subset.py revision 9d2481bedf905fbca566d99a8abfce9892a692e4 
1# Copyright 2013 Google, Inc. All Rights Reserved.
2#
3# Google Author(s): Behdad Esfahbod
4
5"""Python OpenType Layout Subsetter.
6
7Later grown into full OpenType subsetter, supporting all standard tables.
8"""
9
10from __future__ import print_function, division
11from fontTools.misc.py23 import *
12from fontTools import ttLib
13from fontTools.ttLib.tables import otTables
14from fontTools.misc import psCharStrings
15from fontTools.pens import basePen
16import sys
17import struct
18import time
19import array
20
21
22def _add_method(*clazzes):
23  """Returns a decorator function that adds a new method to one or
24  more classes."""
25  def wrapper(method):
26    for clazz in clazzes:
27      assert clazz.__name__ != 'DefaultTable', 'Oops, table class not found.'
28      assert not hasattr(clazz, method.__name__), \
29          "Oops, class '%s' has method '%s'." % (clazz.__name__,
30                                                 method.__name__)
31      setattr(clazz, method.__name__, method)
32    return None
33  return wrapper
34
35def _uniq_sort(l):
36  return sorted(set(l))
37
38def _set_update(s, *others):
39  # Jython's set.update only takes one other argument.
40  # Emulate real set.update...
41  for other in others:
42    s.update(other)
43
44
45@_add_method(otTables.Coverage)
46def intersect(self, glyphs):
47  "Returns ascending list of matching coverage values."
48  return [i for i,g in enumerate(self.glyphs) if g in glyphs]
49
50@_add_method(otTables.Coverage)
51def intersect_glyphs(self, glyphs):
52  "Returns set of intersecting glyphs."
53  return set(g for g in self.glyphs if g in glyphs)
54
55@_add_method(otTables.Coverage)
56def subset(self, glyphs):
57  "Returns ascending list of remaining coverage values."
58  indices = self.intersect(glyphs)
59  self.glyphs = [g for g in self.glyphs if g in glyphs]
60  return indices
61
62@_add_method(otTables.Coverage)
63def remap(self, coverage_map):
64  "Remaps coverage."
65  self.glyphs = [self.glyphs[i] for i in coverage_map]
66
67@_add_method(otTables.ClassDef)
68def intersect(self, glyphs):
69  "Returns ascending list of matching class values."
70  return _uniq_sort(
71     ([0] if any(g not in self.classDefs for g in glyphs) else []) +
72      [v for g,v in self.classDefs.items() if g in glyphs])
73
74@_add_method(otTables.ClassDef)
75def intersect_class(self, glyphs, klass):
76  "Returns set of glyphs matching class."
77  if klass == 0:
78    return set(g for g in glyphs if g not in self.classDefs)
79  return set(g for g,v in self.classDefs.items()
80              if v == klass and g in glyphs)
81
82@_add_method(otTables.ClassDef)
83def subset(self, glyphs, remap=False):
84  "Returns ascending list of remaining classes."
85  self.classDefs = dict((g,v) for g,v in self.classDefs.items() if g in glyphs)
86  # Note: while class 0 has the special meaning of "not matched",
87  # if no glyph will ever /not match/, we can optimize class 0 out too.
88  indices = _uniq_sort(
89     ([0] if any(g not in self.classDefs for g in glyphs) else []) +
90      list(self.classDefs.values()))
91  if remap:
92    self.remap(indices)
93  return indices
94
95@_add_method(otTables.ClassDef)
96def remap(self, class_map):
97  "Remaps classes."
98  self.classDefs = dict((g,class_map.index(v))
99                         for g,v in self.classDefs.items())
100
101@_add_method(otTables.SingleSubst)
102def closure_glyphs(self, s, cur_glyphs=None):
103  if cur_glyphs is None: cur_glyphs = s.glyphs
104  if self.Format in [1, 2]:
105    s.glyphs.update(v for g,v in self.mapping.items() if g in cur_glyphs)
106  else:
107    assert 0, "unknown format: %s" % self.Format
108
109@_add_method(otTables.SingleSubst)
110def subset_glyphs(self, s):
111  if self.Format in [1, 2]:
112    self.mapping = dict((g,v) for g,v in self.mapping.items()
113                        if g in s.glyphs and v in s.glyphs)
114    return bool(self.mapping)
115  else:
116    assert 0, "unknown format: %s" % self.Format
117
118@_add_method(otTables.MultipleSubst)
119def closure_glyphs(self, s, cur_glyphs=None):
120  if cur_glyphs is None: cur_glyphs = s.glyphs
121  if self.Format == 1:
122    indices = self.Coverage.intersect(cur_glyphs)
123    _set_update(s.glyphs, *(self.Sequence[i].Substitute for i in indices))
124  else:
125    assert 0, "unknown format: %s" % self.Format
126
127@_add_method(otTables.MultipleSubst)
128def subset_glyphs(self, s):
129  if self.Format == 1:
130    indices = self.Coverage.subset(s.glyphs)
131    self.Sequence = [self.Sequence[i] for i in indices]
132    # Now drop rules generating glyphs we don't want
133    indices = [i for i,seq in enumerate(self.Sequence)
134         if all(sub in s.glyphs for sub in seq.Substitute)]
135    self.Sequence = [self.Sequence[i] for i in indices]
136    self.Coverage.remap(indices)
137    self.SequenceCount = len(self.Sequence)
138    return bool(self.SequenceCount)
139  else:
140    assert 0, "unknown format: %s" % self.Format
141
142@_add_method(otTables.AlternateSubst)
143def closure_glyphs(self, s, cur_glyphs=None):
144  if cur_glyphs is None: cur_glyphs = s.glyphs
145  if self.Format == 1:
146    _set_update(s.glyphs, *(vlist for g,vlist in self.alternates.items()
147                            if g in cur_glyphs))
148  else:
149    assert 0, "unknown format: %s" % self.Format
150
151@_add_method(otTables.AlternateSubst)
152def subset_glyphs(self, s):
153  if self.Format == 1:
154    self.alternates = dict((g,vlist)
155                           for g,vlist in self.alternates.items()
156                           if g in s.glyphs and
157                              all(v in s.glyphs for v in vlist))
158    return bool(self.alternates)
159  else:
160    assert 0, "unknown format: %s" % self.Format
161
162@_add_method(otTables.LigatureSubst)
163def closure_glyphs(self, s, cur_glyphs=None):
164  if cur_glyphs is None: cur_glyphs = s.glyphs
165  if self.Format == 1:
166    _set_update(s.glyphs, *([seq.LigGlyph for seq in seqs
167                             if all(c in s.glyphs for c in seq.Component)]
168                            for g,seqs in self.ligatures.items()
169                            if g in cur_glyphs))
170  else:
171    assert 0, "unknown format: %s" % self.Format
172
173@_add_method(otTables.LigatureSubst)
174def subset_glyphs(self, s):
175  if self.Format == 1:
176    self.ligatures = dict((g,v) for g,v in self.ligatures.items()
177                          if g in s.glyphs)
178    self.ligatures = dict((g,[seq for seq in seqs
179                              if seq.LigGlyph in s.glyphs and
180                                 all(c in s.glyphs for c in seq.Component)])
