Improve search algorithm testing

Author: lucasimi

src/tdamapper/search.py

@@ -120,8 +120,6 @@ def search(self, x: PointLike) -> List[int]:
         :param x: A query point for which we want to find neighbors.
         :return: The indices of the neighbors contained in the dataset.
         """
-        if self._vptree is None:
-            return []
         neighs = self._vptree.ball_search(
             (-1, x),
             self._radius,
@@ -219,8 +217,6 @@ def search(self, x: PointLike) -> List[int]:
         :param x: A query point for which we want to find neighbors.
         :return: The indices of the neighbors contained in the dataset.
         """
-        if self._vptree is None:
-            return []
         neighs = self._vptree.knn_search((-1, x), self._neighbors)
         return [x for (x, _) in neighs]
 
@@ -306,7 +302,7 @@ def _gamma_n_inv(self, x):
     def _get_bounds(
         self, arr: NDArray[np.float64]
     ) -> Optional[Tuple[NDArray[np.float64], NDArray[np.float64], NDArray[np.float64]]]:
-        if (arr is None) or len(arr) == 0:
+        if len(arr) == 0:
             return None
         _min, _max = arr[0], arr[0]
         eps = np.finfo(np.float64).eps
@@ -330,7 +326,9 @@ def fit(self, arr: ArrayLike) -> CubicalSearch:
             raise ValueError("The parameter overlap_frac is expected to be > 0.0")
         if self.overlap_frac is not None and self.overlap_frac > 0.5:
             warn_user("The parameter overlap_frac is expected to be <= 0.5")
-        arr_ = np.asarray(arr).reshape(len(arr), -1).astype(float)
+        arr_ = np.asarray(arr)
+        if len(arr) > 0:
+            arr_ = np.asarray(arr).reshape(len(arr), -1).astype(float)
         dim = 1 if arr_.ndim == 1 else arr_.shape[1]
         self._n_intervals = self.n_intervals
         self._overlap_frac = (
@@ -339,9 +337,12 @@ def fit(self, arr: ArrayLike) -> CubicalSearch:
             else self._get_overlap_frac(dim, 0.5)
         )
         bounds = self._get_bounds(arr_)
-        if bounds is None:
-            raise ValueError("The dataset is empty or not properly defined.")
-        self._min, self._max, self._delta = bounds
+        if bounds is not None:
+            self._min, self._max, self._delta = bounds
+        else:
+            self._min = np.zeros(dim, dtype=np.float64)
+            self._max = np.ones(dim, dtype=np.float64)
+            self._delta = np.ones(dim, dtype=np.float64)
         radius = 1.0 / (2.0 - 2.0 * self._overlap_frac)
         self._ball_search = BallSearch(
             radius,

src/tdamapper/vptree_flat/ball_search.py

@@ -55,6 +55,8 @@ def search(self) -> List[T]:
         :return: A list of points that are within the specified distance from
             the given point.
         """
+        if self._arr.size() == 0:
+            return []
         return self._search_iter()
 
     def _inside(self, dist: float) -> bool:

src/tdamapper/vptree_flat/builder.py

@@ -102,6 +102,8 @@ def build(self) -> VPArray[T]:
 
         :return: The VPArray instance containing the constructed VP-tree.
         """
+        if self._arr.size() == 0:
+            return self._arr
         self._build_iter()
         return self._arr
 

src/tdamapper/vptree_flat/common.py

@@ -9,15 +9,7 @@
 
 from __future__ import annotations
 
-from typing import (
-    Callable,
-    Generic,
-    Iterable,
-    List,
-    Optional,
-    Protocol,
-    TypeVar,
-)
+from typing import Callable, Generic, Iterable, List, Optional, Protocol, TypeVar
 
 import numpy as np
 from numpy.typing import NDArray

src/tdamapper/vptree_flat/knn_search.py

@@ -59,8 +59,9 @@ def search(self) -> List[T]:
         :return: A list of points that are the k-nearest neighbors of the
             given point.
         """
-        self._search_iter()
-        return self._get_items()
+        if self._arr.size() == 0:
+            return []
+        return self._search_iter()
 
     def _process(self, x: T) -> float:
         dist = self._distance(self._point, x)

tests/test_unit_search.py

@@ -3,7 +3,15 @@
 from tdamapper.search import BallSearch, CubicalSearch, KNNSearch
 
 
-def test_ball_search():
+def test_ball_search_empty():
+    search = BallSearch(radius=1.0)
+    data = np.array([])
+    search.fit(data)
+
+    assert 0 == len(search.search(np.array([0.0])))
+
+
+def test_ball_search_ok():
     search = BallSearch(radius=1.0)
     data = np.array([[0.0], [1.0], [2.0]])
     search.fit(data)
@@ -17,7 +25,15 @@ def test_ball_search():
     assert 1 in search.search(np.array([0.5]))
 
 
-def test_knn_search():
+def test_knn_search_empty():
+    search = KNNSearch(neighbors=2)
+    data = np.array([])
+    search.fit(data)
+
+    assert 0 == len(search.search(np.array([0.0])))
+
+
+def test_knn_search_ok():
     search = KNNSearch(neighbors=2)
     data = np.array([[0.0], [1.0], [2.0], [3.0]])
     search.fit(data)
@@ -31,7 +47,15 @@ def test_knn_search():
     assert 1 in search.search(np.array([0.5]))
 
 
-def test_cubical_search():
+def test_cubical_search_empty():
+    search = CubicalSearch(n_intervals=4, overlap_frac=0.25)
+    data = np.array([])
+    search.fit(data)
+
+    assert 0 == len(search.search(np.array([1.0])))
+
+
+def test_cubical_search_ok():
     search = CubicalSearch(n_intervals=4, overlap_frac=0.25)
     data = np.array([[0.0], [1.0], [2.0], [3.0], [4.0]])
     search.fit(data)