revise plot(MeshLine1)

[gdmcbain]

Some questions about plot(MeshLine), viz.

scikit-fem/skfem/visuals/matplotlib.py

Lines 130 to 131 in d08af2a

	@plot.register(MeshLine)
	def plot_meshline(m: MeshLine, z: ndarray, **kwargs):

were raised in working on the first unsteady one-dimensional example. #674

Because efficiently animating evolving solutions in matplotlib involves something like matplotlib.lines.Line2D.set_data, as in the existing two-dimensional unsteady example

scikit-fem/docs/examples/ex19.py

Line 117 in d08af2a

field.set_array(u)

one needs to know the structure of the data used in the original plot of the initial condition. Whereas this is just the solution's nodal DoFs in two dimensions

scikit-fem/docs/examples/ex19.py

Line 99 in 9608b5f

ax = plot(mesh, u_init[basis.nodal_dofs.flatten()], shading='gouraud')

scikit-fem/docs/examples/ex19.py

Line 117 in 9608b5f

field.set_array(u[basis.nodal_dofs.flatten()])

in one dimension, the initial plot is delegated to plot_meshline which has some rather involved preprocessing

scikit-fem/skfem/visuals/matplotlib.py

Lines 144 to 153 in d08af2a

	for y1, y2, s, t in zip(z[m.t[0]],
	z[m.t[1]],
	m.p[0, m.t[0]],
	m.p[0, m.t[1]]):
	xs.append(s)
	xs.append(t)
	xs.append(None)
	ys.append(y1)
	ys.append(y2)
	ys.append(None)

which needs to be equally applied at each step.

There was a proposal to extract this as a method MeshLine1.segment_data in ad43d00 but this is where questions were raised:

I don't know if this is needed though? Isn't ax.plot(m.p[0], z) enough? This could be just some copypasta from draw_mesh2d where appending None gives faster performance.

The main work of the first unsteady one-dimensional example #674 is to demonstrate how post-processing works in one dimension because one dimension is different to two and three, so if one-dimensional plotting should generally be done differently, that should be sorted out before writing the example.

By

    ax.plot(m.p[0], z)

or a variation with np.argsort, do you mean dispensing with visuals.matplotlib.plot_meshline? I guess that that works well enough in

scikit-fem/docs/examples/ex16.py

Line 49 in d08af2a

ax.plot(x, legendre(n)(x), color=dots.get_color())

[gdmcbain]

Besides 94dc1b8, I've been thinking about another idea to get the various implementations of skfem.visuals.matplotlib.plot to behave more similarly. (The motivation for this is so that one can begin a new project in say one dimension, get it going quickly, and then change up to two or three dimensions, only having to change the least amount of dimension-specific code. Similarly for beginning with P1 elements and changing to something higher order or more exotic.) To be more concrete, I'd like to reduce the differences between exx 19

scikit-fem/docs/examples/ex19.py

Lines 106 to 108 in a1496b6

	ax = plot(mesh, u_init[basis.nodal_dofs.flatten()], shading='gouraud')
	title = ax.set_title('t = 0.00')
	field = ax.get_children()[0] # vertex-based temperature-colour

scikit-fem/docs/examples/ex19.py

Line 122 in a1496b6

field.set_array(u[basis.nodal_dofs.flatten()])

and 39 (which doesn't currently use skfem.visuals)

scikit-fem/docs/examples/ex39.py

Line 94 in a1496b6

line = ax.plot(xi, u_init[sorting], "bo", label="skfem")[0]

scikit-fem/docs/examples/ex39.py

Line 112 in a1496b6

line.set_ydata(u[sorting])

So the still vague idea is that if ax is the matplotlib.axes.Axes returned by skfem.visuals.matplotlib.plot, we might define a method for ax.get_children()[0] that handles updating. Then in ex19 we might update with something like ax.update_plot_element(u[basis.nodal_dofs.flatten()]) and under the hood it would do the equivalent of calling field.set_array(u[basis.nodal_dofs.flatten()]) and in ex39 ax.update_plot_element would effectively call line.set_ydata with the same kind of data (flattened, restricted to nodes, sorted, &c.) as originally passed to plot. I envisage implementing this by defining a closure inside, e.g. plot_meshline, and attaching it as a method to ax before returning it.

[gdmcbain]

…yes, like the addition of the .show method in

scikit-fem/skfem/mesh/mesh.py

Lines 738 to 740 in 8b1be60

	ax = plot(self, *args, **kwargs)
	ax.show = show
	return ax