'])
self.assertEqual(self.search_names('牛乳'), ['牛乳'])
- def test_short_search_escapes_like_metacharacters(self):
- for query in ('%', '_', '\\'):
- with self.subTest(query=query):
- self.assertEqual(self.search_names(query), [])
+ def test_punctuation_only_search_returns_no_results(self):
+ self.assertEqual(self.search_names('%'), [])
def test_search_finds_a_correctly_spelled_name_from_a_typo(self):
names = self.search_names('chickn')
@@ -382,5 +387,18 @@ class IngredientSearchRankingApiTestCase(BaseTestCase, ApiBaseTestCase):
def test_search_ranks_whole_words_above_inside_word_matches(self):
names = self.search_names('rice')
- self.assert_ranked_before(names, 'Rice', 'Salty Liquorice')
- self.assert_ranked_before(names, 'Rice Pudding', 'Salty Liquorice')
+ self.assert_ranked_before(
+ names,
+ 'Brown rice, long grain, cooked without salt',
+ 'Licorice',
+ )
+
+ def test_search_ranks_unicode_case_insensitive_exact_matches_first(self):
+ names = self.search_names('istanbul')
+
+ self.assertEqual(names[:2], ['İstanbul', 'Istanbul spice'])
+
+ def test_search_does_not_rank_generic_substrings_above_fuzzy_matches(self):
+ names = self.search_names('buttr')
+
+ self.assert_ranked_before(names, 'Butter', 'Apple Buttrr Spread')
diff --git a/wger/utils/db.py b/wger/utils/db.py
index 8fea834d8..d9d010230 100644
--- a/wger/utils/db.py
+++ b/wger/utils/db.py
@@ -21,6 +21,7 @@ from django.db import models
from django.db.models.lookups import (
Contains,
Exact,
+ StartsWith,
)
# wger
@@ -51,6 +52,13 @@ class PostgresILikeContains(Contains):
return f'ILIKE {rhs}'
+class PostgresILikeStartsWith(StartsWith):
+ """Case-insensitive prefix matching using PostgreSQL's native ILIKE operator."""
+
+ def get_rhs_op(self, connection, rhs):
+ return f'ILIKE {rhs}'
+
+
def is_postgres_db():
return 'postgres' in settings.DATABASES['default']['ENGINE']