mirror of
https://github.com/redis/redis.git
synced 2026-06-11 09:52:05 -04:00
674b829981
Hi, this PR implements the following changes: 1. The EPSILON option of VSIM is now documented. 2. The EPSILON behavior was fixed: the score was incorrectly divided by two in the meaning, with a 0-2 interval provided by the underlying cosine similarity, instead of the 0-1 interval. So an EPSILON of 0.2 only returned elements with a distance between 1 and 0.9 instead of 1 and 0.8. This is a *breaking change* but the command was not documented so far, and it is a fix, as the user sees the similarity score so was a total mismatch. I believe this fix should definitely be back ported as soon as possible. 3. There are now tests. Thanks for checking, Salvatore
77 lines
4.7 KiB
Python
77 lines
4.7 KiB
Python
from test import TestCase
|
|
|
|
class EpsilonOption(TestCase):
|
|
def getname(self):
|
|
return "VSIM EPSILON option filtering"
|
|
|
|
def estimated_runtime(self):
|
|
return 0.1
|
|
|
|
def test(self):
|
|
# Add vectors as shown in the example
|
|
# Vector 'a' at (1, 1) - normalized to (0.707, 0.707)
|
|
result = self.redis.execute_command('VADD', self.test_key, 'VALUES', '2', '1', '1', 'a')
|
|
assert result == 1, "VADD should return 1 for item 'a'"
|
|
|
|
# Vector 'b' at (0, 1) - normalized to (0, 1)
|
|
result = self.redis.execute_command('VADD', self.test_key, 'VALUES', '2', '0', '1', 'b')
|
|
assert result == 1, "VADD should return 1 for item 'b'"
|
|
|
|
# Vector 'c' at (0, 0) - this will be a zero vector, might be handled specially
|
|
result = self.redis.execute_command('VADD', self.test_key, 'VALUES', '2', '0', '0', 'c')
|
|
assert result == 1, "VADD should return 1 for item 'c'"
|
|
|
|
# Vector 'd' at (0, -1) - normalized to (0, -1)
|
|
result = self.redis.execute_command('VADD', self.test_key, 'VALUES', '2', '0', '-1', 'd')
|
|
assert result == 1, "VADD should return 1 for item 'd'"
|
|
|
|
# Vector 'e' at (-1, -1) - normalized to (-0.707, -0.707)
|
|
result = self.redis.execute_command('VADD', self.test_key, 'VALUES', '2', '-1', '-1', 'e')
|
|
assert result == 1, "VADD should return 1 for item 'e'"
|
|
|
|
# Test without EPSILON - should return all items
|
|
result = self.redis.execute_command('VSIM', self.test_key, 'VALUES', '2', '1', '1', 'WITHSCORES')
|
|
# Result is a flat list: [elem1, score1, elem2, score2, ...]
|
|
elements_all = [result[i].decode() for i in range(0, len(result), 2)]
|
|
scores_all = [float(result[i]) for i in range(1, len(result), 2)]
|
|
|
|
assert len(elements_all) == 5, f"Should return 5 elements without EPSILON, got {len(elements_all)}"
|
|
assert elements_all[0] == 'a', "First element should be 'a' (most similar)"
|
|
assert scores_all[0] == 1.0, "Score for 'a' should be 1.0 (identical)"
|
|
|
|
# Test with EPSILON 0.5 - should return only elements with similarity >= 0.5 (distance < 0.5)
|
|
result = self.redis.execute_command('VSIM', self.test_key, 'VALUES', '2', '1', '1', 'WITHSCORES', 'EPSILON', '0.5')
|
|
elements_epsilon_0_5 = [result[i].decode() for i in range(0, len(result), 2)]
|
|
scores_epsilon_0_5 = [float(result[i]) for i in range(1, len(result), 2)]
|
|
|
|
assert len(elements_epsilon_0_5) == 3, f"With EPSILON 0.5, should return 3 elements, got {len(elements_epsilon_0_5)}"
|
|
assert set(elements_epsilon_0_5) == {'a', 'b', 'c'}, f"With EPSILON 0.5, should get a, b, c, got {elements_epsilon_0_5}"
|
|
|
|
# Verify all returned scores are >= 0.5
|
|
for i, score in enumerate(scores_epsilon_0_5):
|
|
assert score >= 0.5, f"Element {elements_epsilon_0_5[i]} has score {score} which is < 0.5"
|
|
|
|
# Test with EPSILON 0.2 - should return only elements with similarity >= 0.8 (distance < 0.2)
|
|
result = self.redis.execute_command('VSIM', self.test_key, 'VALUES', '2', '1', '1', 'WITHSCORES', 'EPSILON', '0.2')
|
|
elements_epsilon_0_2 = [result[i].decode() for i in range(0, len(result), 2)]
|
|
scores_epsilon_0_2 = [float(result[i]) for i in range(1, len(result), 2)]
|
|
|
|
assert len(elements_epsilon_0_2) == 2, f"With EPSILON 0.2, should return 2 elements, got {len(elements_epsilon_0_2)}"
|
|
assert set(elements_epsilon_0_2) == {'a', 'b'}, f"With EPSILON 0.2, should get a, b, got {elements_epsilon_0_2}"
|
|
|
|
# Verify all returned scores are >= 0.8 (since distance < 0.2 means similarity > 0.8)
|
|
for i, score in enumerate(scores_epsilon_0_2):
|
|
assert score >= 0.8, f"Element {elements_epsilon_0_2[i]} has score {score} which is < 0.8"
|
|
|
|
# Test with very small EPSILON - should return only the exact match
|
|
result = self.redis.execute_command('VSIM', self.test_key, 'VALUES', '2', '1', '1', 'WITHSCORES', 'EPSILON', '0.001')
|
|
elements_epsilon_small = [result[i].decode() for i in range(0, len(result), 2)]
|
|
|
|
assert len(elements_epsilon_small) == 1, f"With EPSILON 0.001, should return only 1 element, got {len(elements_epsilon_small)}"
|
|
assert elements_epsilon_small[0] == 'a', "With very small EPSILON, should only get 'a'"
|
|
|
|
# Test with EPSILON 1.0 - should return all elements (since all similarities are between 0 and 1)
|
|
result = self.redis.execute_command('VSIM', self.test_key, 'VALUES', '2', '1', '1', 'WITHSCORES', 'EPSILON', '1.0')
|
|
elements_epsilon_1 = [result[i].decode() for i in range(0, len(result), 2)]
|
|
|
|
assert len(elements_epsilon_1) == 5, f"With EPSILON 1.0, should return all 5 elements, got {len(elements_epsilon_1)}"
|