Correctly parse LOWER_ROW edge weight format (#135)

Author: leonlan

tests/parse/test_parse_distances.py

@@ -2,12 +2,7 @@
 import pytest
 from numpy.testing import assert_almost_equal, assert_equal, assert_raises
 
-from vrplib.parse.parse_distances import (
-    from_eilon,
-    from_lower_row,
-    is_triangular_number,
-    parse_distances,
-)
+from vrplib.parse.parse_distances import parse_distances
 
 
 @pytest.mark.parametrize(
@@ -68,33 +63,24 @@ def test_parse_euclidean_distances(edge_weight_type, desired):
 
 
 @pytest.mark.parametrize(
-    "comment, func", [("Eilon", from_eilon), (None, from_lower_row)]
+    "data",
+    [
+        [[1, 2, 3, 4, 5, 6]],  # single line
+        [[1, 2, 3, 4], [5, 6]],  # ragged lines
+        [[1], [2, 3], [4, 5, 6]],  # proper triangular rows
+    ],
 )
-def test_parse_lower_row(comment, func):
+def test_parse_lower_row(data):
     """
-    Tests if a ``LOWER ROW`` instance is parsed as Eilon instance or regular
-    instance. Eilon instances do not contain a proper lower row matrix, but
-    a lower column matrix instead. The current way of detecting an Eilon
-    instance is by means of the ``COMMENT`` field, which is checked for
-    including "Eilon".
+    Tests that LOWER_ROW instances are parsed correctly regardless of how
+    the values are wrapped across lines. See #134.
     """
-    instance = {
-        "data": np.array([[1], [2, 3], [4, 5, 6]], dtype=object),
-        "edge_weight_type": "EXPLICIT",
-        "edge_weight_format": "LOWER_ROW",
-        "comment": comment,
-    }
-
-    assert_equal(parse_distances(**instance), func(instance["data"]))
-
-
-def test_from_lower_row():
-    """
-    Tests that a lower row triangular matrix is correctly transformed into a
-    full matrix.
-    """
-    triangular_matrix = np.array([[1], [2, 3], [4, 5, 6]], dtype=object)
-    actual = from_lower_row(triangular_matrix)
+    data = np.array(data, dtype=object)
+    actual = parse_distances(
+        data,
+        edge_weight_type="EXPLICIT",
+        edge_weight_format="LOWER_ROW",
+    )
     desired = np.array(
         [
             [0, 1, 2, 4],
@@ -105,31 +91,3 @@ def test_from_lower_row():
     )
 
     assert_equal(actual, desired)
-
-
-def test_from_eilon():
-    """
-    Tests that the distance matrix of Eilon instances is correctly transformed.
-    These distance matrices have entries corresponding to the lower column
-    triangular matrices. But the distance matrix is not a triangular matrix,
-    so they are flattened first.
-    """
-    eilon = np.array([[1, 2, 3, 4], [5, 6]], dtype=object)
-    actual = from_eilon(eilon)
-    desired = np.array(
-        [
-            [0, 1, 2, 3],
-            [1, 0, 4, 5],
-            [2, 4, 0, 6],
-            [3, 5, 6, 0],
-        ]
-    )
-
-    assert_equal(actual, desired)
-
-
-@pytest.mark.parametrize(
-    "n, res", [(1, True), (3, True), (4, False), (630, True), (1000, False)]
-)
-def test_is_triangular_number(n, res):
-    assert_equal(is_triangular_number(n), res)

tests/read/test_read_instance.py

@@ -115,3 +115,25 @@ def test_do_not_compute_edge_weights(tmp_path):
 
