feat: Graph.Nearest_Neighbor_Graph()

src/_igraph/convert.c

@@ -4069,6 +4069,21 @@ int igraphmodule_PyObject_to_pagerank_algo_t(PyObject *o, igraph_pagerank_algo_t
   TRANSLATE_ENUM_WITH(pagerank_algo_tt);
 }
 
+/**
+ * \ingroup python_interface_conversion
+ * \brief Converts a Python object to an igraph \c igraph_metric_t
+ */
+int igraphmodule_PyObject_to_metric_t(PyObject *o, igraph_metric_t *result) {
+  static igraphmodule_enum_translation_table_entry_t metric_tt[] = {
+        {"euclidean", IGRAPH_METRIC_EUCLIDEAN},
+        {"l2", IGRAPH_METRIC_L2}, /* alias to the previous */
+        {"manhattan", IGRAPH_METRIC_MANHATTAN},
+        {"l1", IGRAPH_METRIC_L1}, /* alias to the previous */
+        {0,0}
+    };
+  TRANSLATE_ENUM_WITH(metric_tt);
+}
+
 /**
  * \ingroup python_interface_conversion
  * \brief Converts a Python object to an igraph \c igraph_edge_type_sw_t

src/_igraph/convert.h

@@ -78,6 +78,7 @@ int igraphmodule_PyObject_to_laplacian_normalization_t(PyObject *o, igraph_lapla
 int igraphmodule_PyObject_to_layout_grid_t(PyObject *o, igraph_layout_grid_t *result);
 int igraphmodule_PyObject_to_lpa_variant_t(PyObject *o, igraph_lpa_variant_t *result);
 int igraphmodule_PyObject_to_loops_t(PyObject *o, igraph_loops_t *result);
+int igraphmodule_PyObject_to_metric_t(PyObject *o, igraph_metric_t *result);
 int igraphmodule_PyObject_to_mst_algorithm_t(PyObject *o, igraph_mst_algorithm_t *result);
 int igraphmodule_PyObject_to_neimode_t(PyObject *o, igraph_neimode_t *result);
 int igraphmodule_PyObject_to_pagerank_algo_t(PyObject *o, igraph_pagerank_algo_t *result);

src/_igraph/graphobject.c

@@ -14034,6 +14034,46 @@ PyObject *igraphmodule_Graph_random_walk(igraphmodule_GraphObject * self,
   }
 }
 
+/**********************************************************************
+ * Spatial graphs                                                     *
+ **********************************************************************/
+
+PyObject *igraphmodule_Graph_Nearest_Neighbor_Graph(PyTypeObject *type,
+                                                    PyObject *args, PyObject *kwds) {
+  static char *kwlist[] = {"points", "k", "r", "metric", "directed", NULL};
+  PyObject *points_o = Py_None, *metric_o = Py_None, *directed_o = Py_False;
+  double r = -1;
+  Py_ssize_t k = 1;
+  igraph_matrix_t points;
+  igraphmodule_GraphObject *self;
+  igraph_t graph;
+  igraph_metric_t metric = IGRAPH_METRIC_EUCLIDEAN;
+
+  if (!PyArg_ParseTupleAndKeywords(args, kwds, "O|ndOO", kwlist,
+                                   &points_o, &k, &r, &metric_o, &directed_o)) {
+    return NULL;
+  }
+
+  if (igraphmodule_PyObject_to_metric_t(metric_o, &metric)) {
+    return NULL;
+  }
+
+  if (igraphmodule_PyObject_to_matrix_t(points_o, &points, "points")) {
+    return NULL;
+  }
+
+  if (igraph_nearest_neighbor_graph(&graph, &points, metric, k, r, PyObject_IsTrue(directed_o))) {
+    igraph_matrix_destroy(&points);
+    return igraphmodule_handle_igraph_error(); 
+  }
+
+  igraph_matrix_destroy(&points);
+
+  CREATE_GRAPH_FROM_TYPE(self, graph, type);
+
+  return (PyObject *) self;
+}
+
 /**********************************************************************
  * Special internal methods that you won't need to mess around with   *
  **********************************************************************/
@@ -18894,6 +18934,22 @@ struct PyMethodDef igraphmodule_Graph_methods[] = {
    "  the given length (shorter if the random walk got stuck).\n"
   },
 
+  /**********************/
+  /* SPATIAL GRAPHS     */
+  /**********************/
+  {"Nearest_Neighbor_Graph", (PyCFunction)igraphmodule_Graph_Nearest_Neighbor_Graph,
+   METH_VARARGS | METH_CLASS | METH_KEYWORDS,
+   "Nearest_Neighbor_Graph(points, k=1, r=-1, metric=\"euclidean\", directed=False)\n--\n\n"
+   "Constructs a k nearest neighbor graph of a give point set. Each point is\n"
+   "connected to at most k spatial neighbors within a radius of 1.\n\n"
+   "@param points: coordinates of the points to use, in an arbitrary number of dimensions\n"
+   "@param k: at most how many neighbors to connect to. Pass a negative value to ignore\n"
+   "@param r: only neighbors within this radius are considered. Pass a negative value to ignore\n"
+   "@param metric: the metric to use. C{\"euclidean\"} and C{\"manhattan\"} are supported.\n"
+   "@param directed: whethe to create directed edges.\n"
+   "@return: the nearest neighbor graph.\n"
+  },
+
   /**********************/
   /* INTERNAL FUNCTIONS */
   /**********************/

tests/test_generators.py

@@ -890,6 +890,27 @@ def testDataFrame(self):
             edges = pd.DataFrame(np.array([[0, 1], [1, np.nan], [1, 2]]), dtype="Int64")
             Graph.DataFrame(edges)
 
+    def testNearestNeighborGraph(self):
+        points = [[0,0], [1,2], [-3, -3]]
+
+        g = Graph.Nearest_Neighbor_Graph(points)
+        # expecting 1 - 2, 3 - 1
+        self.assertFalse(g.is_directed())
+        self.assertEqual(g.vcount(), 3)
+        self.assertEqual(g.ecount(), 2)
+
+        g = Graph.Nearest_Neighbor_Graph(points, directed=True)
+        # expecting 1 <-> 2, 3 -> 1
+        self.assertTrue(g.is_directed())
+        self.assertEqual(g.vcount(), 3)
+        self.assertEqual(g.ecount(), 3)
+
+        # expecting a complete graph
+        g = Graph.Nearest_Neighbor_Graph(points, k=2)
+        self.assertFalse(g.is_directed())
+        self.assertEqual(g.vcount(), 3)
+        self.assertTrue(g.is_complete())
+
 
 def suite():
     generator_suite = unittest.defaultTestLoader.loadTestsFromTestCase(GeneratorTests)