usearch 0.17.0 → 0.18.0

package/.github/workflows/prerelease.yml

@@ -42,13 +42,13 @@ jobs:
         run: pytest python/scripts/test.py
 
 
-  test_python_36:
-    name: Test Python 3.6
+  test_python_37:
+    name: Test Python 3.7
     runs-on: ${{ matrix.os }}
     strategy:
       matrix:
         os: [ubuntu-20.04]
-        python-version: ["3.6"]
+        python-version: ["3.7"]
 
     steps:
       - uses: actions/checkout@v3
@@ -62,7 +62,7 @@ jobs:
       - name: Install dependencies
         run: |
           python -m pip install --upgrade pip
-          pip install pytest numpy dataclasses
+          pip install pytest numpy
 
       - name: Build locally
         run: python -m pip install .

package/.github/workflows/release.yml

@@ -125,6 +125,8 @@ jobs:
 
       - name: Build wheels
         run: python -m cibuildwheel
+        env:
+          CIBW_SKIP: cp36-*
 
       - uses: actions/upload-artifact@v3
         with:

package/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "usearch"
-version = "0.17.0"
+version = "0.18.0"
 authors = ["Ashot Vardanian <1983160+ashvardanian@users.noreply.github.com>"]
 description = "Smaller & Faster Single-File Vector Search Engine from Unum"
 edition = "2021"

package/conanfile.py

@@ -4,7 +4,7 @@ from conans import ConanFile
 class USearchConan(ConanFile):
 
     name = 'USearch'
-    version = '0.17.0'
+    version = '0.18.0'
     license = 'Apache 2.0'
     url = 'https://github.com/unum-cloud/usearch'
     description = 'Smaller & Faster Single-File Vector Search Engine from Unum'

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "usearch",
-    "version": "0.17.0",
+    "version": "0.18.0",
     "description": "Smaller & Faster Single-File Vector Search Engine from Unum",
     "author": "Ashot Vardanian",
     "license": "Apache 2.0",

package/python/scripts/test.py

@@ -6,6 +6,7 @@ from usearch.io import load_matrix, save_matrix
 from usearch.eval import recall_members, random_vectors
 
 from usearch.index import Index, SparseIndex, MetricKind, ScalarKind, Matches
+from usearch.index import _normalize_dtype
 from usearch.index import (
     DEFAULT_CONNECTIVITY,
     DEFAULT_EXPANSION_ADD,
@@ -131,7 +132,8 @@ def test_index_batch(
     assert matches.counts.shape[0] == batch_size
     assert np.all(np.sort(index.labels) == np.sort(labels))
 
-    assert recall_members(index, exact=True) == 1
+    if _normalize_dtype(numpy_type) == _normalize_dtype(index_type):
+        assert recall_members(index, exact=True) == 1
 
 
 @pytest.mark.parametrize('ndim', dimensions)

package/python/usearch/eval.py

@@ -1,4 +1,7 @@
-from typing import Tuple, Any, Callable, Union, Optional, List, NamedTuple
+from __future__ import annotations
+from time import time_ns
+from typing import Tuple, Any, Callable, Union, Optional, List
+from dataclasses import dataclass, asdict
 from collections import defaultdict
 
 import numpy as np
@@ -7,17 +10,6 @@ from usearch.io import load_matrix
 from usearch.index import Index, Matches, MetricKind, MetricKindBitwise
 
 
-# For compatibility with Python 3.6
-try:
-    from time import time_ns
-except ImportError:
-    from datetime import datetime
-
-    def time_ns():
-        now = datetime.now()
-        return int(now.timestamp() * 1e9)
-
-
 def random_vectors(
     count: int,
     metric: MetricKind = MetricKind.IP,
@@ -89,7 +81,8 @@ def measure_seconds(f: Callable) -> Tuple[float, Any]:
     return secs, result
 
 
-class Dataset(NamedTuple):
+@dataclass
+class Dataset:
 
     labels: np.ndarray
     vectors: np.ndarray
@@ -163,7 +156,8 @@ class Dataset(NamedTuple):
         return d
 
 
-class TaskResult(NamedTuple):
+@dataclass
+class TaskResult:
 
     add_operations: Optional[int] = None
     add_per_second: Optional[float] = None
@@ -190,7 +184,7 @@ class TaskResult(NamedTuple):
     def search_seconds(self) -> float:
         return self.search_operations / self.search_per_second
 
-    def __add__(self, other):
+    def __add__(self, other: TaskResult):
         result = TaskResult()
         if self.add_operations and other.add_operations:
             result.add_operations = self.add_operations + other.add_operations
@@ -217,7 +211,8 @@ class TaskResult(NamedTuple):
         return result
 
 
-class AddTask(NamedTuple):
+@dataclass
+class AddTask:
 
     labels: np.ndarray
     vectors: np.ndarray
@@ -251,7 +246,7 @@ class AddTask(NamedTuple):
         self.labels = self.labels[new_order]
         self.vectors = self.vectors[new_order, :]
 
-    def slices(self, batch_size: int) -> list:
+    def slices(self, batch_size: int) -> List[AddTask]:
         """Splits this dataset into smaller chunks."""
 
         return [
@@ -260,7 +255,7 @@ class AddTask(NamedTuple):
                 vectors=self.vectors[start_row:start_row+batch_size, :],
             ) for start_row in range(0, self.count, batch_size)]
 
-    def clusters(self, number_of_clusters: int) -> list:
+    def clusters(self, number_of_clusters: int) -> List[AddTask]:
         """Splits this dataset into smaller chunks."""
 
         from sklearn.cluster import KMeans
@@ -280,7 +275,8 @@ class AddTask(NamedTuple):
             ) for rows in partitioning.values()]
 
 
-class SearchTask(NamedTuple):
+@dataclass
+class SearchTask:
 
     queries: np.ndarray
     neighbors: np.ndarray
@@ -295,7 +291,7 @@ class SearchTask(NamedTuple):
             recall_at_one=results.recall_first(self.neighbors[:, 0].flatten()),
         )
 
-    def slices(self, batch_size: int) -> list:
+    def slices(self, batch_size: int) -> List[SearchTask]:
         """Splits this dataset into smaller chunks."""
 
         return [
@@ -305,7 +301,8 @@ class SearchTask(NamedTuple):
             ) for start_row in range(0, self.queries.shape[0], batch_size)]
 
 
-class Evaluation(NamedTuple):
+@dataclass
+class Evaluation:
 
     tasks: List[Union[AddTask, SearchTask]]
     count: int
@@ -315,7 +312,7 @@ class Evaluation(NamedTuple):
     def for_dataset(
             dataset: Dataset,
             batch_size: int = 0,
-            clusters: int = 1):
+            clusters: int = 1) -> Evaluation:
 
         tasks = []
         add = AddTask(
@@ -356,5 +353,5 @@ class Evaluation(NamedTuple):
             index.clear()
         return {
             **index.specs,
-            **task_result._asdict(),
+            **asdict(task_result),
         }