aboutsummaryrefslogtreecommitdiffstats
path: root/contrib/python/fonttools/fontTools/cffLib/width.py
blob: 0ba3ed39bf1c6705d59557bce8835fff6220a0ae (plain) (blame)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
# -*- coding: utf-8 -*-

"""T2CharString glyph width optimizer.

CFF glyphs whose width equals the CFF Private dictionary's ``defaultWidthX``
value do not need to specify their width in their charstring, saving bytes.
This module determines the optimum ``defaultWidthX`` and ``nominalWidthX``
values for a font, when provided with a list of glyph widths."""

from fontTools.ttLib import TTFont
from collections import defaultdict
from operator import add
from functools import reduce


class missingdict(dict):
    def __init__(self, missing_func):
        self.missing_func = missing_func

    def __missing__(self, v):
        return self.missing_func(v)


def cumSum(f, op=add, start=0, decreasing=False):
    keys = sorted(f.keys())
    minx, maxx = keys[0], keys[-1]

    total = reduce(op, f.values(), start)

    if decreasing:
        missing = lambda x: start if x > maxx else total
        domain = range(maxx, minx - 1, -1)
    else:
        missing = lambda x: start if x < minx else total
        domain = range(minx, maxx + 1)

    out = missingdict(missing)

    v = start
    for x in domain:
        v = op(v, f[x])
        out[x] = v

    return out


def byteCost(widths, default, nominal):
    if not hasattr(widths, "items"):
        d = defaultdict(int)
        for w in widths:
            d[w] += 1
        widths = d

    cost = 0
    for w, freq in widths.items():
        if w == default:
            continue
        diff = abs(w - nominal)
        if diff <= 107:
            cost += freq
        elif diff <= 1131:
            cost += freq * 2
        else:
            cost += freq * 5
    return cost


def optimizeWidthsBruteforce(widths):
    """Bruteforce version.  Veeeeeeeeeeeeeeeeery slow.  Only works for smallests of fonts."""

    d = defaultdict(int)
    for w in widths:
        d[w] += 1

    # Maximum number of bytes using default can possibly save
    maxDefaultAdvantage = 5 * max(d.values())

    minw, maxw = min(widths), max(widths)
    domain = list(range(minw, maxw + 1))

    bestCostWithoutDefault = min(byteCost(widths, None, nominal) for nominal in domain)

    bestCost = len(widths) * 5 + 1
    for nominal in domain:
        if byteCost(widths, None, nominal) > bestCost + maxDefaultAdvantage:
            continue
        for default in domain:
            cost = byteCost(widths, default, nominal)
            if cost < bestCost:
                bestCost = cost
                bestDefault = default
                bestNominal = nominal

    return bestDefault, bestNominal


def optimizeWidths(widths):
    """Given a list of glyph widths, or dictionary mapping glyph width to number of
    glyphs having that, returns a tuple of best CFF default and nominal glyph widths.

    This algorithm is linear in UPEM+numGlyphs."""

    if not hasattr(widths, "items"):
        d = defaultdict(int)
        for w in widths:
            d[w] += 1
        widths = d

    keys = sorted(widths.keys())
    minw, maxw = keys[0], keys[-1]
    domain = list(range(minw, maxw + 1))

    # Cumulative sum/max forward/backward.
    cumFrqU = cumSum(widths, op=add)
    cumMaxU = cumSum(widths, op=max)
    cumFrqD = cumSum(widths, op=add, decreasing=True)
    cumMaxD = cumSum(widths, op=max, decreasing=True)

    # Cost per nominal choice, without default consideration.
    nomnCostU = missingdict(
        lambda x: cumFrqU[x] + cumFrqU[x - 108] + cumFrqU[x - 1132] * 3
    )
    nomnCostD = missingdict(
        lambda x: cumFrqD[x] + cumFrqD[x + 108] + cumFrqD[x + 1132] * 3
    )
    nomnCost = missingdict(lambda x: nomnCostU[x] + nomnCostD[x] - widths[x])

    # Cost-saving per nominal choice, by best default choice.
    dfltCostU = missingdict(
        lambda x: max(cumMaxU[x], cumMaxU[x - 108] * 2, cumMaxU[x - 1132] * 5)
    )
    dfltCostD = missingdict(
        lambda x: max(cumMaxD[x], cumMaxD[x + 108] * 2, cumMaxD[x + 1132] * 5)
    )
    dfltCost = missingdict(lambda x: max(dfltCostU[x], dfltCostD[x]))

    # Combined cost per nominal choice.
    bestCost = missingdict(lambda x: nomnCost[x] - dfltCost[x])

    # Best nominal.
    nominal = min(domain, key=lambda x: bestCost[x])

    # Work back the best default.
    bestC = bestCost[nominal]
    dfltC = nomnCost[nominal] - bestCost[nominal]
    ends = []
    if dfltC == dfltCostU[nominal]:
        starts = [nominal, nominal - 108, nominal - 1132]
        for start in starts:
            while cumMaxU[start] and cumMaxU[start] == cumMaxU[start - 1]:
                start -= 1
            ends.append(start)
    else:
        starts = [nominal, nominal + 108, nominal + 1132]
        for start in starts:
            while cumMaxD[start] and cumMaxD[start] == cumMaxD[start + 1]:
                start += 1
            ends.append(start)
    default = min(ends, key=lambda default: byteCost(widths, default, nominal))

    return default, nominal


def main(args=None):
    """Calculate optimum defaultWidthX/nominalWidthX values"""

    import argparse

    parser = argparse.ArgumentParser(
        "fonttools cffLib.width",
        description=main.__doc__,
    )
    parser.add_argument(
        "inputs", metavar="FILE", type=str, nargs="+", help="Input TTF files"
    )
    parser.add_argument(
        "-b",
        "--brute-force",
        dest="brute",
        action="store_true",
        help="Use brute-force approach (VERY slow)",
    )

    args = parser.parse_args(args)

    for fontfile in args.inputs:
        font = TTFont(fontfile)
        hmtx = font["hmtx"]
        widths = [m[0] for m in hmtx.metrics.values()]
        if args.brute:
            default, nominal = optimizeWidthsBruteforce(widths)
        else:
            default, nominal = optimizeWidths(widths)
        print(
            "glyphs=%d default=%d nominal=%d byteCost=%d"
            % (len(widths), default, nominal, byteCost(widths, default, nominal))
        )


if __name__ == "__main__":
    import sys

    if len(sys.argv) == 1:
        import doctest

        sys.exit(doctest.testmod().failed)
    main()