181                           for g,seqs in self.ligatures.items())
182    self.ligatures = dict((g,v) for g,v in self.ligatures.items() if v)
183    return bool(self.ligatures)
184  else:
185    assert 0, "unknown format: %s" % self.Format
186
187@_add_method(otTables.ReverseChainSingleSubst)
188def closure_glyphs(self, s, cur_glyphs=None):
189  if cur_glyphs is None: cur_glyphs = s.glyphs
190  if self.Format == 1:
191    indices = self.Coverage.intersect(cur_glyphs)
192    if(not indices or
193        not all(c.intersect(s.glyphs)
194                 for c in self.LookAheadCoverage + self.BacktrackCoverage)):
195      return
196    s.glyphs.update(self.Substitute[i] for i in indices)
197  else:
198    assert 0, "unknown format: %s" % self.Format
199
200@_add_method(otTables.ReverseChainSingleSubst)
201def subset_glyphs(self, s):
202  if self.Format == 1:
203    indices = self.Coverage.subset(s.glyphs)
204    self.Substitute = [self.Substitute[i] for i in indices]
205    # Now drop rules generating glyphs we don't want
206    indices = [i for i,sub in enumerate(self.Substitute)
207         if sub in s.glyphs]
208    self.Substitute = [self.Substitute[i] for i in indices]
209    self.Coverage.remap(indices)
210    self.GlyphCount = len(self.Substitute)
211    return bool(self.GlyphCount and
212                 all(c.subset(s.glyphs)
213                      for c in self.LookAheadCoverage+self.BacktrackCoverage))
214  else:
215    assert 0, "unknown format: %s" % self.Format
216
217@_add_method(otTables.SinglePos)
218def subset_glyphs(self, s):
219  if self.Format == 1:
220    return len(self.Coverage.subset(s.glyphs))
221  elif self.Format == 2:
222    indices = self.Coverage.subset(s.glyphs)
223    self.Value = [self.Value[i] for i in indices]
224    self.ValueCount = len(self.Value)
225    return bool(self.ValueCount)
226  else:
227    assert 0, "unknown format: %s" % self.Format
228
229@_add_method(otTables.SinglePos)
230def prune_post_subset(self, options):
231  if not options.hinting:
232    # Drop device tables
233    self.ValueFormat &= ~0x00F0
234  return True
235
236@_add_method(otTables.PairPos)
237def subset_glyphs(self, s):
238  if self.Format == 1:
239    indices = self.Coverage.subset(s.glyphs)
240    self.PairSet = [self.PairSet[i] for i in indices]
241    for p in self.PairSet:
242      p.PairValueRecord = [r for r in p.PairValueRecord
243                           if r.SecondGlyph in s.glyphs]
244      p.PairValueCount = len(p.PairValueRecord)
245    self.PairSet = [p for p in self.PairSet if p.PairValueCount]
246    self.PairSetCount = len(self.PairSet)
247    return bool(self.PairSetCount)
248  elif self.Format == 2:
249    class1_map = self.ClassDef1.subset(s.glyphs, remap=True)
250    class2_map = self.ClassDef2.subset(s.glyphs, remap=True)
251    self.Class1Record = [self.Class1Record[i] for i in class1_map]
252    for c in self.Class1Record:
253      c.Class2Record = [c.Class2Record[i] for i in class2_map]
254    self.Class1Count = len(class1_map)
255    self.Class2Count = len(class2_map)
256    return bool(self.Class1Count and
257                 self.Class2Count and
258                 self.Coverage.subset(s.glyphs))
259  else:
260    assert 0, "unknown format: %s" % self.Format
261
262@_add_method(otTables.PairPos)
263def prune_post_subset(self, options):
264  if not options.hinting:
265    # Drop device tables
266    self.ValueFormat1 &= ~0x00F0
267    self.ValueFormat2 &= ~0x00F0
268  return True
269
270@_add_method(otTables.CursivePos)
271def subset_glyphs(self, s):
272  if self.Format == 1:
273    indices = self.Coverage.subset(s.glyphs)
274    self.EntryExitRecord = [self.EntryExitRecord[i] for i in indices]
275    self.EntryExitCount = len(self.EntryExitRecord)
276    return bool(self.EntryExitCount)
277  else:
278    assert 0, "unknown format: %s" % self.Format
279
280@_add_method(otTables.Anchor)
281def prune_hints(self):
282  # Drop device tables / contour anchor point
283  self.Format = 1
284
285@_add_method(otTables.CursivePos)
286def prune_post_subset(self, options):
287  if not options.hinting:
288    for rec in self.EntryExitRecord:
289      if rec.EntryAnchor: rec.EntryAnchor.prune_hints()
290      if rec.ExitAnchor: rec.ExitAnchor.prune_hints()
291  return True
292
293@_add_method(otTables.MarkBasePos)
294def subset_glyphs(self, s):
295  if self.Format == 1:
296    mark_indices = self.MarkCoverage.subset(s.glyphs)
297    self.MarkArray.MarkRecord = [self.MarkArray.MarkRecord[i]
298                                 for i in mark_indices]
299    self.MarkArray.MarkCount = len(self.MarkArray.MarkRecord)
300    base_indices = self.BaseCoverage.subset(s.glyphs)
301    self.BaseArray.BaseRecord = [self.BaseArray.BaseRecord[i]
302                                 for i in base_indices]
303    self.BaseArray.BaseCount = len(self.BaseArray.BaseRecord)
304    # Prune empty classes
305    class_indices = _uniq_sort(v.Class for v in self.MarkArray.MarkRecord)
306    self.ClassCount = len(class_indices)
307    for m in self.MarkArray.MarkRecord:
308      m.Class = class_indices.index(m.Class)
309    for b in self.BaseArray.BaseRecord:
310      b.BaseAnchor = [b.BaseAnchor[i] for i in class_indices]
311    return bool(self.ClassCount and
312                 self.MarkArray.MarkCount and
313                 self.BaseArray.BaseCount)
314  else:
315    assert 0, "unknown format: %s" % self.Format
316
317@_add_method(otTables.MarkBasePos)
318def prune_post_subset(self, options):
319    if not options.hinting:
320      for m in self.MarkArray.MarkRecord:
321        if m.MarkAnchor:
322          m.MarkAnchor.prune_hints()
323      for b in self.BaseArray.BaseRecord:
324        for a in b.BaseAnchor:
325          if a:
326            a.prune_hints()
327    return True
328
329@_add_method(otTables.MarkLigPos)
330def subset_glyphs(self, s):
331  if self.Format == 1:
332    mark_indices = self.MarkCoverage.subset(s.glyphs)
333    self.MarkArray.MarkRecord = [self.MarkArray.MarkRecord[i]
334                                 for i in mark_indices]
335    self.MarkArray.MarkCount = len(self.MarkArray.MarkRecord)
336    ligature_indices = self.LigatureCoverage.subset(s.glyphs)
337    self.LigatureArray.LigatureAttach = [self.LigatureArray.LigatureAttach[i]
338                                         for i in ligature_indices]
339    self.LigatureArray.LigatureCount = len(self.LigatureArray.LigatureAttach)
340    # Prune empty classes
341    class_indices = _uniq_sort(v.Class for v in self.MarkArray.MarkRecord)
342    self.ClassCount = len(class_indices)
343    for m in self.MarkArray.MarkRecord:
344      m.Class = class_indices.index(m.Class)
345    for l in self.LigatureArray.LigatureAttach:
