Blog: #pdf

Progress on Press has been a bit slower lately. I’ve fixed most of the errors I discovered by running the exported PDFs through the 3-Heights PDF validator. I also refactored the code and reorganized the package per Kenneth Reitz’s advice.

I’ve implemented initial support for embedding subsetted fonts (doing the subsetting via fontTools.subset), and while the fonts (including uninstalled fonts) display fine on my macOS box, the PDFs don’t validate properly and the fonts don’t show at all on iOS, which means the embedding isn’t actually working right. Current suspects include the /Differences array (which I’m not generating properly yet) and the CMap (which I haven’t implemented at all yet). I still have to implement ToUnicode as well, so that copying and pasting does what it should, but I’m fairly certain that isn’t what’s causing the fonts to not embed properly.

I’m also trying to figure out color spaces. In general I believe I want the output to be either DeviceRGB or DeviceCMYK, with some way of specifying an output intent, and also an option for the user to embed an ICC profile if they want. I’m part of the way there.

Anyway, the font stuff is far more complicated than I expected going in, but I’m still making progress, and I’m learning a lot.

Press can now generate PDFs. For example:

from press import Press

with Press('press-demo.pdf', size=Press.LETTER, margin=1*Press.INCH) as p:
    # Top and bottom borders
    p.pen(0.25)
    p.line(p.page_min_x, p.page_min_y, p.page_max_x, p.page_min_y)
    p.line(p.page_min_x, p.page_max_y, p.page_max_x, p.page_max_y)

    # Rotated colored rectangle
    p.push()
    p.pen(15)
    p.stroke(hsl=(0, 0.5, 0.5))
    p.fill('#fd0')
    p.translate(p.page_min_x + 4*Press.INCH, p.page_max_y - 4*Press.INCH)
    p.rotate(45)
    p.rect(0, 0, 2*Press.INCH, 2*Press.INCH)
    p.pop()

    # Lines of varying thickness
    p.translate(p.page_min_x + 1*Press.INCH, p.page_min_y + 1*Press.CM)
    for i in range(1, 20):
        p.pen(i / 2)
        p.line(0, 0, 30, 0)
        p.translate(0, 20)

That code generates the following PDF (linked):

I’m working on text/font support now, which is by far the most complicated thing about this project.

Since there isn’t a clean, cross-platform way to select a font via code, I’ve decided to use font maps (inspired by @font-face in CSS):

fontmap = {
    'paths': [ './fonts', '/Library/Fonts', ],
    'Minion Pro': [
        { 'weight': 300, 'italics': False, 'filename': 'MinionPro-Regular.otf', },
        { 'weight': 300, 'italics': True,  'filename': 'MinionPro-It.otf', },
        { 'weight': 600, 'italics': False, 'filename': 'MinionPro-Bold.otf', },
        { 'weight': 600, 'italics': True,  'filename': 'MinionPro-BoldIt.otf', },
    ],
}

with Press('output.pdf', size=Press.LETTER, fontmap=fontmap) as p:
    p.font('Minion Pro', size=24, weight=300, italics=True, dlig=True, smcp=True, tracking=50)
    p.align(Press.LEFT)
    p.text("This is a test.", 50, 50)

Font maps are admittedly extra work, but they do have some advantages as well: you can use fonts you haven’t installed, for example, and you can specify exactly which font files you want to use. And I can’t see any good way around the lack of a cross-platform font selection mechanism (meaning, a way to pass in ‘Minion Pro’ with specific weight and styles, and get a font filename in return).

Anyway, I’m in the middle of reading the PDF spec on CIDFonts and CMaps. It’s … complicated. It makes my head hurt. But it’ll be awesome when it’s done.

I’ve renamed inkpdf to Press (as in printing press).

I reached the point where creating the PDF manually is no longer feasible, so I’ve been working on getting Press to a point where I can implement the PDF generation. The basic structure is in place, sans the PDF part. (That’s next.)

Here’s what a Press script looks like right now:

from press import Press

p = Press('output.pdf', width=6*Press.INCH, height=11*Press.INCH,
          margin=1*Press.INCH)

# Horizontal borders at top and bottom of page
p.stroke('#000')
p.pen(1.0)
p.line(p.page_min_x, p.page_min_y, p.page_max_x, p.page_min_y)
p.line(p.page_max_x, p.page_min_y, p.page_max_x, p.page_max_y)

# Page 2
p.page(2)
p.layer('base')
p.stroke(rgb=(1, 0, 0))
p.line(150, 150, 300, 300)
p.layer('fg')
p.stroke(hsl=(0, 0.5, 0.8))
p.line(300, 300, 450, 150)

# Go back and add another line to page 1
p.page(1)
p.stroke('#025')
p.line(p.page_min_x, p.page_min_y, p.page_min_x, p.page_max_y)

p.save() # this doesn't work yet

You can also do something like this:

with Press('output2.pdf', size=Press.LETTER,
           margin=(0.5*Press.INCH, 1.0*Press.INCH),
           inner_margin=0.5*Press.INCH,
           outer_margin=1.25*Press.INCH,
           bleed=.125*Press.INCH) as p:
    p.line(50, 50, 250, 50)
    # And so on

(Context manager, inner/outer margin, bleed, built-in paper sizes.)

Up next: adding more primitives, designing the font selection mechanism, getting it to generate an actual PDF, embedding fonts, using arbitrary Unicode code points, integrating HarfBuzz, etc.