cost-matrix 0.1.0__py3-none-any.whl → 0.2.0__py3-none-any.whl

cost_matrix/euclidean_matrix.py

@@ -1,5 +1,28 @@
 import numpy as np
 
 
-def euclidean(sources: np.ndarray, destinations: np.ndarray) -> np.ndarray:
-    return np.linalg.norm(sources - destinations, axis=1)
+def euclidean(sources, destinations):
+    """
+    Calculate the Euclidean distance matrix between source and destination points.
+
+    Parameters
+    ----------
+    sources : np.ndarray
+        Array of shape (n, d) containing the source points.
+    destinations : np.ndarray
+        Array of shape (m, d) containing the destination points.
+
+    Returns
+    -------
+    np.ndarray
+        Euclidean distance matrix of shape (n, m).
+    """
+    # Expand dimensions to enable broadcasting
+    sources_expanded = sources[:, np.newaxis, :]
+    destinations_expanded = destinations[np.newaxis, :, :]
+
+    # Compute differences along the last dimension
+    differences = sources_expanded - destinations_expanded
+
+    # Compute Euclidean distances using norm along the last dimension
+    return np.linalg.norm(differences, axis=2)

cost_matrix/manhattan_matrix.py

@@ -2,4 +2,28 @@ import numpy as np
 
 
 def manhattan(sources: np.ndarray, destinations: np.ndarray) -> np.ndarray:
-    return np.sum(np.abs(sources - destinations), axis=1)
+    """
+    Compute the Manhattan distance matrix between source and destination points.
+
+    Parameters
+    ----------
+    sources : np.ndarray
+        Array of shape (n, d) containing the source points.
+    destinations : np.ndarray
+        Array of shape (m, d) containing the destination points.
+
+    Returns
+    -------
+    np.ndarray
+        Manhattan distance matrix of shape (n, m).
+    """
+    # Expand dimensions to enable broadcasting
+    sources_expanded = sources[:, np.newaxis, :]
+    destinations_expanded = destinations[np.newaxis, :, :]
+
+    # Compute absolute differences along the last dimension
+    differences = np.abs(sources_expanded - destinations_expanded)
+
+    # Sum absolute differences along the last dimension to get
+    # Manhattan distance
+    return np.sum(differences, axis=2)

cost_matrix/osrm_matrix.py

@@ -1,3 +1,5 @@
+import os
+from concurrent.futures import ThreadPoolExecutor
 from math import ceil
 
 import numpy as np
@@ -11,55 +13,88 @@ def osrm(
     batch_size: int = 150,
     cost_type: str = "distances",
 ) -> np.ndarray:
-    """Compute the OSRM cost matrix between sources and destinations"""
-
+    """
+    Compute the OSRM cost matrix between sources and destinations
+
+    Parameters
+    ----------
+    sources : np.ndarray
+        Array of shape (n, d) containing the source points.
+    destinations : np.ndarray
+        Array of shape (m, d) containing the destination points.
+    server_address : str
+        Address of the OSRM server. Default is "http://router.project-osrm.org".
+    batch_size : int
+        Number of points to send in each request. Default is 150.
+    cost_type : str
+        Type of cost to be computed. Default is "distances". Options are
+        "distances" and "durations".
+
+    Returns
+    -------
+    np.ndarray
+        OSRM cost matrix of shape (n, m).
+    """
     num_sources = sources.shape[0]
     num_destinations = destinations.shape[0]
-    cost_matrix = np.zeros((num_sources, num_destinations))
+    cost_matrix = np.zeros((num_sources, num_destinations), dtype=np.float32)
 
     num_batches_i = ceil(num_sources / batch_size)
     num_batches_j = ceil(num_destinations / batch_size)
 
-    for i in range(num_batches_i):
-        start_i = i * batch_size
-        end_i = min((i + 1) * batch_size, num_sources)
+    cpu_count = os.cpu_count() or 0
+    default_max = cpu_count * 2 if cpu_count > 0 else 8
+    max_workers = min(default_max, 16)
+
+    futures = []
+    positions = []
+
+    with ThreadPoolExecutor(max_workers=max_workers) as executor:
+        for i in range(num_batches_i):
+            start_i = i * batch_size
+            end_i = min((i + 1) * batch_size, num_sources)
+
+            for j in range(num_batches_j):
+                start_j = j * batch_size
+                end_j = min((j + 1) * batch_size, num_destinations)
 