346      for c in l.ComponentRecord:
347        c.LigatureAnchor = [c.LigatureAnchor[i] for i in class_indices]
348    return bool(self.ClassCount and
349                 self.MarkArray.MarkCount and
350                 self.LigatureArray.LigatureCount)
351  else:
352    assert 0, "unknown format: %s" % self.Format
353
354@_add_method(otTables.MarkLigPos)
355def prune_post_subset(self, options):
356    if not options.hinting:
357      for m in self.MarkArray.MarkRecord:
358        if m.MarkAnchor:
359          m.MarkAnchor.prune_hints()
360      for l in self.LigatureArray.LigatureAttach:
361        for c in l.ComponentRecord:
362          for a in c.LigatureAnchor:
363            if a:
364              a.prune_hints()
365    return True
366
367@_add_method(otTables.MarkMarkPos)
368def subset_glyphs(self, s):
369  if self.Format == 1:
370    mark1_indices = self.Mark1Coverage.subset(s.glyphs)
371    self.Mark1Array.MarkRecord = [self.Mark1Array.MarkRecord[i]
372                                  for i in mark1_indices]
373    self.Mark1Array.MarkCount = len(self.Mark1Array.MarkRecord)
374    mark2_indices = self.Mark2Coverage.subset(s.glyphs)
375    self.Mark2Array.Mark2Record = [self.Mark2Array.Mark2Record[i]
376                                   for i in mark2_indices]
377    self.Mark2Array.MarkCount = len(self.Mark2Array.Mark2Record)
378    # Prune empty classes
379    class_indices = _uniq_sort(v.Class for v in self.Mark1Array.MarkRecord)
380    self.ClassCount = len(class_indices)
381    for m in self.Mark1Array.MarkRecord:
382      m.Class = class_indices.index(m.Class)
383    for b in self.Mark2Array.Mark2Record:
384      b.Mark2Anchor = [b.Mark2Anchor[i] for i in class_indices]
385    return bool(self.ClassCount and
386                 self.Mark1Array.MarkCount and
387                 self.Mark2Array.MarkCount)
388  else:
389    assert 0, "unknown format: %s" % self.Format
390
391@_add_method(otTables.MarkMarkPos)
392def prune_post_subset(self, options):
393    if not options.hinting:
394      # Drop device tables or contour anchor point
395      for m in self.Mark1Array.MarkRecord:
396        if m.MarkAnchor:
397          m.MarkAnchor.prune_hints()
398      for b in self.Mark2Array.Mark2Record:
399        for m in b.Mark2Anchor:
400          if m:
401            m.prune_hints()
402    return True
403
404@_add_method(otTables.SingleSubst,
405             otTables.MultipleSubst,
406             otTables.AlternateSubst,
407             otTables.LigatureSubst,
408             otTables.ReverseChainSingleSubst,
409             otTables.SinglePos,
410             otTables.PairPos,
411             otTables.CursivePos,
412             otTables.MarkBasePos,
413             otTables.MarkLigPos,
414             otTables.MarkMarkPos)
415def subset_lookups(self, lookup_indices):
416  pass
417
418@_add_method(otTables.SingleSubst,
419             otTables.MultipleSubst,
420             otTables.AlternateSubst,
421             otTables.LigatureSubst,
422             otTables.ReverseChainSingleSubst,
423             otTables.SinglePos,
424             otTables.PairPos,
425             otTables.CursivePos,
426             otTables.MarkBasePos,
427             otTables.MarkLigPos,
428             otTables.MarkMarkPos)
429def collect_lookups(self):
430  return []
431
432@_add_method(otTables.SingleSubst,
433             otTables.MultipleSubst,
434             otTables.AlternateSubst,
435             otTables.LigatureSubst,
436             otTables.ContextSubst,
437             otTables.ChainContextSubst,
438             otTables.ReverseChainSingleSubst,
439             otTables.SinglePos,
440             otTables.PairPos,
441             otTables.CursivePos,
442             otTables.MarkBasePos,
443             otTables.MarkLigPos,
444             otTables.MarkMarkPos,
445             otTables.ContextPos,
446             otTables.ChainContextPos)
447def prune_pre_subset(self, options):
448  return True
449
450@_add_method(otTables.SingleSubst,
451             otTables.MultipleSubst,
452             otTables.AlternateSubst,
453             otTables.LigatureSubst,
454             otTables.ReverseChainSingleSubst,
455             otTables.ContextSubst,
456             otTables.ChainContextSubst,
457             otTables.ContextPos,
458             otTables.ChainContextPos)
459def prune_post_subset(self, options):
460  return True
461
462@_add_method(otTables.SingleSubst,
463             otTables.AlternateSubst,
464             otTables.ReverseChainSingleSubst)
465def may_have_non_1to1(self):
466  return False
467
468@_add_method(otTables.MultipleSubst,
469             otTables.LigatureSubst,
470             otTables.ContextSubst,
471             otTables.ChainContextSubst)
472def may_have_non_1to1(self):
473  return True
474
475@_add_method(otTables.ContextSubst,
476             otTables.ChainContextSubst,
477             otTables.ContextPos,
478             otTables.ChainContextPos)
479def __classify_context(self):
480
481  class ContextHelper(object):
482    def __init__(self, klass, Format):
483      if klass.__name__.endswith('Subst'):
484        Typ = 'Sub'
485        Type = 'Subst'
486      else:
487        Typ = 'Pos'
488        Type = 'Pos'
489      if klass.__name__.startswith('Chain'):
490        Chain = 'Chain'
491      else:
492        Chain = ''
493      ChainTyp = Chain+Typ
494
495      self.Typ = Typ
496      self.Type = Type
497      self.Chain = Chain
498      self.ChainTyp = ChainTyp
499
500      self.LookupRecord = Type+'LookupRecord'
501
502      if Format == 1:
503        Coverage = lambda r: r.Coverage
504        ChainCoverage = lambda r: r.Coverage
505        ContextData = lambda r:(None,)
506        ChainContextData = lambda r:(None, None, None)
507        RuleData = lambda r:(r.Input,)
508        ChainRuleData = lambda r:(r.Backtrack, r.Input, r.LookAhead)
509        SetRuleData = None
510        ChainSetRuleData = None
511      elif Format == 2:
512        Coverage = lambda r: r.Coverage
513        ChainCoverage = lambda r: r.Coverage
514        ContextData = lambda r:(r.ClassDef,)
515        ChainContextData = lambda r:(r.LookAheadClassDef,
516                                      r.InputClassDef,
517                                      r.BacktrackClassDef)
518        RuleData = lambda r:(r.Class,)
519        ChainRuleData = lambda r:(r.LookAhead, r.Input, r.Backtrack)
520        def SetRuleData(r, d):(r.Class,) = d
521        def ChainSetRuleData(r, d):(r.LookAhead, r.Input, r.Backtrack) = d
522      elif Format == 3:
523        Coverage = lambda r: r.Coverage[0]
524        ChainCoverage = lambda r: r.InputCoverage[0]
525        ContextData = None
526        ChainContextData = None