     instance = read_instance(tmp_path / name, "solomon", False)
     assert_("edge_weight" not in instance)
+
+
+def test_read_explicit_lower_row_instance_objective():
+    """
+    Tests that the E-n13-k4 instance with EXPLICIT LOWER_ROW edge weights
+    is read correctly by verifying the known optimal solution cost of 247.
+    """
+    instance = read_instance("tests/data/E-n13-k4.vrp")
+    edge_weight = instance["edge_weight"]
+
+    # Known optimal solution routes (0-indexed customer IDs).
+    # Depot is node 0; customers are nodes 1-12.
+    routes = [[1], [8, 5, 3], [9, 12, 10, 6], [11, 4, 7, 2]]
+
+    total_cost = 0
+    for route in routes:
+        total_cost += edge_weight[0, route[0]]
+        for idx in range(len(route) - 1):
+            total_cost += edge_weight[route[idx], route[idx + 1]]
+        total_cost += edge_weight[route[-1], 0]
+
+    assert_equal(total_cost, 247)

vrplib/parse/parse_distances.py

@@ -1,5 +1,3 @@
-from itertools import combinations
-
 import numpy as np
 
 
@@ -61,11 +59,6 @@ def parse_distances(
 
     if edge_weight_type == "EXPLICIT":
         if edge_weight_format == "LOWER_ROW":
-            # TODO Eilon instances edge weight specifications are incorrect in
-            # (C)VRPLIB format. Find a better way to identify Eilon instances.
-            if comment is not None and "Eilon" in comment:
-                return from_eilon(data)
-
             return from_lower_row(data)
 
         if edge_weight_format == "FULL_MATRIX":
@@ -98,56 +91,32 @@ def pairwise_euclidean(coords: np.ndarray) -> np.ndarray:
     return np.sqrt(sq_dist)
 
 
-def from_lower_row(triangular: np.ndarray) -> np.ndarray:
+def from_lower_row(data: np.ndarray) -> np.ndarray:
     """
-    Computes a full distances matrix from a lower row triangular matrix.
-    The triangular matrix should not contain the diagonal.
+    Computes a full distances matrix from a LOWER_ROW edge weight section.
+
+    The input is treated as a continuous 1D stream of values (as specified
+    by TSPLIB95), regardless of how the values are wrapped across lines.
 
     Parameters
     ----------
-    triangular
-        A list of lists, each list representing the entries of a row in a
-        lower triangular matrix without diagonal entries.
+    data
+        Edge weight data, possibly as a ragged array of rows.
 
     Returns
     -------
     np.ndarray
-        A n-by-n distances matrix.
-    """
-    n = len(triangular) + 1
-    distances = np.zeros((n, n))
-
-    for i in range(n - 1):
-        distances[i + 1, : i + 1] = triangular[i]
-
-    return distances + distances.T
-
-
-def from_eilon(edge_weights: np.ndarray) -> np.ndarray:
+        An n-by-n distances matrix.
     """
-    Computes a full distances matrix from the Eilon instances with "LOWER_ROW"
-    edge weight format. The specification is incorrect, instead the edge weight
-    section needs to be parsed as a flattend, column-wise triangular matrix.
+    flattened = np.concatenate(data).astype(float)
 
-    See https://github.com/leonlan/VRPLIB/issues/40.
-    """
-    flattened = [dist for row in edge_weights for dist in row]
-    n = int((2 * len(flattened)) ** 0.5) + 1  # The (n+1)-th triangular number
+    # The flattened data represents the lower triangle of a symmetric matrix.
+    # See https://en.wikipedia.org/wiki/Triangular_number.
+    # m = n * (n - 1) / 2 => n = (1 + sqrt(1 + 8m)) / 2
+    n = (1 + int((1 + 8 * flattened.size) ** 0.5)) // 2
 
     distances = np.zeros((n, n))
-    indices = sorted([(i, j) for (i, j) in combinations(range(n), r=2)])
-
-    for idx, (i, j) in enumerate(indices):
-        d_ij = flattened[idx]
-        distances[i, j] = d_ij
-        distances[j, i] = d_ij
+    distances[np.tril_indices(n, k=-1)] = flattened
+    distances += distances.T
 
     return distances
-
-
-def is_triangular_number(n):
-    """
-    Checks if n is a triangular number.
-    """
-    i = int((2 * n) ** 0.5)
-    return i * (i + 1) == 2 * n