-        for j in range(num_batches_j):
-            start_j = j * batch_size
-            end_j = min((j + 1) * batch_size, num_destinations)
-            sources_batch = sources[start_i:end_i]
-            destinations_batch = destinations[start_j:end_j]
+                sources_batch = sources[start_i:end_i]
+                destinations_batch = destinations[start_j:end_j]
 
-            cost_matrix[start_i:end_i, start_j:end_j] = (
-                _get_batch_osrm_distance(
+                future = executor.submit(
+                    _fetch_matrix_batch,
                     sources_batch,
                     destinations_batch,
                     server_address,
-                    cost_type=cost_type,
+                    cost_type,
                 )
-            )
+                futures.append(future)
+                positions.append((start_i, end_i, start_j, end_j))
+
+        for future, (start_i, end_i, start_j, end_j) in zip(
+            futures, positions, strict=False
+        ):
+            cost_matrix[start_i:end_i, start_j:end_j] = future.result()
 
     return cost_matrix
 
 
-def _get_batch_osrm_distance(
+def _fetch_matrix_batch(
     sources_batch: np.ndarray,
     destinations_batch: np.ndarray,
     server_address: str,
     cost_type: str,
 ):
     """Request the OSRM cost matrix for a given batch"""
-
-    url = _format_osrm_url(
+    url = _format_url(
         sources_batch, destinations_batch, server_address, cost_type
     )
-    resp = requests.get(url)
-    resp.raise_for_status()
-
-    return np.array(resp.json()[cost_type])
+    response = requests.get(url)
+    response.raise_for_status()
+    return np.array(response.json()[cost_type], dtype=np.float32)
 
 
-def _format_osrm_url(
+def _format_url(
     sources_batch: np.ndarray,
     destinations_batch: np.ndarray,
     server_address: str,
@@ -93,11 +128,8 @@ def _format_osrm_url(
     "distance"), but the returned JSON follows the plural form (e.g.,
     "distances"). Thus, we ignore the last letter of the input type
     """
-
     url_cost_type = cost_type[:-1]
-    sources_coord = ";".join(
-        f"{source[1]},{source[0]}" for source in sources_batch
-    )
+    sources_coord = ";".join(f"{lng},{lat}" for (lat, lng) in sources_batch)
 
     # If sources == destinations, return a simpler URL early. Notice it needs
     # at least two points, otherwise OSRM complains
@@ -106,14 +138,12 @@ def _format_osrm_url(
         and sources_batch.shape[0] > 1
     ):
         return (
-            f"{server_address}/table/v1/driving/"
-            f"{sources_coord}"
+            f"{server_address}/table/v1/driving/{sources_coord}"
             f"?annotations={url_cost_type}"
         )
 
     destinations_coord = ";".join(
-        f"{destination[1]},{destination[0]}"
-        for destination in destinations_batch
+        f"{lng},{lat}" for (lat, lng) in destinations_batch
     )
     locations_coord = sources_coord + ";" + destinations_coord
 
@@ -121,7 +151,6 @@ def _format_osrm_url(
     # sources = 0,1,...,N' and destinations = N'+1,N'+2...N'+M'
     num_sources = sources_batch.shape[0]
     num_destinations = destinations_batch.shape[0]
-
     sources_indices = ";".join(str(index) for index in range(num_sources))
     destinations_indices = ";".join(
         str(index)
@@ -129,8 +158,8 @@ def _format_osrm_url(
     )
 
     return (
-        f"{server_address}/table/v1/driving/"
-        f"{locations_coord}"
-        f"?sources={sources_indices}&destinations={destinations_indices}"
+        f"{server_address}/table/v1/driving/{locations_coord}"
+        f"?sources={sources_indices}"
+        f"&destinations={destinations_indices}"
         f"&annotations={url_cost_type}"
     )

cost_matrix/spherical_matrix.py

@@ -4,7 +4,21 @@ EARTH_RADIUS_METERS = 6371000
 
 
 def spherical(sources: np.ndarray, destinations: np.ndarray) -> np.ndarray:
-    """Distance matrix using the Spherical distance"""
+    """
+    Compute the distance matrix using the Spherical distance]
+
+    Parameters
+    ----------
+    sources : np.ndarray
+        Array of shape (n, d) containing the source points.
+    destinations : np.ndarray
+        Array of shape (m, d) containing the destination points.
+
+    Returns
+    -------
+    np.ndarray
+        Spherical distance matrix of shape (n, m).
+    """
 
     sources_rad = np.radians(sources)
     destinations_rad = np.radians(destinations)