527        RuleData = lambda r: r.Coverage
528        ChainRuleData = lambda r:(r.LookAheadCoverage +
529                                   r.InputCoverage +
530                                   r.BacktrackCoverage)
531        SetRuleData = None
532        ChainSetRuleData = None
533      else:
534        assert 0, "unknown format: %s" % Format
535
536      if Chain:
537        self.Coverage = ChainCoverage
538        self.ContextData = ChainContextData
539        self.RuleData = ChainRuleData
540        self.SetRuleData = ChainSetRuleData
541      else:
542        self.Coverage = Coverage
543        self.ContextData = ContextData
544        self.RuleData = RuleData
545        self.SetRuleData = SetRuleData
546
547      if Format == 1:
548        self.Rule = ChainTyp+'Rule'
549        self.RuleCount = ChainTyp+'RuleCount'
550        self.RuleSet = ChainTyp+'RuleSet'
551        self.RuleSetCount = ChainTyp+'RuleSetCount'
552        self.Intersect = lambda glyphs, c, r: [r] if r in glyphs else []
553      elif Format == 2:
554        self.Rule = ChainTyp+'ClassRule'
555        self.RuleCount = ChainTyp+'ClassRuleCount'
556        self.RuleSet = ChainTyp+'ClassSet'
557        self.RuleSetCount = ChainTyp+'ClassSetCount'
558        self.Intersect = lambda glyphs, c, r: c.intersect_class(glyphs, r)
559
560        self.ClassDef = 'InputClassDef' if Chain else 'ClassDef'
561        self.ClassDefIndex = 1 if Chain else 0
562        self.Input = 'Input' if Chain else 'Class'
563
564  if self.Format not in [1, 2, 3]:
565    return None  # Don't shoot the messenger; let it go
566  if not hasattr(self.__class__, "__ContextHelpers"):
567    self.__class__.__ContextHelpers = {}
568  if self.Format not in self.__class__.__ContextHelpers:
569    helper = ContextHelper(self.__class__, self.Format)
570    self.__class__.__ContextHelpers[self.Format] = helper
571  return self.__class__.__ContextHelpers[self.Format]
572
573@_add_method(otTables.ContextSubst,
574             otTables.ChainContextSubst)
575def closure_glyphs(self, s, cur_glyphs=None):
576  if cur_glyphs is None: cur_glyphs = s.glyphs
577  c = self.__classify_context()
578
579  indices = c.Coverage(self).intersect(s.glyphs)
580  if not indices:
581    return []
582  cur_glyphs = c.Coverage(self).intersect_glyphs(s.glyphs);
583
584  if self.Format == 1:
585    ContextData = c.ContextData(self)
586    rss = getattr(self, c.RuleSet)
587    rssCount = getattr(self, c.RuleSetCount)
588    for i in indices:
589      if i >= rssCount or not rss[i]: continue
590      for r in getattr(rss[i], c.Rule):
591        if not r: continue
592        if all(all(c.Intersect(s.glyphs, cd, k) for k in klist)
593          for cd,klist in zip(ContextData, c.RuleData(r))):
594          chaos = False
595          for ll in getattr(r, c.LookupRecord):
596            if not ll: continue
597            seqi = ll.SequenceIndex
598            if chaos:
599              pos_glyphs = s.glyphs
600            else:
601              if seqi == 0:
602                pos_glyphs = set([c.Coverage(self).glyphs[i]])
603              else:
604                pos_glyphs = set([r.Input[seqi - 1]])
605            lookup = s.table.LookupList.Lookup[ll.LookupListIndex]
606            chaos = chaos or lookup.may_have_non_1to1()
607            lookup.closure_glyphs(s, cur_glyphs=pos_glyphs)
608  elif self.Format == 2:
609    ClassDef = getattr(self, c.ClassDef)
610    indices = ClassDef.intersect(cur_glyphs)
611    ContextData = c.ContextData(self)
612    rss = getattr(self, c.RuleSet)
613    rssCount = getattr(self, c.RuleSetCount)
614    for i in indices:
615      if i >= rssCount or not rss[i]: continue
616      for r in getattr(rss[i], c.Rule):
617        if not r: continue
618        if all(all(c.Intersect(s.glyphs, cd, k) for k in klist)
619          for cd,klist in zip(ContextData, c.RuleData(r))):
620          chaos = False
621          for ll in getattr(r, c.LookupRecord):
622            if not ll: continue
623            seqi = ll.SequenceIndex
624            if chaos:
625              pos_glyphs = s.glyphs
626            else:
627              if seqi == 0:
628                pos_glyphs = ClassDef.intersect_class(cur_glyphs, i)
629              else:
630                pos_glyphs = ClassDef.intersect_class(s.glyphs,
631                                                      getattr(r, c.Input)[seqi - 1])
632            lookup = s.table.LookupList.Lookup[ll.LookupListIndex]
633            chaos = chaos or lookup.may_have_non_1to1()
634            lookup.closure_glyphs(s, cur_glyphs=pos_glyphs)
635  elif self.Format == 3:
636    if not all(x.intersect(s.glyphs) for x in c.RuleData(self)):
637      return []
638    r = self
639    chaos = False
640    for ll in getattr(r, c.LookupRecord):
641      if not ll: continue
642      seqi = ll.SequenceIndex
643      if chaos:
644        pos_glyphs = s.glyphs
645      else:
646        if seqi == 0:
647          pos_glyphs = cur_glyphs
648        else:
649          pos_glyphs = r.InputCoverage[seqi].intersect_glyphs(s.glyphs)
650      lookup = s.table.LookupList.Lookup[ll.LookupListIndex]
651      chaos = chaos or lookup.may_have_non_1to1()
652      lookup.closure_glyphs(s, cur_glyphs=pos_glyphs)
653  else:
654    assert 0, "unknown format: %s" % self.Format
655
656@_add_method(otTables.ContextSubst,
657             otTables.ContextPos,
658             otTables.ChainContextSubst,
659             otTables.ChainContextPos)
660def subset_glyphs(self, s):
661  c = self.__classify_context()
662
663  if self.Format == 1:
664    indices = self.Coverage.subset(s.glyphs)
665    rss = getattr(self, c.RuleSet)
666    rss = [rss[i] for i in indices]
667    for rs in rss:
668      if not rs: continue
669      ss = getattr(rs, c.Rule)
670      ss = [r for r in ss
671            if r and all(all(g in s.glyphs for g in glist)
672              for glist in c.RuleData(r))]
673      setattr(rs, c.Rule, ss)
674      setattr(rs, c.RuleCount, len(ss))
675    # Prune empty subrulesets
676    rss = [rs for rs in rss if rs and getattr(rs, c.Rule)]
677    setattr(self, c.RuleSet, rss)
678    setattr(self, c.RuleSetCount, len(rss))
679    return bool(rss)
680  elif self.Format == 2:
681    if not self.Coverage.subset(s.glyphs):
682      return False
683    ContextData = c.ContextData(self)
684    klass_maps = [x.subset(s.glyphs, remap=True) for x in ContextData]
685
686    # Keep rulesets for class numbers that survived.
687    indices = klass_maps[c.ClassDefIndex]
688    rss = getattr(self, c.RuleSet)
689    rssCount = getattr(self, c.RuleSetCount)
690    rss = [rss[i] for i in indices if i < rssCount]
691    del rssCount
692    # Delete, but not renumber, unreachable rulesets.
