TransformVolumeData.py: support instance methods as kernels

Useful to look at mesh.extents() in ParaView
without having to write a kernel file.

src/Visualization/Python/TransformVolumeData.py

@@ -682,8 +682,14 @@ def parse_kernels(kernels, exec_files, map_input_names, interactive=False):
             # A module path was specified. Import function from the module
             kernel_module_path, kernel_function = kernel.rsplit(".", maxsplit=1)
             kernel_module = importlib.import_module(kernel_module_path)
+            if ":" in kernel_function:
+                kernel_class, kernel_function = kernel_function.split(":")
+                kernel_class = getattr(kernel_module, kernel_class)
+                kernel_function = getattr(kernel_class, kernel_function)
+            else:
+                kernel_function = getattr(kernel_module, kernel_function)
             yield Kernel(
-                getattr(kernel_module, kernel_function),
+                kernel_function,
                 output_name=None,
                 map_input_names=map_input_names,
                 interactive=interactive,
@@ -717,7 +723,7 @@ def parse_kernels(kernels, exec_files, map_input_names, interactive=False):
     multiple=True,
     help=(
         "Python function to apply to the volume data. "
-        "Specify as 'path.to.module.function_name', where the "
+        "Specify as 'path.to.module:function_name', where the "
         "module must be available to import. "
         "Alternatively, specify just 'function_name' if the "
         "function is defined in one of the '--exec' / '-e' "
@@ -811,16 +817,17 @@ def shift_magnitude(
     'parse_kernel_arg' function for all supported argument types, and
     'parse_kernel_output' for all supported return types.
 
-    The kernels can be loaded from any available Python module. Examples of
-    useful kernels:
+    The kernels can be loaded from any available Python module. Specify them as
+    'path.to.module:function_name' or 'path.to.module.class:function_name'. You
+    can also load a Python file that defines kernels with the '--exec' / '-e'
+    option and then just specify the kernel as 'function_name'.
+    Examples of useful kernels:
 
     - Anything in 'spectre.PointwiseFunctions'
-    - 'spectre.NumericalAlgorithms.LinearOperators.relative_truncation_error'
+    - 'spectre.Spectral.Mesh3D:extents'
+    - 'spectre.NumericalAlgorithms.LinearOperators:relative_truncation_error'
       and 'absolute_truncation_error'
 
-    You can also execute a Python file that defines kernels with the '--exec' /
-    '-e' option.
-
     ## Input and output dataset names
 
     You will be prompted to specify dataset names for input and output tensors,

tests/Unit/Visualization/Python/Test_TransformVolumeData.py

@@ -20,6 +20,7 @@
 from spectre.Spectral import Mesh
 from spectre.Visualization.TransformVolumeData import (
     Kernel,
+    parse_kernels,
     parse_pybind11_signatures,
     snake_case_to_camel_case,
     transform_volume_data,
@@ -226,6 +227,17 @@ def test_integrate(self):
         # particularly precise
         npt.assert_allclose(integrals["Sinusoid"], 64.0, rtol=1e-2)
 
+    def test_parse_kernels(self):
+        (mesh_kernel,) = list(
+            parse_kernels(
+                ["spectre.Spectral.Mesh3D:number_of_grid_points"],
+                exec_files=[],
+                map_input_names={},
+                interactive=False,
+            )
+        )
+        self.assertEqual(mesh_kernel.callable, Mesh[3].number_of_grid_points)
+
     def test_cli(self):
         runner = CliRunner()
         cli_flags = [