usearch 2.8.7 → 2.8.9

package/python/scripts/test_tooling.py

@@ -1,133 +0,0 @@
-import os
-
-import pytest
-import numpy as np
-
-from usearch.io import load_matrix, save_matrix
-from usearch.index import search
-from usearch.eval import random_vectors
-
-from usearch.index import Match, Matches, BatchMatches, Index, Indexes
-
-
-dimensions = [3, 97, 256]
-batch_sizes = [1, 77, 100]
-
-
-@pytest.mark.parametrize("rows", batch_sizes)
-@pytest.mark.parametrize("cols", dimensions)
-def test_serializing_fbin_matrix(rows: int, cols: int):
-    """
-    Test the serialization of floating point binary matrix.
-
-    :param int rows: The number of rows in the matrix.
-    :param int cols: The number of columns in the matrix.
-    """
-    original = np.random.rand(rows, cols).astype(np.float32)
-    save_matrix(original, "tmp.fbin")
-    reconstructed = load_matrix("tmp.fbin")
-    assert np.allclose(original, reconstructed)
-    os.remove("tmp.fbin")
-
-
-@pytest.mark.parametrize("rows", batch_sizes)
-@pytest.mark.parametrize("cols", dimensions)
-def test_serializing_ibin_matrix(rows: int, cols: int):
-    """
-    Test the serialization of integer binary matrix.
-
-    :param int rows: The number of rows in the matrix.
-    :param int cols: The number of columns in the matrix.
-    """
-    original = np.random.randint(0, rows + 1, size=(rows, cols)).astype(np.int32)
-    save_matrix(original, "tmp.ibin")
-    reconstructed = load_matrix("tmp.ibin")
-    assert np.allclose(original, reconstructed)
-    os.remove("tmp.ibin")
-
-
-@pytest.mark.parametrize("rows", batch_sizes)
-@pytest.mark.parametrize("cols", dimensions)
-@pytest.mark.parametrize("k", [1, 5])
-@pytest.mark.parametrize("reordered", [False, True])
-def test_exact_search(rows: int, cols: int, k: int, reordered: bool):
-    """
-    Test exact search.
-
-    :param int rows: The number of rows in the matrix.
-    :param int cols: The number of columns in the matrix.
-    """
-    original = np.random.rand(rows, cols)
-    keys = np.arange(rows)
-    k = min(k, rows)
-
-    if reordered:
-        reordered_keys = np.arange(rows)
-        np.random.shuffle(reordered_keys)
-    else:
-        reordered_keys = keys
-
-    matches: BatchMatches = search(original, original[reordered_keys], k, exact=True)
-    top_matches = (
-        [int(m.keys[0]) for m in matches] if rows > 1 else [int(matches.keys[0])]
-    )
-    assert top_matches == list(reordered_keys)
-
-    matches: Matches = search(original, original[-1], k, exact=True)
-    top_match = int(matches.keys[0])
-    assert top_match == keys[-1]
-
-
-def test_matches_creation_and_methods():
-    matches = Matches(
-        keys=np.array([1, 2]),
-        distances=np.array([0.5, 0.6]),
-        visited_members=2,
-        computed_distances=2,
-    )
-    assert len(matches) == 2
-    assert matches[0] == Match(key=1, distance=0.5)
-    assert matches.to_list() == [(1, 0.5), (2, 0.6)]
-
-
-def test_batch_matches_creation_and_methods():
-    keys = np.array([[1, 2], [3, 4]])
-    distances = np.array([[0.5, 0.6], [0.7, 0.8]])
-    counts = np.array([2, 2])
-    batch_matches = BatchMatches(
-        keys=keys,
-        distances=distances,
-        counts=counts,
-        visited_members=2,
-        computed_distances=2,
-    )
-
-    assert len(batch_matches) == 2
-    assert batch_matches[0].keys.tolist() == [1, 2]
-    assert batch_matches[0].distances.tolist() == [0.5, 0.6]
-    assert batch_matches.to_list() == [(1, 0.5), (2, 0.6), (3, 0.7), (4, 0.8)]
-
-
-def test_multi_index():
-    ndim = 10
-    index_a = Index(ndim=ndim)
-    index_b = Index(ndim=ndim)
-
-    vectors = random_vectors(count=3, ndim=ndim)
-    index_a.add(42, vectors[0])
-    index_b.add(43, vectors[1])
-
-    indexes = Indexes([index_a, index_b])
-
-    # Search top 10 for 1
-    matches = indexes.search(vectors[2], 10)
-    assert len(matches) == 2
-
-    # Search top 1 for 1
-    matches = indexes.search(vectors[2], 1)
-    assert len(matches) == 1
-
-    # Search top 10 for 3
-    matches = indexes.search(vectors, 10)
-    assert len(matches) == 3
-    assert len(matches[0].keys) == 2