693    indices = getattr(self, c.ClassDef).intersect(self.Coverage.glyphs)
694    rss = [rss if i in indices else None for i,rss in enumerate(rss)]
695    while rss and rss[-1] is None:
696      del rss[-1]
697
698    for rs in rss:
699      if not rs: continue
700      ss = getattr(rs, c.Rule)
701      ss = [r for r in ss
702            if r and all(all(k in klass_map for k in klist)
703              for klass_map,klist in zip(klass_maps, c.RuleData(r)))]
704      setattr(rs, c.Rule, ss)
705      setattr(rs, c.RuleCount, len(ss))
706
707      # Remap rule classes
708      for r in ss:
709        c.SetRuleData(r, [[klass_map.index(k) for k in klist]
710               for klass_map,klist in zip(klass_maps, c.RuleData(r))])
711    return bool(rss)
712  elif self.Format == 3:
713    return all(x.subset(s.glyphs) for x in c.RuleData(self))
714  else:
715    assert 0, "unknown format: %s" % self.Format
716
717@_add_method(otTables.ContextSubst,
718             otTables.ChainContextSubst,
719             otTables.ContextPos,
720             otTables.ChainContextPos)
721def subset_lookups(self, lookup_indices):
722  c = self.__classify_context()
723
724  if self.Format in [1, 2]:
725    for rs in getattr(self, c.RuleSet):
726      if not rs: continue
727      for r in getattr(rs, c.Rule):
728        if not r: continue
729        setattr(r, c.LookupRecord,
730                 [ll for ll in getattr(r, c.LookupRecord)
731                  if ll and ll.LookupListIndex in lookup_indices])
732        for ll in getattr(r, c.LookupRecord):
733          if not ll: continue
734          ll.LookupListIndex = lookup_indices.index(ll.LookupListIndex)
735  elif self.Format == 3:
736    setattr(self, c.LookupRecord,
737             [ll for ll in getattr(self, c.LookupRecord)
738              if ll and ll.LookupListIndex in lookup_indices])
739    for ll in getattr(self, c.LookupRecord):
740      if not ll: continue
741      ll.LookupListIndex = lookup_indices.index(ll.LookupListIndex)
742  else:
743    assert 0, "unknown format: %s" % self.Format
744
745@_add_method(otTables.ContextSubst,
746             otTables.ChainContextSubst,
747             otTables.ContextPos,
748             otTables.ChainContextPos)
749def collect_lookups(self):
750  c = self.__classify_context()
751
752  if self.Format in [1, 2]:
753    return [ll.LookupListIndex
754      for rs in getattr(self, c.RuleSet) if rs
755      for r in getattr(rs, c.Rule) if r
756      for ll in getattr(r, c.LookupRecord) if ll]
757  elif self.Format == 3:
758    return [ll.LookupListIndex
759      for ll in getattr(self, c.LookupRecord) if ll]
760  else:
761    assert 0, "unknown format: %s" % self.Format
762
763@_add_method(otTables.ExtensionSubst)
764def closure_glyphs(self, s, cur_glyphs=None):
765  if self.Format == 1:
766    self.ExtSubTable.closure_glyphs(s, cur_glyphs)
767  else:
768    assert 0, "unknown format: %s" % self.Format
769
770@_add_method(otTables.ExtensionSubst)
771def may_have_non_1to1(self):
772  if self.Format == 1:
773    return self.ExtSubTable.may_have_non_1to1()
774  else:
775    assert 0, "unknown format: %s" % self.Format
776
777@_add_method(otTables.ExtensionSubst,
778             otTables.ExtensionPos)
779def prune_pre_subset(self, options):
780  if self.Format == 1:
781    return self.ExtSubTable.prune_pre_subset(options)
782  else:
783    assert 0, "unknown format: %s" % self.Format
784
785@_add_method(otTables.ExtensionSubst,
786             otTables.ExtensionPos)
787def subset_glyphs(self, s):
788  if self.Format == 1:
789    return self.ExtSubTable.subset_glyphs(s)
790  else:
791    assert 0, "unknown format: %s" % self.Format
792
793@_add_method(otTables.ExtensionSubst,
794             otTables.ExtensionPos)
795def prune_post_subset(self, options):
796  if self.Format == 1:
797    return self.ExtSubTable.prune_post_subset(options)
798  else:
799    assert 0, "unknown format: %s" % self.Format
800
801@_add_method(otTables.ExtensionSubst,
802             otTables.ExtensionPos)
803def subset_lookups(self, lookup_indices):
804  if self.Format == 1:
805    return self.ExtSubTable.subset_lookups(lookup_indices)
806  else:
807    assert 0, "unknown format: %s" % self.Format
808
809@_add_method(otTables.ExtensionSubst,
810             otTables.ExtensionPos)
811def collect_lookups(self):
812  if self.Format == 1:
813    return self.ExtSubTable.collect_lookups()
814  else:
815    assert 0, "unknown format: %s" % self.Format
816
817@_add_method(otTables.Lookup)
818def closure_glyphs(self, s, cur_glyphs=None):
819  for st in self.SubTable:
820    if not st: continue
821    st.closure_glyphs(s, cur_glyphs)
822
823@_add_method(otTables.Lookup)
824def prune_pre_subset(self, options):
825  ret = False
826  for st in self.SubTable:
827    if not st: continue
828    if st.prune_pre_subset(options): ret = True
829  return ret
830
831@_add_method(otTables.Lookup)
832def subset_glyphs(self, s):
833  self.SubTable = [st for st in self.SubTable if st and st.subset_glyphs(s)]
834  self.SubTableCount = len(self.SubTable)
835  return bool(self.SubTableCount)
836
837@_add_method(otTables.Lookup)
838def prune_post_subset(self, options):
839  ret = False
840  for st in self.SubTable:
841    if not st: continue
842    if st.prune_post_subset(options): ret = True
843  return ret
844
845@_add_method(otTables.Lookup)
846def subset_lookups(self, lookup_indices):
847  for s in self.SubTable:
848    s.subset_lookups(lookup_indices)
849
850@_add_method(otTables.Lookup)
851def collect_lookups(self):
852  return _uniq_sort(sum((st.collect_lookups() for st in self.SubTable
853                         if st), []))
854
855@_add_method(otTables.Lookup)
856def may_have_non_1to1(self):
857  return any(st.may_have_non_1to1() for st in self.SubTable if st)
858
859@_add_method(otTables.LookupList)
860def prune_pre_subset(self, options):
861  ret = False
862  for l in self.Lookup:
863    if not l: continue
864    if l.prune_pre_subset(options): ret = True
865  return ret
866
867@_add_method(otTables.LookupList)
868def subset_glyphs(self, s):
869  "Returns the indices of nonempty lookups."
870  return [i for i,l in enumerate(self.Lookup) if l and l.subset_glyphs(s)]
871
872@_add_method(otTables.LookupList)
873def prune_post_subset(self, options):
874  ret = False
875  for l in self.Lookup:
876    if not l: continue
877    if l.prune_post_subset(options): ret = True
878  return ret
879
880@_add_method(otTables.LookupList)
881def subset_lookups(self, lookup_indices):
882  self.Lookup = [self.Lookup[i] for i in lookup_indices
883                 if i < self.LookupCount]
884  self.LookupCount = len(self.Lookup)
885  for l in self.Lookup:
886    l.subset_lookups(lookup_indices)
887
888@_add_method(otTables.LookupList)
889def closure_lookups(self, lookup_indices):
890  lookup_indices = _uniq_sort(lookup_indices)
891  recurse = lookup_indices
892  while True:
893    recurse_lookups = sum((self.Lookup[i].collect_lookups()
894                            for i in recurse if i < self.LookupCount), [])
895    recurse_lookups = [l for l in recurse_lookups
896                       if l not in lookup_indices and l < self.LookupCount]
897    if not recurse_lookups:
898      return _uniq_sort(lookup_indices)
899    recurse_lookups = _uniq_sort(recurse_lookups)
900    lookup_indices.extend(recurse_lookups)
901    recurse = recurse_lookups
902
903@_add_method(otTables.Feature)
904def subset_lookups(self, lookup_indices):
905  self.LookupListIndex = [l for l in self.LookupListIndex
906                          if l in lookup_indices]
907  # Now map them.
908  self.LookupListIndex = [lookup_indices.index(l)
909                          for l in self.LookupListIndex]
910  self.LookupCount = len(self.LookupListIndex)
911  return self.LookupCount or self.FeatureParams
912
913@_add_method(otTables.Feature)
914def collect_lookups(self):
915  return self.LookupListIndex[:]
916
917@_add_method(otTables.FeatureList)
918def subset_lookups(self, lookup_indices):
919  "Returns the indices of nonempty features."
920  feature_indices = [i for i,f in enumerate(self.FeatureRecord)
921                     if f.Feature.subset_lookups(lookup_indices)]
922  self.subset_features(feature_indices)
923  return feature_indices
924
925@_add_method(otTables.FeatureList)
926def collect_lookups(self, feature_indices):
927  return _uniq_sort(sum((self.FeatureRecord[i].Feature.collect_lookups()
928                         for i in feature_indices
929                          if i < self.FeatureCount), []))
930
931@_add_method(otTables.FeatureList)
932def subset_features(self, feature_indices):
933  self.FeatureRecord = [self.FeatureRecord[i] for i in feature_indices]
934  self.FeatureCount = len(self.FeatureRecord)
935  return bool(self.FeatureCount)
936
937@_add_method(otTables.DefaultLangSys,
938             otTables.LangSys)
939def subset_features(self, feature_indices):
940  if self.ReqFeatureIndex in feature_indices:
941    self.ReqFeatureIndex = feature_indices.index(self.ReqFeatureIndex)
942  else:
943    self.ReqFeatureIndex = 65535
944  self.FeatureIndex = [f for f in self.FeatureIndex if f in feature_indices]
945  # Now map them.
