PNMImage.flip() documentation doesn't match method, can't transpose with x/y mirror about panda3d HOT 2 CLOSED

Can't repro this using the following code:

from panda3d.core import *

a = PNMImage(2, 2)

a.set_xel_val(0, 0, 0)
a.set_xel_val(0, 1, 1)
a.set_xel_val(1, 0, 2)
a.set_xel_val(1, 1, 3)

print("Original:")
print(a.get_xel_val(0, 0), a.get_xel_val(0, 1), a.get_xel_val(1, 0), a.get_xel_val(1, 1))

a.flip(True, True, True)

print("Post transform:")
print(a.get_xel_val(0, 0), a.get_xel_val(0, 1), a.get_xel_val(1, 0), a.get_xel_val(1, 1))

It works as expected. Looking at the code I also see nothing wrong.

from panda3d.

You're right. The problem I'm actually running into is that the ordering of the operations seems to be inverted in my reference data (produced from the Tiled map editor). It appears that Tiled transposes from top-left to bottom-right (so [0, 1, 2, 3] turns into [0, 2, 1, 3]) whereas Panda3D transposes bottom-left to top-right ([0, 1, 2, 3] -> [3, 1, 2, 0]).

So the right approach in my case is to reverse the flip_x/flip_y when also transposing. So this seems to work:

                if flags.flipped_diagonally:
                    tmpimg.flip(flip_x=flags.flipped_vertically,
                                flip_y=flags.flipped_horizontally,
                                transpose=flags.flipped_diagonally)
                else:
                    tmpimg.flip(flip_x=flags.flipped_horizontally,
                                flip_y=flags.flipped_vertically,
                                transpose=flags.flipped_diagonally)

Otherwise a less efficient solution would be to perform a separate flip() and only transpose, followed by a flip for mirroring X/Y only.

That makes sense now. Thanks for your time.

from panda3d.