Inconsistent return values from getRealPosition methods

[calebreister]

I've been working on a fairly complicated footprint generator script. Part of this script utilizes the tree shown below.

+ Node
  * Translation [x: -27.25, y: -2.485]
    * Rotation [r: 180]
      * PolygoneLine [[x: -0.5, y: 0.5] ,[x: 0.0, y: 0.0] ,[x: 0.5, y: 0.5]]
        * Line (fp_line (start -0.5 0.5) (end 0 0) (layer F.Fab) (width 0.1))
        * Line (fp_line (start 0 0) (end 0.5 0.5) (layer F.Fab) (width 0.1))

However, attempting to rotate the PolygoneLine object produced an AttributeError:

Traceback (most recent call last):
  File "./QStrip_Vertical.py", line 300, in <module>
    generate_one_footprint(fp_params, config, args.library)
  File "./QStrip_Vertical.py", line 246, in generate_one_footprint
    KicadFileHandler(fp).writeFile(filename)
  File "kicad-footprint-generator/scripts/Connector/Connector_Samtec/../../../KicadModTree/FileHandler.py", line 54, in writeFile
    output = self.serialize(**kwargs)
  File "kicad-footprint-generator/scripts/Connector/Connector_Samtec/../../../KicadModTree/KicadFileHandler.py", line 103, in serialize
    sexpr.extend(self._serializeTree())
  File "kicad-footprint-generator/scripts/Connector/Connector_Samtec/../../../KicadModTree/KicadFileHandler.py", line 143, in _serializeTree
    sexpr.append(self._callSerialize(node))
  File "/home/caleb/GitHub/kicad-footprint-generator/scripts/Connector/Connector_Samtec/../../../KicadModTree/KicadFileHandler.py", line 161, in _callSerialize
    return getattr(self, method_name)(node)
  File "kicad-footprint-generator/scripts/Connector/Connector_Samtec/../../../KicadModTree/KicadFileHandler.py", line 220, in _serialize_Line
    sexpr += self._serialize_LinePoints(node)
  File "kicad-footprint-generator/scripts/Connector/Connector_Samtec/../../../KicadModTree/KicadFileHandler.py", line 211, in _serialize_LinePoints
    ['start', start_pos.x, start_pos.y],
AttributeError: 'tuple' object has no attribute 'x'

After looking at the inner workings of KicadFileHandler.py, Translation.py, and Rotation.py, I found the source of the error: Rotation.getRealPosition() always returns a tuple, and many of the serialization methods expect Node.getRealPosition() to return a Vector2D object. However, if a rotation argument is included, getRealPosition() functions generally return a tuple containing the position and an angle in degrees.

For now, I made a minor modification to KicadFileHandler._serialize_LinePoints() that resolves my issue. However, it would probably be a good idea to make all getRealPosition() methods return a single type. I see several ways to do this:

• Change all rotation=None default parameters to rotation=0 and modify all _serialize methods to accept a tuple from getRealPosition().
• Several classes redefine getRealPosition() to return a dict. For instance, Translation includes this definition:

   translation_coordinate = {'x': parsed_coordinate.x + self.offset_x,
                             'y': parsed_coordinate.y + self.offset_y}
  

  Adding an 'r' or 'theta' element to such a dict (even if it is zero) would be inconsequential in cases where only the coordinates are required (Vector2D ignores the extra element), and it should be fairly easy to modify the _serialize functions that require the rotation.

[pointhi]

I think the cleanest solution would be using a transformation matrice. This should remove quite some of those hacks.

[calebreister]

@pointhi How would that affect backward-compatibility? I don't know much about the inner workings of the API, but there are definitely a variety of things that could be cleaned up.* However, it is much easier to use than the KiCad Python API. If a rework is in order, it may also be a good time to add more kwargs and provide better documentation for the existing ones. Here are a few ideas/suggestions:

• Merge the filled and line shape functions (example: merge PolygoneLine into Polygon with a fill=True kwarg).
• Provide an option to make a mirrored copy of a shape.
• Create a Transform node capable of translating, scaling, and rotating child nodes

If you plan on performing matrix operations, have you considered adding numpy as a dependency?

* For instance, what is a "polygone?"

[calebreister]

@pointhi I can't make any guarantees, but I'm willing to try reworking some of the API code. It would help to have some input on my ideas and an answer to the numpy question.

[poeschlr]

I have added translation and rotation functions for the vector classes in #351 (using transformation matrixes)

So you might not really need to do much but call these functions on all points of the node. (which i also did for a lot of primitives already in that pull request.)