946  self.FeatureIndex = [feature_indices.index(f) for f in self.FeatureIndex
947                       if f in feature_indices]
948  self.FeatureCount = len(self.FeatureIndex)
949  return bool(self.FeatureCount or self.ReqFeatureIndex != 65535)
950
951@_add_method(otTables.DefaultLangSys,
952             otTables.LangSys)
953def collect_features(self):
954  feature_indices = self.FeatureIndex[:]
955  if self.ReqFeatureIndex != 65535:
956    feature_indices.append(self.ReqFeatureIndex)
957  return _uniq_sort(feature_indices)
958
959@_add_method(otTables.Script)
960def subset_features(self, feature_indices):
961  if(self.DefaultLangSys and
962      not self.DefaultLangSys.subset_features(feature_indices)):
963    self.DefaultLangSys = None
964  self.LangSysRecord = [l for l in self.LangSysRecord
965                        if l.LangSys.subset_features(feature_indices)]
966  self.LangSysCount = len(self.LangSysRecord)
967  return bool(self.LangSysCount or self.DefaultLangSys)
968
969@_add_method(otTables.Script)
970def collect_features(self):
971  feature_indices = [l.LangSys.collect_features() for l in self.LangSysRecord]
972  if self.DefaultLangSys:
973    feature_indices.append(self.DefaultLangSys.collect_features())
974  return _uniq_sort(sum(feature_indices, []))
975
976@_add_method(otTables.ScriptList)
977def subset_features(self, feature_indices):
978  self.ScriptRecord = [s for s in self.ScriptRecord
979                       if s.Script.subset_features(feature_indices)]
980  self.ScriptCount = len(self.ScriptRecord)
981  return bool(self.ScriptCount)
982
983@_add_method(otTables.ScriptList)
984def collect_features(self):
985  return _uniq_sort(sum((s.Script.collect_features()
986                         for s in self.ScriptRecord), []))
987
988@_add_method(ttLib.getTableClass('GSUB'))
989def closure_glyphs(self, s):
990  s.table = self.table
991  if self.table.ScriptList:
992    feature_indices = self.table.ScriptList.collect_features()
993  else:
994    feature_indices = []
995  if self.table.FeatureList:
996    lookup_indices = self.table.FeatureList.collect_lookups(feature_indices)
997  else:
998    lookup_indices = []
999  if self.table.LookupList:
1000    while True:
1001      orig_glyphs = s.glyphs.copy()
1002      for i in lookup_indices:
1003        if i >= self.table.LookupList.LookupCount: continue
1004        if not self.table.LookupList.Lookup[i]: continue
1005        self.table.LookupList.Lookup[i].closure_glyphs(s)
1006      if orig_glyphs == s.glyphs:
1007        break
1008  del s.table
1009
1010@_add_method(ttLib.getTableClass('GSUB'),
1011             ttLib.getTableClass('GPOS'))
1012def subset_glyphs(self, s):
1013  s.glyphs = s.glyphs_gsubed
1014  if self.table.LookupList:
1015    lookup_indices = self.table.LookupList.subset_glyphs(s)
1016  else:
1017    lookup_indices = []
1018  self.subset_lookups(lookup_indices)
1019  self.prune_lookups()
1020  return True
1021
1022@_add_method(ttLib.getTableClass('GSUB'),
1023             ttLib.getTableClass('GPOS'))
1024def subset_lookups(self, lookup_indices):
1025  """Retains specified lookups, then removes empty features, language
1026     systems, and scripts."""
1027  if self.table.LookupList:
1028    self.table.LookupList.subset_lookups(lookup_indices)
1029  if self.table.FeatureList:
1030    feature_indices = self.table.FeatureList.subset_lookups(lookup_indices)
1031  else:
1032    feature_indices = []
1033  if self.table.ScriptList:
1034    self.table.ScriptList.subset_features(feature_indices)
1035
1036@_add_method(ttLib.getTableClass('GSUB'),
1037             ttLib.getTableClass('GPOS'))
1038def prune_lookups(self):
1039  "Remove unreferenced lookups"
1040  if self.table.ScriptList:
1041    feature_indices = self.table.ScriptList.collect_features()
1042  else:
1043    feature_indices = []
1044  if self.table.FeatureList:
1045    lookup_indices = self.table.FeatureList.collect_lookups(feature_indices)
1046  else:
1047    lookup_indices = []
1048  if self.table.LookupList:
1049    lookup_indices = self.table.LookupList.closure_lookups(lookup_indices)
1050  else:
1051    lookup_indices = []
1052  self.subset_lookups(lookup_indices)
1053
1054@_add_method(ttLib.getTableClass('GSUB'),
1055             ttLib.getTableClass('GPOS'))
1056def subset_feature_tags(self, feature_tags):
1057  if self.table.FeatureList:
1058    feature_indices = [i for i,f in
1059                       enumerate(self.table.FeatureList.FeatureRecord)
1060                       if f.FeatureTag in feature_tags]
1061    self.table.FeatureList.subset_features(feature_indices)
1062  else:
1063    feature_indices = []
1064  if self.table.ScriptList:
1065    self.table.ScriptList.subset_features(feature_indices)
1066
1067@_add_method(ttLib.getTableClass('GSUB'),
1068             ttLib.getTableClass('GPOS'))
1069def prune_pre_subset(self, options):
1070  # Drop undesired features
1071  if '*' not in options.layout_features:
1072    self.subset_feature_tags(options.layout_features)
1073  # Drop unreferenced lookups
1074  self.prune_lookups()
1075  # Prune lookups themselves
1076  if self.table.LookupList:
1077    self.table.LookupList.prune_pre_subset(options);
1078  return True
1079
1080@_add_method(ttLib.getTableClass('GSUB'),
1081             ttLib.getTableClass('GPOS'))
1082def prune_post_subset(self, options):
1083  table = self.table
1084  if table.ScriptList and not table.ScriptList.ScriptRecord:
1085    table.ScriptList = None
1086  if table.FeatureList and not table.FeatureList.FeatureRecord:
1087    table.FeatureList = None
1088  # Prune lookups themselves
1089  if table.LookupList:
1090    table.LookupList.prune_post_subset(options);
1091    if not table.LookupList.Lookup:
1092      table.LookupList = None
1093  return True
1094
1095@_add_method(ttLib.getTableClass('GDEF'))
1096def subset_glyphs(self, s):
1097  glyphs = s.glyphs_gsubed
1098  table = self.table
1099  if table.LigCaretList:
1100    indices = table.LigCaretList.Coverage.subset(glyphs)
1101    table.LigCaretList.LigGlyph = [table.LigCaretList.LigGlyph[i]
1102                                   for i in indices]
1103    table.LigCaretList.LigGlyphCount = len(table.LigCaretList.LigGlyph)
1104  if table.MarkAttachClassDef:
1105    table.MarkAttachClassDef.classDefs = dict((g,v) for g,v in
1106                                              table.MarkAttachClassDef.
1107                                                classDefs.items()
1108                                              if g in glyphs)
1109  if table.GlyphClassDef:
1110    table.GlyphClassDef.classDefs = dict((g,v) for g,v in
1111                                         table.GlyphClassDef.
1112                                           classDefs.items()
1113                                         if g in glyphs)
1114  if table.AttachList:
1115    indices = table.AttachList.Coverage.subset(glyphs)
1116    GlyphCount = table.AttachList.GlyphCount
1117    table.AttachList.AttachPoint = [table.AttachList.AttachPoint[i]
1118                                    for i in indices
1119                                    if i < GlyphCount]
1120    table.AttachList.GlyphCount = len(table.AttachList.AttachPoint)
1121  if hasattr(table, "MarkGlyphSetsDef") and table.MarkGlyphSetsDef:
1122    for coverage in table.MarkGlyphSetsDef.Coverage:
1123      coverage.subset(glyphs)
1124    # TODO: The following is disabled.  If enabling, we need to go fixup all
1125    # lookups that use MarkFilteringSet and map their set.
