fix to dbscan tolerances for euclidean cases

hexrd/findorientations.py

@@ -316,10 +316,13 @@ def quat_distance(x, y):
 
 @clustering_algorithm('dbscan')
 def cluster_dbscan(qfib_r, qsym, cl_radius, min_samples):
+    # CAVEAT: the euclidean misorientation of two quaternions  
+    # is ~2x smaller than the true quaternion misorientation
+    # for magnitudes < ~5deg
     _check_dbscan()
     dbscan = sklearn.cluster.dbscan
     pts = qfib_r.T
-    _, labels = dbscan(pts, eps=np.radians(cl_radius),
+    _, labels = dbscan(pts, eps=0.5*np.radians(cl_radius),
                        min_samples=min_samples, metric='minkowski', p=2)
     labels = _normalize_labels_from_dbscan(labels)
     qbar = _compute_centroids_dense(labels, qfib_r, qsym)
@@ -329,10 +332,13 @@ def cluster_dbscan(qfib_r, qsym, cl_radius, min_samples):
 
 @clustering_algorithm('ort-dbscan')
 def cluster_ort_dbscan(qfib_r, qsym, cl_radius, min_samples):
+    # CAVEAT: the euclidean misorientation of the vector parts of two
+    # quaternion is ~2.01x smaller than the true quaternion misorientation 
+    # for magnitudes < ~5deg
     _check_dbscan()
     dbscan = sklearn.cluster.dbscan
     pts = qfib_r[1:, :].T
-    _, labels = dbscan(pts, eps=0.9*np.radians(cl_radius),
+    _, labels = dbscan(pts, eps=0.49*np.radians(cl_radius),
                        min_samples=min_samples, metric='minkowski', p=2)
     labels = _normalize_labels_from_dbscan(labels)
     qbar = _compute_centroids_dense(labels, qfib_r, qsym)
@@ -345,8 +351,8 @@ def cluster_sph_dbscan(qfib_r, qsym, cl_radius, min_samples):
     _check_dbscan()
     num_ors = qfib_r.shape[1]
     if num_ors > 25000:
-        msg = 'Size too big for sph-dbscan. Defaulting to ort-dbscan.'
-        raise ClusteringError(msg, alternative='ort-dbscan')
+        msg = 'Size too big for sph-dbscan. Defaulting to dbscan.'
+        raise ClusteringError(msg, alternative='dbscan')
 
     dbscan = sklearn.cluster.dbscan
     pts = qfib_r.T