Add spenc as a lib rather than using via an import

environment.yml

@@ -11,7 +11,6 @@ dependencies:
   - geopandas>=0.7.0
   - matplotlib
   - scikit-learn=0.22
-  - spenc
   - pytest
   - pytest-cov
   - pulp

spopt/region/spenc.py

@@ -1,46 +1,47 @@
 from ..BaseClass import BaseSpOptHeuristicSolver
-import numpy as np
-from warnings import warn
-from spenc import SPENC
+from .spenclib import SPENC
 
 
 class Spenc(BaseSpOptHeuristicSolver):
-	"""
-	Spatially encouraged spectral clustering.
-	:cite:`wolf2018`
-	"""
-	def __init__(self, gdf, w, attrs_name, n_clusters=5, random_state=None, gamma=1):
-		"""
-
-		Parameters
-		----------
-
-		gdf : geopandas.GeoDataFrame
-
-		w : libpywal.weights.W instance
-		spatial weights matrix
-
-		attrs_name : list
-		Strings for attribute names (cols of ``geopandas.GeoDataFrame``).
-
-		n_clusters : int, optional, default: 5
-		The number of clusters to form.
-
-		gamma: int, default:1
-
-		"""
-		self.gdf = gdf
-		self.w = w
-		self.attrs_name = attrs_name
-		self.n_clusters = n_clusters
-		self.gamma = gamma
-		self.random_state = random_state
-
-	def solve(self):
-		"""Solve the spenc"""
-		data = self.gdf
-		X = data[self.attrs_name].values
-		#_import_tryer("spenc", "SPENC", "spenc")
-		model = SPENC(n_clusters=self.n_clusters, random_state=self.random_state, gamma=self.gamma)
-		model.fit(X, self.w.sparse)
-		self.labels_ = model.labels_
+    """
+    Spatially encouraged spectral clustering.
+    :cite:`wolf2018`
+    """
+
+    def __init__(self, gdf, w, attrs_name, n_clusters=5, random_state=None,
+                 gamma=1):
+        """
+
+        Parameters
+        ----------
+
+        gdf : geopandas.GeoDataFrame
+
+        w : libpywal.weights.W instance
+        spatial weights matrix
+
+        attrs_name : list
+        Strings for attribute names (cols of ``geopandas.GeoDataFrame``).
+
+        n_clusters : int, optional, default: 5
+        The number of clusters to form.
+
+        gamma: int, default:1
+
+        """
+        self.gdf = gdf
+        self.w = w
+        self.attrs_name = attrs_name
+        self.n_clusters = n_clusters
+        self.gamma = gamma
+        self.random_state = random_state
+
+    def solve(self):
+        """Solve the spenc"""
+        data = self.gdf
+        X = data[self.attrs_name].values
+        model = SPENC(n_clusters=self.n_clusters,
+                      random_state=self.random_state,
+                      gamma=self.gamma)
+        model.fit(X, self.w.sparse)
+        self.labels_ = model.labels_

spopt/region/spenclib/__init__.py

@@ -0,0 +1 @@
+from .abstracts import SPENC

spopt/region/spenclib/scores.py

@@ -0,0 +1,21 @@
+import numpy as np
+
+def boundary_fraction(W, labels, X = None):
+    """
+    """
+    boundary = 0
+    for row, own_label in zip(W,labels):
+        neighbor_labels = labels[row.nonzero()[-1]]
+        boundary += (neighbor_labels != own_label).any().astype(int)
+    return boundary / W.shape[0]
+
+def boundary_score(W, labels, X = None):
+    """
+    Returns a version of boundary_fraction unbounded on the negative end using
+    the log of the fraction:
+
+    np.log(boundary_fraction(W, labels))
+
+    This is solely for testing purposes. 
+    """
+    return np.log(boundary_fraction(W, labels, X = None))

spopt/region/spenclib/utils.py

@@ -0,0 +1,32 @@
+import scipy.sparse.csgraph as csg
+import scipy.sparse as sp
+from warnings import warn as Warn
+import numpy as np
+
+def check_weights(W, X=None, transform = None):
+    if X is not None:
+        assert W.shape[0] == X.shape[0], "W does not have the same number of samples as X"
+    graph = sp.csc_matrix(W)
+    graph.eliminate_zeros()
+    components, labels = csg.connected_components(graph)
+    if components > 1:
+    	Warn('Spatial affinity matrix is disconnected, and has {} subcomponents.'
+    		 'This will certainly affect the solution output.')
+    return W
+
+def lattice(x,y):
+    """
+    Construct a lattice of unit squares of dimension (x,y)
+    """
+    from shapely.geometry import Polygon
+    import geopandas as gpd
+    x = np.arange(x)*1.0
+    y = np.arange(y)*1.0
+    pgons = []
+    for i in x:
+        for j in y:
+            ll,lr,ur,ul = (i,j), (i+1,j),\
+                          (i+1,j+1), (i,j+1)
+            #print([ll,lr,ur,ul])
+            pgons.append(Polygon([ll,lr,ur,ul]))
+    return gpd.GeoDataFrame({'geometry':pgons})