IMHO this way of writing is more concise:

Note also that I've used np.isin() instead of np.in1d() because self.codes are 1D.

Edit: I'm wondering if

xys = self.vertices[np.isin(self.codes, [Path.LINETO, Path.MOVETO])]

would be sufficient. Here, the codes of points at line starts cannot be STOP, CURVE3, CURVE4. The only grey area would be CLOSEPOLY followed by LINETO, in which case the line start would be a CLOSEPOLY node. IMHO that's not quite well defined CLOSEPOLY should always be followed by MOVETO.

I think they can be CURVE{3,4}, as there will always be a final control point for those curves (the curve endpoint) which one can then draw a straight line from.

But the elif conditions explicitly excludes CURVE{3,4} here.

Sorry for the lag. A couple of things:

The only grey area would be CLOSEPOLY followed by LINETO, in which case the line start would be a CLOSEPOLY node. IMHO that's not quite well defined CLOSEPOLY should always be followed by MOVETO.

While "undefined" is a reasonable guess, AFAICT, all backends currently handle this behavior consistently. That is, CLOSEPOLY's vertex is ignored and it is treated as "(start_point, LINETO)" plus adding the extra cap.

Which means that CLOSEPOLY followed by LINETO is well-defined, and does the natural thing:

import matplotlib.pyplot as plt
from matplotlib.path import Path
from matplotlib.patches import PathPatch

p = Path(
    [[0, 0], [0, 1], [1, 1], [5, 5], [1, 0], [0, -1], [-5, 5]], 
    [Path.MOVETO, Path.LINETO, Path.LINETO, Path.CLOSEPOLY, 
     Path.LINETO, Path.LINETO, Path.CLOSEPOLY]
)
fig, ax = plt.subplots()
ax.add_patch(PathPatch(p))
ax.set_xlim([-0.5, 1.5])
ax.set_ylim([-1.5, 1.5])

And, unless I'm missing something, then

I'm wondering if

xys = self.vertices[np.isin(self.codes, [Path.LINETO, Path.MOVETO])]

would be sufficient.

tl;dr: yes. We should just be doing this.