Scalars not properly ordered about pyvista-support

[banesullivan]

So this issue has to do with the ordering of your scalars and how you want them on your structured grid. Simply raveling the scalars in an 'F' order resolves this issue. The color mapping and plotting are all doing what they're supposed to, just where the scalars lie on the mesh is incorrect.

The reason the second example works and shows the correct values at their expected locations on the mesh is that those values are totally symmetric - thus the array, when raveled in the C direction (NumPy's default), would have the same values as the F ordered array.

It's important to note that when you pass the array as the scalars argument in the add_mesh function, .ravel() is called internally on that array - you need to ravel it beforehand since it is F-ordered.

To create the correct visualization, ravel the data array so that it makes a flat vector in F order and associate it with the mesh before plotting.

grid = pv.StructuredGrid(x, y, z)
# Add scalars to the mesh and ravel with F order
grid['scalars'] = z.ravel(order='f')

plotter = pv.Plotter()
plotter.add_mesh(grid, scalars='scalars', cmap='viridis', lighting=False)
plotter.add_axes()
plotter.set_scale(xscale=(np.max(z)/np.max(x)),
                  yscale=(np.max(z)/np.max(y)))
plotter.show_bounds(grid='back',
                    location='outer',
                    ticks='both',
                    bounds=[np.min(x), np.max(x),
                            np.min(y), np.max(y),
                            np.min(z), np.max(z)])
plotter.show()

And to prove the symmetry of the second example:

x_samp = np.linspace(-10, 10, 20)
y_samp = np.linspace(-10, 10, 20)
x, y = np.meshgrid(x_samp, y_samp)
a = -0.0001
z = a*(np.abs(np.sin(x)*np.sin(y)*np.exp(np.abs(100-np.sqrt(x**2 + y**2)/np.pi))) + 1)**0.1

assert np.allclose(z.ravel(order='F'), z.ravel(order='c'))

Some additional tips

• avoid using the add_scalar_bar method on the plotter - this is meant more for internal use. The add_mesh call can control the scalar bar better if the scalars are associated with the mesh like I do in the snippet above
• Be sure to associate scalars with the mesh before plotting - this helps managing those arrays and makes editing the scalar bar and other view properties a bit easier. It's good practice to explicitly use the .point_arrays or .cell_arrays depending on where you want the values to lie.

[ezalorpro]

indeed that solves everything, and with the explicit asignacion of grid['scalars'], add_scalar_bar is not needed anymore.

Thanks

[banesullivan]

explicit asignacion of grid['scalars'], add_scalar_bar is not needed anymore.

Correct, or if you passed the array as the scalars argument to the add_mesh call with a stitle name, then the scalar bar would appear:

plotter.add_mesh(grid, scalars=z.ravel(order='f'), stitle='My Scalars', cmap='viridis', lighting=False)