1126    #indices = table.MarkGlyphSetsDef.Coverage = [c for c in table.MarkGlyphSetsDef.Coverage if c.glyphs]
1127  return True
1128
1129@_add_method(ttLib.getTableClass('GDEF'))
1130def prune_post_subset(self, options):
1131  table = self.table
1132  if table.LigCaretList and not table.LigCaretList.LigGlyphCount:
1133    table.LigCaretList = None
1134  if table.MarkAttachClassDef and not table.MarkAttachClassDef.classDefs:
1135    table.MarkAttachClassDef = None
1136  if table.GlyphClassDef and not table.GlyphClassDef.classDefs:
1137    table.GlyphClassDef = None
1138  if table.AttachList and not table.AttachList.GlyphCount:
1139    table.AttachList = None
1140  if hasattr(table, "MarkGlyphSetsDef") and table.MarkGlyphSetsDef and not table.MarkGlyphSetsDef.Coverage:
1141    table.MarkGlyphSetsDef = None
1142    if table.Version == 0x00010002/0x10000:
1143      table.Version = 1.0
1144  return bool(table.LigCaretList or
1145              table.MarkAttachClassDef or
1146              table.GlyphClassDef or
1147              table.AttachList or
1148              (table.Version >= 0x00010002/0x10000 and table.MarkGlyphSetsDef))
1149
1150@_add_method(ttLib.getTableClass('kern'))
1151def prune_pre_subset(self, options):
1152  # Prune unknown kern table types
1153  self.kernTables = [t for t in self.kernTables if hasattr(t, 'kernTable')]
1154  return bool(self.kernTables)
1155
1156@_add_method(ttLib.getTableClass('kern'))
1157def subset_glyphs(self, s):
1158  glyphs = s.glyphs_gsubed
1159  for t in self.kernTables:
1160    t.kernTable = dict(((a,b),v) for (a,b),v in t.kernTable.items()
1161                       if a in glyphs and b in glyphs)
1162  self.kernTables = [t for t in self.kernTables if t.kernTable]
1163  return bool(self.kernTables)
1164
1165@_add_method(ttLib.getTableClass('vmtx'))
1166def subset_glyphs(self, s):
1167  self.metrics = dict((g,v) for g,v in self.metrics.items() if g in s.glyphs)
1168  return bool(self.metrics)
1169
1170@_add_method(ttLib.getTableClass('hmtx'))
1171def subset_glyphs(self, s):
1172  self.metrics = dict((g,v) for g,v in self.metrics.items() if g in s.glyphs)
1173  return True # Required table
1174
1175@_add_method(ttLib.getTableClass('hdmx'))
1176def subset_glyphs(self, s):
1177  self.hdmx = dict((sz,dict((g,v) for g,v in l.items() if g in s.glyphs))
1178                   for sz,l in self.hdmx.items())
1179  return bool(self.hdmx)
1180
1181@_add_method(ttLib.getTableClass('VORG'))
1182def subset_glyphs(self, s):
1183  self.VOriginRecords = dict((g,v) for g,v in self.VOriginRecords.items()
1184                             if g in s.glyphs)
1185  self.numVertOriginYMetrics = len(self.VOriginRecords)
1186  return True  # Never drop; has default metrics
1187
1188@_add_method(ttLib.getTableClass('post'))
1189def prune_pre_subset(self, options):
1190  if not options.glyph_names:
1191    self.formatType = 3.0
1192  return True # Required table
1193
1194@_add_method(ttLib.getTableClass('post'))
1195def subset_glyphs(self, s):
1196  self.extraNames = []  # This seems to do it
1197  return True # Required table
1198
1199@_add_method(ttLib.getTableModule('glyf').Glyph)
1200def remapComponentsFast(self, indices):
1201  if not self.data or struct.unpack(">h", self.data[:2])[0] >= 0:
1202    return  # Not composite
1203  data = array.array("B", self.data)
1204  i = 10
1205  more = 1
1206  while more:
1207    flags =(data[i] << 8) | data[i+1]
1208    glyphID =(data[i+2] << 8) | data[i+3]
1209    # Remap
1210    glyphID = indices.index(glyphID)
1211    data[i+2] = glyphID >> 8
1212    data[i+3] = glyphID & 0xFF
1213    i += 4
1214    flags = int(flags)
1215
1216    if flags & 0x0001: i += 4  # ARG_1_AND_2_ARE_WORDS
1217    else: i += 2
1218    if flags & 0x0008: i += 2  # WE_HAVE_A_SCALE
1219    elif flags & 0x0040: i += 4  # WE_HAVE_AN_X_AND_Y_SCALE
1220    elif flags & 0x0080: i += 8  # WE_HAVE_A_TWO_BY_TWO
1221    more = flags & 0x0020  # MORE_COMPONENTS
1222
1223  self.data = data.tostring()
1224
1225@_add_method(ttLib.getTableClass('glyf'))
1226def closure_glyphs(self, s):
1227  decompose = s.glyphs
1228  while True:
1229    components = set()
1230    for g in decompose:
1231      if g not in self.glyphs:
1232        continue
1233      gl = self.glyphs[g]
1234      for c in gl.getComponentNames(self):
1235        if c not in s.glyphs:
1236          components.add(c)
1237    components = set(c for c in components if c not in s.glyphs)
1238    if not components:
1239      break
1240    decompose = components
1241    s.glyphs.update(components)
1242
1243@_add_method(ttLib.getTableClass('glyf'))
1244def prune_pre_subset(self, options):
1245  if options.notdef_glyph and not options.notdef_outline:
1246    g = self[self.glyphOrder[0]]
1247    # Yay, easy!
1248    g.__dict__.clear()
1249    g.data = ""
1250  return True
1251
1252@_add_method(ttLib.getTableClass('glyf'))
1253def subset_glyphs(self, s):
1254  self.glyphs = dict((g,v) for g,v in self.glyphs.items() if g in s.glyphs)
1255  indices = [i for i,g in enumerate(self.glyphOrder) if g in s.glyphs]
1256  for v in self.glyphs.values():
1257    if hasattr(v, "data"):
1258      v.remapComponentsFast(indices)
1259    else:
1260      pass  # No need
1261  self.glyphOrder = [g for g in self.glyphOrder if g in s.glyphs]
1262  # Don't drop empty 'glyf' tables, otherwise 'loca' doesn't get subset.