{cost_matrix-0.1.0.dist-info → cost_matrix-0.2.0.dist-info}/METADATA

@@ -1,12 +1,11 @@
-Metadata-Version: 2.1
+Metadata-Version: 2.3
 Name: cost-matrix
-Version: 0.1.0
+Version: 0.2.0
 Summary: Designed to simplify the creation of cost matrices for optimization problems.
 Author: Luan
 Author-email: llvdmoraes@gmail.com
-Requires-Python: >=3.9,<3.13
+Requires-Python: >=3.10,<3.13
 Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.9
 Classifier: Programming Language :: Python :: 3.10
 Classifier: Programming Language :: Python :: 3.11
 Classifier: Programming Language :: Python :: 3.12
@@ -69,4 +68,8 @@ osrm_duration_matrix = cost_matrix.osrm(
     batch_size=250
 )
 print(osrm_duration_matrix)
+```
+
+## GitHub
+* [Repository](https://github.com/luanleonardo/cost-matrix)
 

cost_matrix-0.2.0.dist-info/RECORD

@@ -0,0 +1,8 @@
+cost_matrix/__init__.py,sha256=9EaPsXe0SIuym34-CPmzUPSGJuBd2AkQZHqjcwlUywE,202
+cost_matrix/euclidean_matrix.py,sha256=PiOJFjalkvk1p4ktcCvFD7EeXswvsnm8f-oze1gFRlk,844
+cost_matrix/manhattan_matrix.py,sha256=OgOepYRfskarZmNNarl58YqNRw7WXYnjztWV2BbjGBI,907
+cost_matrix/osrm_matrix.py,sha256=mlj61VoMKb0hisDl4xdNa_-9gw0RgtILmr4GpjIFg3k,5495
+cost_matrix/spherical_matrix.py,sha256=9_y5XCpdAEiWcvljPX7kFYSUt0nU3Gq4DOwr7LlbhQY,1271
+cost_matrix-0.2.0.dist-info/METADATA,sha256=IZyDlKzSf0C4bnMICS8CgmO6Lrb95eXlXNkK0wVC1is,2702
+cost_matrix-0.2.0.dist-info/WHEEL,sha256=b4K_helf-jlQoXBBETfwnf4B04YC67LOev0jo4fX5m8,88
+cost_matrix-0.2.0.dist-info/RECORD,,

{cost_matrix-0.1.0.dist-info → cost_matrix-0.2.0.dist-info}/WHEEL

@@ -1,4 +1,4 @@
 Wheel-Version: 1.0
-Generator: poetry-core 1.9.0
+Generator: poetry-core 2.1.3
 Root-Is-Purelib: true
 Tag: py3-none-any

cost_matrix-0.1.0.dist-info/RECORD

@@ -1,8 +0,0 @@
-cost_matrix/__init__.py,sha256=9EaPsXe0SIuym34-CPmzUPSGJuBd2AkQZHqjcwlUywE,202
-cost_matrix/euclidean_matrix.py,sha256=a8uLh0COVb5duaKV6DMfG9HklkJX2dOARKfnnWxLxsc,155
-cost_matrix/manhattan_matrix.py,sha256=bWTTiIgvb0PDWSdfCXuffHEIgb9nNMb4w7Qwhe3EbaQ,155
-cost_matrix/osrm_matrix.py,sha256=VuuoQVRGxMjoN_eUfs9yur9lwINg8JZxB3SX-6cvBFk,4327
-cost_matrix/spherical_matrix.py,sha256=SxaihjkKyrp0den5w-VsimGcjjQEi-q-_cvxeTM3xyY,946
-cost_matrix-0.1.0.dist-info/METADATA,sha256=4TVrFT0TlklV7jcvRAJarKqi-TE2_TPMtSJJtVZTmIQ,2676
-cost_matrix-0.1.0.dist-info/WHEEL,sha256=sP946D7jFCHeNz5Iq4fL4Lu-PrWrFsgfLXbbkciIZwg,88
-cost_matrix-0.1.0.dist-info/RECORD,,