Merge pull request #1265 from nschloe/separate-int-data

Separate int data

setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = meshio
-version = 5.2.4
+version = 5.2.5
 author = Nico Schlömer et al.
 author_email = nico.schloemer@gmail.com
 description = I/O for many mesh formats

src/meshio/_mesh.py

@@ -313,7 +313,7 @@ def read(cls, path_or_buf, file_format=None):
         warn("meshio.Mesh.read is deprecated, use meshio.read instead")
         return read(path_or_buf, file_format)
 
-    def sets_to_int_data(self):
+    def cell_sets_to_data(self):
         # If possible, convert cell sets to integer cell data. This is possible if all
         # cells appear exactly in one group.
         default_value = -1
@@ -341,8 +341,10 @@ def sets_to_int_data(self):
             self.cell_data[data_name] = intfun
             self.cell_sets = {}
 
+    def point_sets_to_data(self):
         # now for the point sets
         # Go for -1 as the default value. (NaN is not int.)
+        default_value = -1
         if len(self.point_sets) > 0:
             intfun = np.full(len(self.points), default_value, dtype=int)
             for i, cc in enumerate(self.point_sets.values()):
@@ -358,73 +360,61 @@ def sets_to_int_data(self):
             self.point_data[data_name] = intfun
             self.point_sets = {}
 
-    def int_data_to_sets(self, keys: list[str] | None = None):
-        """Convert all int data to {point,cell}_sets, where possible."""
-        rm_keys = []
-        for key, data in self.cell_data.items():
-            if keys is not None:
-                if key not in keys:
-                    continue
-
-            # handle all int and uint data
-            if not all(v.dtype.kind in ["i", "u"] for v in data):
-                continue
-
-            rm_keys.append(key)
-
-            # this call can be rather expensive
-            tags = np.unique(np.concatenate(data))
-
-            # try and get the names by splitting the key along "-" (this is how
-            # sets_to_int_data() forms the key)
-            names = key.split("-")
-            # remove duplicates and preserve order
-            # <https://stackoverflow.com/a/7961390/353337>:
-            names = list(dict.fromkeys(names))
-            if len(names) != len(tags):
-                # alternative names
-                names = [f"set-{key}-{tag}" for tag in tags]
-
-            # TODO there's probably a better way besides np.where, something from
-            # np.unique or np.sort
-            for name, tag in zip(names, tags):
-                self.cell_sets[name] = [np.where(d == tag)[0] for d in data]
+    # This used to be int_data_to_sets(), converting _all_ cell and point data.
+    # This is not useful in many cases, as one usually only wants one
+    # particular data array (e.g., "MaterialIDs") converted to sets.
+    def cell_data_to_sets(self, key: str):
+        """Convert point_data to cell_sets."""
+        data = self.cell_data[key]
+
+        # handle all int and uint data
+        if not all(v.dtype.kind in ["i", "u"] for v in data):
+            raise RuntimeError(f"cell_data['{key}'] is not int data.")
+
+        tags = np.unique(np.concatenate(data))
+
+        # try and get the names by splitting the key along "-" (this is how
+        # sets_to_int_data() forms the key)
+        names = key.split("-")
+        # remove duplicates and preserve order
+        # <https://stackoverflow.com/a/7961390/353337>:
+        names = list(dict.fromkeys(names))
+        if len(names) != len(tags):
+            # alternative names
+            names = [f"set-{key}-{tag}" for tag in tags]
+
+        # TODO there's probably a better way besides np.where, something from
+        # np.unique or np.sort
+        for name, tag in zip(names, tags):
+            self.cell_sets[name] = [np.where(d == tag)[0] for d in data]
 
         # remove the cell data
-        for key in rm_keys:
-            del self.cell_data[key]
-
-        # now point data
-        rm_keys = []
-        for key, data in self.point_data.items():
-            if keys is not None:
-                if key not in keys:
-                    continue
+        del self.cell_data[key]
 
-            # handle all int and uint data
-            if not all(v.dtype.kind in ["i", "u"] for v in data):
-                continue
+    def point_data_to_sets(self, key: str):
+        """Convert point_data to point_sets."""
+        data = self.point_data[key]
 
-            rm_keys.append(key)
+        # handle all int and uint data
+        if not all(v.dtype.kind in ["i", "u"] for v in data):
+            raise RuntimeError(f"point_data['{key}'] is not int data.")
 