package/python/usearch/client.py

@@ -1,124 +0,0 @@
-from typing import Union, Optional, List
-
-import numpy as np
-from ucall.client import Client
-
-from usearch.index import Matches
-
-
-def _vector_to_ascii(vector: np.ndarray) -> Optional[str]:
-    if vector.dtype != np.int8 and vector.dtype != np.uint8 and vector.dtype != np.byte:
-        return None
-    if not np.all((vector >= 0) | (vector <= 100)):
-        return None
-
-    # Let's map [0, 100] to the range from [23, 123],
-    # poking 60 and replacing with the 124.
-    vector += 23
-    vector[vector == 60] = 124
-    ascii = str(vector)
-    return ascii
-
-
-class IndexClient:
-    def __init__(
-        self, uri: str = "127.0.0.1", port: int = 8545, use_http: bool = True
-    ) -> None:
-        self.client = Client(uri=uri, port=port, use_http=use_http)
-
-    def add_one(self, key: int, vector: np.ndarray):
-        assert isinstance(key, int)
-        assert isinstance(vector, np.ndarray)
-        vector = vector.flatten()
-        ascii = _vector_to_ascii(vector)
-        if ascii:
-            self.client.add_ascii(key=key, string=ascii)
-        else:
-            self.client.add_one(key=key, vectors=vector)
-
-    def add_many(self, keys: np.ndarray, vectors: np.ndarray):
-        assert isinstance(keys, int)
-        assert isinstance(vectors, np.ndarray)
-        assert keys.ndim == 1 and vectors.ndim == 2
-        assert keys.shape[0] == vectors.shape[0]
-        self.client.add_many(keys=keys, vectors=vectors)
-
-    def add(self, keys: Union[np.ndarray, int], vectors: np.ndarray):
-        if isinstance(keys, int) or len(keys) == 1:
-            return self.add_one(keys, vectors)
-        else:
-            return self.add_many(keys, vectors)
-
-    def search_one(self, vector: np.ndarray, count: int) -> Matches:
-        matches: List[dict] = []
-        vector = vector.flatten()
-        ascii = _vector_to_ascii(vector)
-        if ascii:
-            matches = self.client.search_ascii(string=ascii, count=count)
-        else:
-            matches = self.client.search_one(vector=vector, count=count)
-
-        print(matches.data)
-        matches = matches.json
-
-        keys = np.array((1, count), dtype=np.uint32)
-        distances = np.array((1, count), dtype=np.float32)
-        counts = np.array((1), dtype=np.uint32)
-        for col, result in enumerate(matches):
-            keys[0, col] = result["key"]
-            distances[0, col] = result["distance"]
-        counts[0] = len(matches)
-
-        return keys, distances, counts
-
-    def search_many(self, vectors: np.ndarray, count: int) -> Matches:
-        batch_size: int = vectors.shape[0]
-        list_of_matches: List[List[dict]] = self.client.search_many(
-            vectors=vectors, count=count
-        )
-
-        keys = np.array((batch_size, count), dtype=np.uint32)
-        distances = np.array((batch_size, count), dtype=np.float32)
-        counts = np.array((batch_size), dtype=np.uint32)
-        for row, matches in enumerate(list_of_matches):
-            for col, result in enumerate(matches):
-                keys[row, col] = result["key"]
-                distances[row, col] = result["distance"]
-            counts[row] = len(results)
-
-        return keys, distances, counts
-
-    def search(self, vectors: np.ndarray, count: int) -> Matches:
-        if vectors.ndim == 1 or (vectors.ndim == 2 and vectors.shape[0] == 1):
-            return self.search_one(vectors, count)
-        else:
-            return self.search_many(vectors, count)
-
-    def __len__(self):
-        return self.client.size().json()
-
-    @property
-    def ndim(self):
-        return self.client.ndim().json()
-
-    def capacity(self):
-        return self.client.capacity().json()
-
-    def connectivity(self):
-        return self.client.connectivity().json()
-
-    def load(self, path: str):
-        raise NotImplementedError()
-
-    def view(self, path: str):
-        raise NotImplementedError()
-
-    def save(self, path: str):
-        raise NotImplementedError()
-
-
-if __name__ == "__main__":
-    index = IndexClient()
-    index.add(42, np.array([0.4] * 256, dtype=np.float32))
-    results = index.search(np.array([0.4] * 256, dtype=np.float32), 10)
-    print(results)