extract_ttfs.py

#!/usr/bin/env python

from fontTools.ttLib import TTFont
import sys
import json

# map of characters to extract
metrics_to_extract = {
    # Font name
    "AMS-Regular": {
        u"\u21e2": None,  # \dashrightarrow
        u"\u21e0": None,  # \dashleftarrow
    },
    "Main-Regular": {
        # Skew and italic metrics can't be easily parsed from the TTF. Instead,
        # we map each character to a "base character", which is a character
        # from the same font with correct italic and skew metrics. A character
        # maps to None if it doesn't have a base.

        u"\u2260": None,  # \neq
        u"\u2245": None,  # \cong
        u"\u0020": None,  # space
        u"\u00a0": None,  # nbsp
        u"\u2026": None,  # \ldots
        u"\u22ef": None,  # \cdots
        u"\u22f1": None,  # \ddots
        u"\u22ee": None,  # \vdots
        u"\u22ee": None,  # \vdots
        u"\u22a8": None,  # \models
        u"\u22c8": None,  # \bowtie
        u"\u2250": None,  # \doteq
        u"\u23b0": None,  # \lmoustache
        u"\u23b1": None,  # \rmoustache
        u"\u27ee": None,  # \lgroup
        u"\u27ef": None,  # \rgroup
        u"\u27f5": None,  # \longleftarrow
        u"\u27f8": None,  # \Longleftarrow
        u"\u27f6": None,  # \longrightarrow
        u"\u27f9": None,  # \Longrightarrow
        u"\u27f7": None,  # \longleftrightarrow
        u"\u27fa": None,  # \Longleftrightarrow
        u"\u21a6": None,  # \mapsto
        u"\u27fc": None,  # \longmapsto
        u"\u21a9": None,  # \hookleftarrow
        u"\u21aa": None,  # \hookrightarrow
        u"\u21cc": None,  # \rightleftharpoons
    },
    "Size1-Regular": {
        u"\u222c": u"\u222b",  # \iint, based on \int
        u"\u222d": u"\u222b",  # \iiint, based on \int
    },
    "Size2-Regular": {
        u"\u222c": u"\u222b",  # \iint, based on \int
        u"\u222d": u"\u222b",  # \iiint, based on \int
    },
}


def main():
    start_json = json.load(sys.stdin)

    for font, chars in metrics_to_extract.iteritems():
        fontInfo = TTFont("../static/fonts/KaTeX_" + font + ".ttf")
        glyf = fontInfo["glyf"]
        unitsPerEm = float(fontInfo["head"].unitsPerEm)

        # We keep ALL Unicode cmaps, not just fontInfo["cmap"].getcmap(3, 1).
        # This is playing it extra safe, since it reports inconsistencies.
        # Platform 0 is Unicode, platform 3 is Windows. For platform 3,
        # encoding 1 is UCS-2 and encoding 10 is UCS-4.
        cmap = [t.cmap for t in fontInfo["cmap"].tables
                if (t.platformID == 0)
                or (t.platformID == 3 and t.platEncID in (1, 10))]

        for char, base_char in chars.iteritems():
            code = ord(char)
            names = set(t.get(code) for t in cmap)
            if not names:
                sys.stderr.write(
                    "Codepoint {} of font {} maps to no name\n"
                    .format(code, font))
                continue
            if len(names) != 1:
                sys.stderr.write(
                    "Codepoint {} of font {} maps to multiple names: {}\n"
                    .format(code, font, ", ".join(sorted(names))))
                continue
            name = names.pop()

            height = depth = italic = skew = 0
            glyph = glyf[name]
            if glyph.numberOfContours:
                height = glyph.yMax
                depth = -glyph.yMin
            if base_char:
                base_char_str = str(ord(base_char))
                base_metrics = start_json[font][base_char_str]
                italic = base_metrics["italic"]
                skew = base_metrics["skew"]

            start_json[font][str(code)] = {
                "height": height / unitsPerEm,
                "depth": depth / unitsPerEm,
                "italic": italic,
                "skew": skew,
            }

    sys.stdout.write(
        json.dumps(start_json, separators=(',', ':'), sort_keys=True))

if __name__ == "__main__":
    main()