-            # this call can be rather expensive
-            tags = np.unique(data)
+        tags = np.unique(data)
 
-            # try and get the names by splitting the key along "-" (this is how
-            # sets_to_int_data() forms the key
-            names = key.split("-")
-            # remove duplicates and preserve order
-            # <https://stackoverflow.com/a/7961390/353337>:
-            names = list(dict.fromkeys(names))
-            if len(names) != len(tags):
-                # alternative names
-                names = [f"set{tag}" for tag in tags]
+        # try and get the names by splitting the key along "-" (this is how
+        # sets_to_int_data() forms the key
+        names = key.split("-")
+        # remove duplicates and preserve order
+        # <https://stackoverflow.com/a/7961390/353337>:
+        names = list(dict.fromkeys(names))
+        if len(names) != len(tags):
+            # alternative names
+            names = [f"set-key-{tag}" for tag in tags]
 
-            # TODO there's probably a better way besides np.where, something from
-            # np.unique or np.sort
-            for name, tag in zip(names, tags):
-                self.point_sets[name] = np.where(data == tag)[0]
+        # TODO there's probably a better way besides np.where, something from
+        # np.unique or np.sort
+        for name, tag in zip(names, tags):
+            self.point_sets[name] = np.where(data == tag)[0]
 
         # remove the cell data
-        for key in rm_keys:
-            del self.point_data[key]
+        del self.point_data[key]

tests/test_mesh.py

@@ -30,7 +30,8 @@ def test_sets_to_int_data():
     mesh = helpers.add_point_sets(mesh)
     mesh = helpers.add_cell_sets(mesh)
 
-    mesh.sets_to_int_data()
+    mesh.point_sets_to_data()
+    mesh.cell_sets_to_data()
 
     assert mesh.cell_sets == {}
     assert_equal(mesh.cell_data, {"grain0-grain1": [[0, 0, 1, 1, 1]]})
@@ -39,7 +40,8 @@ def test_sets_to_int_data():
     assert_equal(mesh.point_data, {"fixed-loose": [0, 0, 0, 1, 1, 1, 1]})
 
     # now back to set data
-    mesh.int_data_to_sets()
+    mesh.cell_data_to_sets("grain0-grain1")
+    mesh.point_data_to_sets("fixed-loose")
 
     assert mesh.cell_data == {}
     assert_equal(mesh.cell_sets, {"grain0": [[0, 1]], "grain1": [[2, 3, 4]]})
@@ -56,7 +58,7 @@ def test_sets_to_int_data_warning():
         cell_sets={"tag": [[0]]},
     )
     with pytest.warns(UserWarning):
-        mesh.sets_to_int_data()
+        mesh.cell_sets_to_data()
     assert np.all(mesh.cell_data["tag"] == np.array([[0, -1]]))
 
     mesh = meshio.Mesh(
@@ -65,7 +67,7 @@ def test_sets_to_int_data_warning():
         point_sets={"tag": [[0, 1, 3]]},
     )
     with pytest.warns(UserWarning):
-        mesh.sets_to_int_data()
+        mesh.point_sets_to_data()
 
     assert np.all(mesh.point_data["tag"] == np.array([[0, 0, -1, 0]]))
 
@@ -74,7 +76,7 @@ def test_int_data_to_sets():
     mesh = helpers.tri_mesh
     mesh.cell_data = {"grain0-grain1": [np.array([0, 1])]}
 
-    mesh.int_data_to_sets()
+    mesh.cell_data_to_sets("grain0-grain1")
 
     assert_equal(mesh.cell_sets, {"grain0": [[0]], "grain1": [[1]]})
 
@@ -92,8 +94,8 @@ def test_gh_1165():
         },
     )
 
-    mesh.sets_to_int_data()
-    mesh.int_data_to_sets()
+    mesh.cell_sets_to_data()
+    mesh.cell_data_to_sets("test-sets")
 
     assert_equal(mesh.cell_sets, {"test": [[], [1]], "sets": [[0, 1], [0, 2, 3]]})