Inferring if a set of lines form a rectangle

[qwertynik]

Using the get_drawings method fetched all the drawings on a page. Some rectangles, as seen on the PDF, are represented as lines in the output of get_drawings. The coordinates can be used to guess if the lines are forming a rectangle. Are there other ways to know with greater accuracy that a group of lines form a rectangle?

Posting snippet of the output of the get_drawings method. In the output 'itemsByShape' and 'items' are additional keys added based on the contents of the items key for the path.

{
  "coordinates": {
    "x0": 110.0,
    "y0": 426.3626708984375,
    "x1": 490.0,
    "y1": 474.3626708984375
  },
  "closePath": false,
  "even_odd": false,
  "fill_opacity": 1.0,
  "fill": [
    0.44999998807907104,
    0.08100000023841858,
    0.3109999895095825
  ],
  "seqno": 113,
  "color": null,
  "width": 0,
  "type": "f",
  "lineCap": [
    0
  ],
  "lineJoin": 0,
  "dashes": "[] 0",
  "stroke_opacity": 1,
  "items": [
    {
      "x0": 110.0,
      "y0": 426.3626708984375,
      "x1": 110.0,
      "y1": 426.3626708984375
    },
    {
      "x0": 110.0,
      "y0": 426.3626708984375,
      "x1": 110.0,
      "y1": 474.3626708984375
    },
    {
      "x0": 110.0,
      "y0": 474.3626708984375,
      "x1": 490.0,
      "y1": 474.3626708984375
    },
    {
      "x0": 490.0,
      "y0": 474.3626708984375,
      "x1": 490.0,
      "y1": 426.3626708984375
    }
  ],
  "itemsByShape": {
    "l": [
      0,
      1,
      2,
      3
    ]
  }
}

RectangleDrawnUsingLinesOriginal.pdf

Interestingly, the group of lines is recognized as a rectangle in the Foxit PDF editor. Assuming that there should be something pointing towards the fact that it is a rectangle.

[qwertynik]

@JorjMcKie
Would be great if you could add your insights here.

[qwertynik]

@JorjMcKie
Can you add clarity here? That would help.

Found another use case - triangles are also auto-detected.
singleTriangle.pdf

However, had found only two lines in the path.

import fitz

doc = fitz.open("output/singleTriangle.pdf")
page = doc[0]
paths = page.get_drawings()  # get all drawings on page

print(f"Total paths: {len(paths)}")

# loop through drawings and select relevant ones
for i, path in enumerate(paths):
    pathRect = path["rect"]

    items = path["items"]
    totalItems = len(items)

    print("\n\nPath Details:")
    print(pathRect)
    print('Fill details: ', path['fill'], path['color'])
    print("Total items: ", totalItems)
    for item in path['items']:
        print(item)

('l', Point(199.07200622558594, 285.02899169921875), Point(210.281005859375, 273.82000732421875))
('l', Point(210.281005859375, 273.82000732421875), Point(199.07200622558594, 262.6109924316406))

[JorjMcKie]

The underlying data basis of the method only consists of either lines or curves.
Inside the method get_drawings(), some logic already identifies rectangles and quads in many, but not all cases.
Detection of other shapes, such as circles, ellipses, triangles, etc. are your burden I am afraid.
I have no insight that goes beyond the trivial / obvious unfortunately.

[qwertynik]

@JorjMcKie

[JorjMcKie]

In the method, the above is translated to three lines with connected end points (plus any stroke / fill colors).
Any drawing / image editing app presumably will contain support for all the basic shapes and be able to track down where lines and curves connect, intersect etc.

[qwertynik]

Hmm, makes sense @JorjMcKie
Are you aware of any open-source command-line applications that would be doing this?

[JorjMcKie]

No, but it may be worthwhile looking at wxPython. They have a few accompanying tools to make basic drawings.

[qwertynik]

Thanks, will checkout. However, it looks like wxPython supports drawing and does not particularly detect the shape a given vector in the PDF.