#!/usr/bin/python
# -*- coding: utf-8 -*-
"""A spike to see about font cascades with a fallback in Reportlab.

The issue is that in Dercuano I'm using ET Book
(etbook/et-book-roman-old-style-figures.ttf and friends) and I use a
lot of characters that are not in ET Book.  So for those characters I
need to switch to a different font, such as FreeSerif.

Fundamentally this is not too terrible: I need to iterate over the
characters of the string I want to draw, switching fonts
appropriately.  This also requires me to do the same switcheroo for
string width in order to fill paragraphs properly.

"""
import random
import sys

from reportlab.pdfgen.canvas import Canvas
from reportlab.pdfbase.ttfonts import TTFont
from reportlab.pdfbase import pdfmetrics
from reportlab.pdfbase.cidfonts import UnicodeCIDFont


class Cascade:
    """A font cascade is similar to a font (modulo point size) with fallback.

    For code points not in the most preferred font, it falls back to
    less-preferred fonts, finally using the fallback font if nothing
    else has it.  To do this, it must be able to figure out if a font
    has a given code point or not, which I don't know how to do for
    arbitrary Reportlab fonts.  But I do know how to do it for
    TrueType fonts, so those are the ones you can use in the cascade
    list.

    """
    def __init__(self, fonts, fallback):
        """Parameter ``fonts`` is a TTFont list, which have .face.charToGlyph.
        Parameter ``fallback`` can be any Reportlab font; it will be used
        peremptorily for any codepoint not found in ``fonts``.
        """
        self.fonts = fonts
        self.fallback = fallback
        self.string_cache = {}
        self.char_cache = {}

    def find_fonts(self, string):
        "Map a string of chars to (char, font) pairs."
        char_cache = self.char_cache
        for c in string:
            if c in char_cache:
                yield c, char_cache[c]
                continue
            
            for f in self.fonts:
                if ord(c) in f.face.charToGlyph:
                    if len(char_cache) > 1024:
                        char_cache.clear()
                    char_cache[c] = f
                    yield c, f
                    break
            else:
                yield c, self.fallback

    def _map(self, string):
        last_font = object()
        rv = []
        for c, f in self.find_fonts(string):
            if f == last_font:
                chars.append(c)
            else:
                chars = [c]
                rv.append((f, chars))
                last_font = f

        return tuple((f, ''.join(chars)) for f, chars in rv)

    def map(self, string):
        "Map a string to a tuple of (font, substring) pairs."
        if string not in self.string_cache:
            if len(self.string_cache) > 1024:
                self.string_cache.clear()
            self.string_cache[string] = self._map(string)
        return self.string_cache[string]

    def width(self, string, size):
        "Equivalent of font.stringWidth or canvas.stringWidth."
        return sum(f.stringWidth(s, size) for f, s in self.map(string))

    def textOut(self, textobject, string, size):
        "Equivalent of textobject.textOut."
        for f, s in self.map(string):
            print(u"Setting '%s' in %s" % (s, f.fontName))
            textobject.setFont(f.fontName, size)
            textobject.textOut(s)
    
def main():
    c = Canvas('hellocascade.pdf', pagesize=(72*3, 72*2), invariant=True)
    s = [u'⸘eh‽  Liberation Serif covers ≈, †, ∞,',
         u'←↓↑→, ² and ³, and Greek, KΚορη, but',
         u'not ⁻⁶ or ɑ or ₂ or ⁴⁸ or ℤ.  Whee! 你好!']
    f = TTFont('ET-Book-Roman', 'etbook/et-book-roman-old-style-figures.ttf')
    pdfmetrics.registerFont(f)
    fallback0 = TTFont('DejaVuSerif', '/usr/share/fonts/truetype/dejavu/DejaVuSerif.ttf')
    pdfmetrics.registerFont(fallback0)
    fallback1 = TTFont('FreeSerif', '/usr/share/fonts/truetype/freefont/FreeSerif.ttf')
    pdfmetrics.registerFont(fallback1)
    stsong_light = UnicodeCIDFont('STSong-Light')
    pdfmetrics.registerFont(stsong_light)

    cascade = Cascade([f, fallback1], stsong_light)

    c.setFont(f.fontName, 12)
    t = c.beginText(6, 72*2-18)
    for line in s:
        print(u"Width of '%s' is %s." % (line, cascade.width(line, size=12)))
        cascade.textOut(t, line, size=12)
        t.textLine()
        
    words = [u'πορτοκαλη', u'Περσεφόνη', u'Άδης', u'ουχ', u'ο', u'η',
             u'βούλομαι', u'ἐλεύσομαι', u'Σοκρατης', u'Πλατον', u'Διογηνης',
             u'τα', u'είναι', u'μία', u'你好', u'我们', u'不', u'boludo',
             u'estás', u'en', u'el', u'ñoqui', u'pelotuda', u'qué',
             u'carajo', u'sos', u'hadix', u'颱風', u'天气', u'水',
             u'ℤ', u'⇒', u'∃', u'∈', u'†', u'2.5ℓ', u'v⃗',
             # Enough Farsi to see that RtL and Arabic ligatures are totally
             # not handled:
             u'ن سینا‎', u'الجبر‎', u'محمد بن',
             # Devanagari, including combining vowels, seems to be reasonably
             # okay, though the vertical alignment leaves something to be desired:
             u'जुड़े', u'से', u'अल्गोरिद्म']

    generator = random.Random(int(sys.argv[1]) if len(sys.argv) > 1 else 1)
    line = '   '
    capitalize = True
    for i in range(30):
        word = generator.choice(words)
        if capitalize:
            word = word.capitalize()
            capitalize = False

        if not generator.randrange(5):
            word += '. '
            capitalize = True

        possible = line + ' '  + word
        if cascade.width(possible, size=12) > 72*3 - 12:
            cascade.textOut(t, line, size=12)
            t.textLine()
            line = word
        else:
            line = possible

    cascade.textOut(t, line if capitalize else line + '.', size=12)

    c.drawText(t)
    c.save()

if __name__ == '__main__':
    main()