Advice regarding rendered width/height of text

[konradmalik]

Hi, thanks for a great library!

I have a question on how should I approach the below problem:

• I have a bunch of TEXT entities extracted from a DXF drawing
• I need to know what would be the width and height of the given TEXT entity if it was rendered on a specific-sized canvas

So, for example:
I have a TEXT entity with text: "HELLO" that has some insert point and other standard TEXT attributes. How can I calculate it's pixel-based height and width if it was rendered on a 1000x1000 pixel canvas? Is this even possible?

I understand that this includes font-specific stuff, to keep things simple (simpler?) I can assume that the default MatplotlibBackend settings are ok for me (the PNG I generated with default settings looks very good for my use case).

Any ideas? I'll be honest, I'm not really fluent in all this DWG/DXF stuff.

[mbway]

(disclaimer: I'm not exactly sure what your requirements are here and with more context, a simpler solution might be preferable but I'm answering the question as-given)

Making matplotlib use a fixed size canvas is a little tricky but is possible if you set the figure dpi to a reasonable value (the value doesn't matter so long as it's not extreme) then set the figure size (in inches) such that size_pixels == size_inches * dpi.
Then depending on what you want you have to be careful to set correct data limits on the axis to remove margins.

Once you have done that you will be able to render your text to a fixed-size canvas. To get the size of the text once rendered there are several things you could do, but probably the easiest would be to just iterate over the patches/artists in the axes after you have finished rendering and query the bounding boxes there. You can then use the ax.transFigure transform to transform the bounding box to the 'figure coordinate system' where (0, 0) is the bottom left and (1, 1) is the top right. Then multiply by size_pixels and invert y so that top-left is (0, 0) then you have the bounding box in pixels.

as you can see, matplotlib doesn't have the best usability for doing pixel-perfect rendering. An additional backend which draws directly to a fixed canvas might be a nice addition but isn't something I personally require so haven't looked into much.

[mozman]

The DXF format is not designed to preserve the text layout across CAD applications. If you've ever imported a DXF drawing into any other CAD application, you will know that no other CAD application renders the text and dimensions in the same way as AutoCAD, although BricsCAD comes very close. Perhaps the DXF format and ezdxf are not the right tools for this task.

You can use the text2path add-on to get the text bounding box in drawing units for a single TEXT entity:

import ezdxf.path
from ezdxf.addons import text2path

doc = ezdxf.readfile("your.dxf")
msp = doc.modelspace()

for text in msp.query("TEXT"):
    bbox = ezdxf.path.bbox(text2path.make_paths_from_entity(text))
    print(f"width={bbox.size.x}, height={bbox.size.y}")

Converting this drawing units to pixels is up to you.

[konradmalik]

Thank you both for your suggestions. I ended up using what @mozman proposed. I agree with the mentioned drawbacks. However, this snippet is precisely what I needed for my use case.