Fixing plot indexing

src/ansys/mapdl/core/post.py

@@ -351,7 +351,7 @@ def element_values(self, item, comp="", option="AVG") -> np.ndarray:
         Returns
         -------
         numpy.ndarray
-            Numpy array containing the requested element values for ta
+            Numpy array containing the requested element values for a
             given item and component.
 
         Notes
@@ -632,27 +632,63 @@ def _plot_cell_scalars(self, scalars, show_elem_numbering=False, **kwargs):
 
         surf = self._mapdl.mesh._surf
 
-        # as ``disp`` returns the result for all nodes, we need all node numbers
+        # as ``disp`` returns the result for all nodes/elems, we need all node/elem numbers
         # and to index to the output node numbers
         if hasattr(self._mapdl.mesh, "enum_all"):
-            enum = self._mapdl.mesh.enum
+            enum = self._mapdl.mesh.enum_all
         else:
             enum = self._all_enum
 
-        # it's possible that there are duplicated element numbers,
-        # therefore we need to get the unique values and a reverse index
+        #######################################################################
+        # Bool operations
+        # ===============
+        # I'm going to explain this clearly because it can be confusing for the
+        # future developers (me).
+        # This explanation is based in scalars (`element_values`) NOT having the
+        # full elements (selected and not selected) size.
+        #
+        # First, it's possible that there are duplicated element numbers,
+        # in the surf object returned by Pyvista.
+        # Therefore we need to get the unique values and a reverse index, to
+        # later convert the MAPDL values to Pyvista values.
         uni, ridx = np.unique(surf["ansys_elem_num"], return_inverse=True)
-        mask = np.isin(enum, uni, assume_unique=True)
-
-        if scalars.size != mask.size:
-            scalars = scalars[self.selected_elements]
-        scalars = scalars[mask][ridx]
+        # which means that, uni is the id of mapdl elements in the polydata
+        # object. These elements does not need to be in order, and there can be
+        # duplicated!
+        # Hence:
+        # uni[ridx] = surf["ansys_elem_num"]
+        #
+        # Let's notice that:
+        # * enum[self.selected_elements] is mapdl selected elements ids in MAPDL notation.
+        #
+        # Theoretical approach
+        # --------------------
+        # The theoretical approach will be using an intermediate array of the
+        # size of the MAPDL total number of elements (we do not care about selected).
+        #
+        values = np.zeros(enum.shape)
+        #
+        # Then assign the MAPDL values for the selected element (scalars)
+        #
+        values[self.selected_elements] = scalars
+        #
+        # Because values are in order, but with python notation, then we can do:
+        #
+        surf_values = values[
+            uni - 1
+        ]  # -1 to set MAPDL element indexing to python indexing
+        #
+        # Then to account for the original Pyvista object:
+        #
+        surf_values = surf_values[ridx]
+        #
+        #######################################################################
 
         meshes = [
             {
                 "mesh": surf.copy(deep=False),  # deep=False for ipyvtk-simple
                 "scalar_bar_args": {"title": kwargs.pop("stitle", "")},
-                "scalars": scalars,
+                "scalars": surf_values,
             }
         ]
 

tests/test_post.py

@@ -765,13 +765,44 @@ def test_plot_incomplete_element_selection(mapdl, contact_solve):
     mapdl.esel("S", "ELEM", "", 1, mapdl.mesh.n_elem // 2)
     assert mapdl.post_processing.plot_element_displacement() is None
 
+    mapdl.nsel("S", "NODE", "", 1, mapdl.mesh.n_elem // 2, 2)
+    assert mapdl.post_processing.plot_element_displacement() is None
+
+    mapdl.nsel("S", "NODE", "", 5, mapdl.mesh.n_elem // 2, 2)
+    assert mapdl.post_processing.plot_element_displacement() is None
+
+    mapdl.vsel("s", "", "", 1)
+    mapdl.eslv("s")
+    assert mapdl.post_processing.plot_element_displacement() is None
+
+    mapdl.vsel("s", "", "", 2)
+    mapdl.eslv("s")
+    assert mapdl.post_processing.plot_element_displacement() is None
+
 
 @requires("pyvista")
-@pytest.mark.xfail(strict=False, reason="The image regression is failing. See #2435")
-def test_plot_incomplete_nodal_selection(mapdl, contact_solve):
+def test_plot_incomplete_nodal_selection(mapdl, contact_solve, verify_image_cache):
+    verify_image_cache.skip = True
+
     mapdl.nsel("S", "NODE", "", 1, mapdl.mesh.n_node // 2)
     assert mapdl.post_processing.plot_nodal_displacement() is None
 
+    mapdl.nsel("S", "NODE", "", 1, mapdl.mesh.n_node // 2, 2)
+    assert mapdl.post_processing.plot_nodal_displacement() is None
+
+    mapdl.nsel("S", "NODE", "", 5, mapdl.mesh.n_node // 2, 2)
+    assert mapdl.post_processing.plot_nodal_displacement() is None
+
+    mapdl.vsel("s", "", "", 1)
+    mapdl.eslv("S")
+    mapdl.nsle("S")
+    assert mapdl.post_processing.plot_nodal_displacement() is None
+
+    mapdl.vsel("s", "", "", 2)
+    mapdl.eslv("S")
+    mapdl.nsle("S")
+    assert mapdl.post_processing.plot_nodal_displacement() is None
+
 
 @requires("pyvista")
 def test_general_plotter_returns(mapdl, static_solve, verify_image_cache):