Offset footprint with angle in mind

[ReFil]

I'm trying to automate correcting the origin for some Choc hotswap sockets, however it seems like the rot_footprint filter doesn't offset with respect to the current angle of the component, it just applies a fixed x and Y offset, which leads to inconsistent positioning depending on the angle of the component, and doesn't work at all on a board with the same footprint at multiple angles. Is there a flag i'm missing to have it respect the current angle of the component?

[set-soft]

Please create a small example showing your issue. I.e. containing 2 components with different rotation. Include the schematic, PCB and KiBot config.

The code takes the rotation:

    def apply_offset_value(self, comp, angle, pos_offset_x, pos_offset_y):
        if angle:
            rotation = radians(angle)
            rsin = sin(rotation)
            rcos = cos(rotation)
            comp.pos_offset_x = pos_offset_x * rcos - pos_offset_y * rsin
            comp.pos_offset_y = pos_offset_x * rsin + pos_offset_y * rcos
            logger.debugl(2, f'- rotating offset {angle} degrees: {comp.pos_offset_x}, {comp.pos_offset_y} mm')
        else:
            comp.pos_offset_x = pos_offset_x
            comp.pos_offset_y = pos_offset_y
        # The signs here matches matthewlai/JLCKicadTools offsets because the database comes from this plugin
        comp.pos_offset_x = -GS.from_mm(comp.pos_offset_x)
        comp.pos_offset_y = GS.from_mm(comp.pos_offset_y)
        logger.debugl(2, f'- final offset {comp.pos_offset_x}, {comp.pos_offset_y} KiCad IUs')

But it has too many options to provide compatibility with different tools. So the combination of options is really important.

[ReFil]

Unfortunately that hasn't solved it, I had to change the +ve offsets to -ve in the yaml file so the unrotated component was in the correct location but the rotated one is in the same position it was before. By my maths it should be moved 3.8 and 3.72 mm in x and y respectively but instead it is being moved 0.69 and 5.412 as it is currently according to the logs. did i make a different mistake in the file
angle-problem-demo.zip

[set-soft]

Which math are you using?

[ReFil]

My apologies for the crude diagram, here's how I calculated my x and y offsets.
The absolute distance from the origin to the offset origin remains the same irrelevant of the rotation of the component (5.456mm), and the angle is referenced to the vertical axis (27.27 degrees between the hypotenuse and downward y axis).
I applied the component's rotation in polar coordinate space (so 47.27 degrees between hypotenuse and downwards y axis), then converted back to cartesian.
I made a slight mistake in the maths in my previous message, it should be 3.702 y and 4.008 x offset using this method, which visually looks to be more correct. I think the current method applies the current rotation in the other direction as it is 7.27 degrees between hypotenuse and downwards y axis but length is correct.

[set-soft]

This is the computation KiBot is doing, just with the reverse angle sign. The problem is that this must work for other components. Not just this one. I'm trying with C2935951 which needs both: rotation and offset, and I can figure out how to handle it yet.
Note that you can use schematic fields to control the offset and rotation of individual components to handle cases where KiBot can solve the problem. In this case I recommend using bennymeg_mode: false.

[ReFil]

Ok no worries, thank you for all your assistance

[set-soft]

Ok, I think I found the problem. For the bottom we need to change the angle sign and also the X offset sign. This solves the problem.

I added a test case to the repo here. This includes rotations, offsets and rotations+offsets for the top and bottom.

With the last fix (found in dev branch and images) the test case seems to work really well when uploaded to JLCPCB.

Top side:

Bottom side:

As you can see I included the component you used. Note that I used the following filter:

  - name: rotations
    comment: 'Test case'
    type: rot_footprint
    negative_bottom: false
    mirror_bottom: true
    rotations_and_offsets:
       - regex: 'choc_v1_HS'
         offset_x: -2.5
         offset_y: 4.85