Add FPDF.bezier() method exposing PaintedPath.curve_to

[awmc000]

Begins implementing #496

Checklist:

• The GitHub pipeline is OK (green),
  meaning that both pylint (static code analyzer) and black (code formatter) are happy with the changes of this PR.

• A unit test is covering the code added / modified by this PR

• This PR is ready to be merged

• In case of a new feature, docstrings have been added, with also some documentation in the docs/ folder

• A mention of the change is present in CHANGELOG.md

I do think this could be added as is, but since it only supports cubic Bezier curves in its current state, maybe others would disagree and say this should support any number of control points? That is my next goal.

By submitting this pull request, I confirm that my contribution is made under the terms of the GNU LGPL 3.0 license.

[gmischler]

Cool new feature, thanks @awmc000!

But as always: We want more! 😁

You seem to have noticed the naming confusion yourself:
The method is called bezier(), internally it creates a quadratic bezier, while the test file is called cubic_bezier_curve.pdf

It would actually be nice if the method could create both bezier types. After all, as far as I can tell, the only API change would be to allow point lists with three or four elements, which you can then use to select the appropriate drawing method (and check if the user has supplied a valid number of points at the same time).

The other surprise is that your code only seems to allow closed paths.
Unfortunately, a lot of the functionality in the drawing module is not explicitly documented. However, if you look at the function open_path_drawing() in test/drawing/test_drawing.py, you'll find that you can create open paths with path.style.auto_close = False.
You could expose that option with a closed=False parameter (or default it to True, if you prefer).

[awmc000]

Thank you for the helpful feedback. I'll sort out the issues you mentioned and extend it to support point lists of 3 or 4 points, closed/open paths, and either type of Bezier curves.

[Lucas-C]

Thank you for the helpful feedback. I'll sort out the issues you mentioned and extend it to support point lists of 3 or 4 points, closed/open paths, and either type of Bezier curves.

Thank YOU for taking the time to contribute to fpdf2! 🙂

@allcontributors please add @awmc000 for code.

[allcontributors]

@Lucas-C

I've put up a pull request to add @awmc000! 🎉

[gmischler]

Please check the pylint output and fix any issues it brings up.
You'll also need to rebase your branch, and resolve any conflicts that brings up.

And then I have another wishlist item...
The other shape methods all have a style parameter, which accepts an enum.RenderStyle (or a string), and controls whether to draw the outline of the shape, its area fill, or both. In the drawing module, this seems to translate into the paint_rule field of the GraphicsStyle class, which gets populated with an enum.PathPaintRule value. I'm not sure if we have an example yet on how this translation is best done.
This detail is (hopefully) the last one that is necessary to make your new feature actually useable.

[awmc000]

OK, got it, so I will figure out how to convert a RenderStyle to a PathPaintRule, so that the latter represents the same settings as the former does? And thanks for reminding me about Pylint.

[awmc000]

It looks like the values for RenderStyle and PathPaintRule correspond but just have different names, and that PathPaintRule supports some behaviour RenderStyle doesn't, namely FILL_EVENODD, STROKE_FILL_EVENODD, DONT_PAINT (and AUTO checks PaintedPath.style?).

• DRAW corresponds to STROKE = "S"
• FILL to FILL_NONZERO = "F"
• DRAW_FILL to STROKE_FILL_NONZERO = "B"

So it looks what I need to do is select the PathPaintRule enum value STROKE if the style is D, FILL_NONZERO if the style is F and STROKE_FILL_NONZERO if the style is DF. Does this all sound right?

[awmc000]

I tested that and it seems to work for a basic test in the shell.

>>> from fpdf import FPDF
>>> pdf = FPDF()
>>> pdf.add_page()
>>> pdf.bezier([(20, 20), (20, 10), (30, 30)], style="FD")
>>> pdf.set_fill_color((200, 5, 5))
>>> pdf.bezier([(20, 120), (20, 110), (30, 130)], style="F")
>>> pdf.bezier([(20, 220), (20, 210), (30, 230)], style="D")
>>> pdf.output("stylescr2.pdf")

I get a curve with a fill and stroke, one with just a fill, and one with just the stroke.

[gmischler]

I get a curve with a fill and stroke, one with just a fill, and one with just the stroke.

Looks great!
This appears to be pretty much complete now.

There's yet another thing you might want to test, though.
I think that PaintedPath() automatically picks up the current settings of FPDF.line_width and FPDF.dash_pattern (just like it does with draw_color and fill_color). But it would be good to verify that with an explicit test, so that we know for certain that the new method behaves exactly like the other shapes in all relevant aspects. This will also help prevent regressions during future changes to the code.

Oh, and you should remove any generate=True from your tests before checking in. It currently prevents the tests from running through in the pipeline.

[gmischler]

Two last little nitpicks:

you need to run black on your code, it blocks the testing pipeline.

And I just noticed that you still have a debug_stream=None parameter in the code. While it may be helpful during development that the drawing module offers this option, I don't think that it should be exposed in the public API.

[awmc000]

Should I be getting a debug_stream from somewhere else? Or just remove it? (Removed for now.)

[gmischler]

Should I be getting a debug_stream from somewhere else? Or just remove it? (Removed for now.)

Removing it is fine, as long as you do it in the docstring as well...

[awmc000]

Thanks for catching that.

[gmischler]

There's still an unresolved conflict in CHANGELOG.md (even if github doesn't seem to detect it).

[gmischler]

Thanks for this helpful feature addition, @awmc000 !

merging now.