1263  return True
1264
1265@_add_method(ttLib.getTableClass('glyf'))
1266def prune_post_subset(self, options):
1267  if not options.hinting:
1268    for v in self.glyphs.values():
1269      v.removeHinting()
1270  return True
1271
1272@_add_method(ttLib.getTableClass('CFF '))
1273def prune_pre_subset(self, options):
1274  cff = self.cff
1275  # CFF table must have one font only
1276  cff.fontNames = cff.fontNames[:1]
1277
1278  if options.notdef_glyph and not options.notdef_outline:
1279    for fontname in cff.keys():
1280      font = cff[fontname]
1281      c,_ = font.CharStrings.getItemAndSelector('.notdef')
1282      # XXX we should preserve the glyph width
1283      c.bytecode = '\x0e' # endchar
1284      c.program = None
1285
1286  return True # bool(cff.fontNames)
1287
1288@_add_method(ttLib.getTableClass('CFF '))
1289def subset_glyphs(self, s):
1290  cff = self.cff
1291  for fontname in cff.keys():
1292    font = cff[fontname]
1293    cs = font.CharStrings
1294
1295    # Load all glyphs
1296    for g in font.charset:
1297      if g not in s.glyphs: continue
1298      c,sel = cs.getItemAndSelector(g)
1299
1300    if cs.charStringsAreIndexed:
1301      indices = [i for i,g in enumerate(font.charset) if g in s.glyphs]
1302      csi = cs.charStringsIndex
1303      csi.items = [csi.items[i] for i in indices]
1304      csi.count = len(csi.items)
1305      del csi.file, csi.offsets
1306      if hasattr(font, "FDSelect"):
1307        sel = font.FDSelect
1308        sel.format = None
1309        sel.gidArray = [sel.gidArray[i] for i in indices]
1310      cs.charStrings = dict((g,indices.index(v))
1311                            for g,v in cs.charStrings.items()
1312                            if g in s.glyphs)
1313    else:
1314      cs.charStrings = dict((g,v)
1315                            for g,v in cs.charStrings.items()
1316                            if g in s.glyphs)
1317    font.charset = [g for g in font.charset if g in s.glyphs]
1318    font.numGlyphs = len(font.charset)
1319
1320  return True # any(cff[fontname].numGlyphs for fontname in cff.keys())
1321
1322@_add_method(psCharStrings.T2CharString)
1323def subset_subroutines(self, subrs, gsubrs):
1324  p = self.program
1325  assert len(p)
1326  for i in range(1, len(p)):
1327    if p[i] == 'callsubr':
1328      assert isinstance(p[i-1], int)
1329      p[i-1] = subrs._used.index(p[i-1] + subrs._old_bias) - subrs._new_bias
1330    elif p[i] == 'callgsubr':
1331      assert isinstance(p[i-1], int)
1332      p[i-1] = gsubrs._used.index(p[i-1] + gsubrs._old_bias) - gsubrs._new_bias
1333
1334@_add_method(psCharStrings.T2CharString)
1335def drop_hints(self):
1336  hints = self._hints
1337
1338  if hints.has_hint:
1339    self.program = self.program[hints.last_hint:]
1340    if hasattr(self, 'width'):
1341      # Insert width back if needed
1342      if self.width != self.private.defaultWidthX:
1343        self.program.insert(0, self.width - self.private.nominalWidthX)
1344
1345  if hints.has_hintmask:
1346    i = 0
1347    p = self.program
1348    while i < len(p):
1349      if p[i] in ['hintmask', 'cntrmask']:
1350        assert i + 1 <= len(p)
1351        del p[i:i+2]
1352        continue
1353      i += 1
1354
1355  # TODO: we currently don't drop calls to "empty" subroutines.
1356
1357  assert len(self.program)
1358
1359  del self._hints
1360
1361class _MarkingT2Decompiler(psCharStrings.SimpleT2Decompiler):
1362
1363  def __init__(self, localSubrs, globalSubrs):
1364    psCharStrings.SimpleT2Decompiler.__init__(self,
1365                                              localSubrs,
1366                                              globalSubrs)
1367    for subrs in [localSubrs, globalSubrs]:
1368      if subrs and not hasattr(subrs, "_used"):
1369        subrs._used = set()
1370
1371  def op_callsubr(self, index):
1372    self.localSubrs._used.add(self.operandStack[-1]+self.localBias)
1373    psCharStrings.SimpleT2Decompiler.op_callsubr(self, index)
1374
1375  def op_callgsubr(self, index):
1376    self.globalSubrs._used.add(self.operandStack[-1]+self.globalBias)
1377    psCharStrings.SimpleT2Decompiler.op_callgsubr(self, index)
1378
1379class _DehintingT2Decompiler(psCharStrings.SimpleT2Decompiler):
1380
1381  class Hints(object):
1382    def __init__(self):
1383      # Whether calling this charstring produces any hint stems
1384      self.has_hint = False
1385      # Index to start at to drop all hints
1386      self.last_hint = 0
1387      # Index up to which we know more hints are possible.  Only
1388      # relevant if status is 0 or 1.
1389      self.last_checked = 0
1390      # The status means:
1391      # 0: after dropping hints, this charstring is empty
1392      # 1: after dropping hints, there may be more hints continuing after this
1393      # 2: no more hints possible after this charstring
1394      self.status = 0
1395      # Has hintmask instructions; not recursive
1396      self.has_hintmask = False
1397    pass
1398
1399  def __init__(self, css, localSubrs, globalSubrs):
1400    self._css = css
1401    psCharStrings.SimpleT2Decompiler.__init__(self,
1402                                              localSubrs,
1403                                              globalSubrs)
1404
1405  def execute(self, charString):
1406    old_hints = charString._hints if hasattr(charString, '_hints') else None
1407    charString._hints = self.Hints()
1408
1409    psCharStrings.SimpleT2Decompiler.execute(self, charString)
1410
1411    hints = charString._hints
1412
1413    if hints.has_hint or hints.has_hintmask:
1414      self._css.add(charString)
1415
1416    if hints.status != 2:
1417      # Check from last_check, make sure we didn't have any operators.
1418      for i in range(hints.last_checked, len(charString.program) - 1):
1419        if isinstance(charString.program[i], str):
1420          hints.status = 2
1421          break;
1422        else:
1423          hints.status = 1 # There's *something* here
1424      hints.last_checked = len(charString.program)
1425
1426    if old_hints:
1427      assert hints.__dict__ == old_hints.__dict__
1428
1429  def op_callsubr(self, index):
1430    subr = self.localSubrs[self.operandStack[-1]+self.localBias]
1431    psCharStrings.SimpleT2Decompiler.op_callsubr(self, index)
1432    self.processSubr(index, subr)
1433
1434  def op_callgsubr(self, index):
1435    subr = self.globalSubrs[self.operandStack[-1]+self.globalBias]
1436    psCharStrings.SimpleT2Decompiler.op_callgsubr(self, index)
1437    self.processSubr(index, subr)
1438
1439  def op_hstem(self, index):
1440    psCharStrings.SimpleT2Decompiler.op_hstem(self, index)
1441    self.processHint(index)
1442  def op_vstem(self, index):
1443    psCharStrings.SimpleT2Decompiler.op_vstem(self, index)
1444    self.processHint(index)
1445  def op_hstemhm(self, index):
1446    psCharStrings.SimpleT2Decompiler.op_hstemhm(self, index)
1447    self.processHint(index)
1448  def op_vstemhm(self, index):
1449    psCharStrings.SimpleT2Decompiler.op_vstemhm(self, index)
1450    self.processHint(index)
1451  def op_hintmask(self, index):
1452    psCharStrings.SimpleT2Decompiler.op_hintmask(self, index)
1453    self.processHintmask(index)
1454  def op_cntrmask(self, index):
1455    psCharStrings.SimpleT2Decompiler.op_cntrmask(self, index)
1456    self.processHintmask(index)
1457
1458  def processHintmask(self, index):
1459    cs = self.callingStack[-1]
1460    hints = cs._hints
1461    hints.has_hintmask = True
1462    if hints.status != 2 and hints.has_hint:
1463      # Check from last_check, see if we may be an implicit vstem
1464      for i in range(hints.last_checked, index - 1):
1465        if isinstance(cs.program[i], str):
1466          hints.status = 2
1467          break;
1468      if hints.status != 2:
1469        # We are an implicit vstem
1470        hints.last_hint = index + 1
1471        hints.status = 0
1472    hints.last_checked = index + 1
1473
1474  def processHint(self, index):
1475    cs = self.callingStack[-1]
1476    hints = cs._hints
1477    hints.has_hint = True
1478    hints.last_hint = index
1479    hints.last_checked = index
1480
1481  def processSubr(self, index, subr):
1482    cs = self.callingStack[-1]
1483    hints = cs._hints
1484    subr_hints = subr._hints
1485
1486    if subr_hints.has_hint:
1487      if hints.status != 2:
1488        hints.has_hint = True
1489        hints.last_checked = index
1490        hints.status = subr_hints.status
1491        # Decide where to chop off from
1492        if subr_hints.status == 0:
1493          hints.last_hint = index
1494        else:
1495          hints.last_hint = index - 2 # Leave the subr call in
1496      else:
1497        # In my understanding, this is a font bug.  Ie. it has hint stems
1498        # *after* path construction.  I've seen this in widespread fonts.
1499        # Best to ignore the hints I suppose...
1500        pass
1501        #assert 0
1502    else:
1503      hints.status = max(hints.status, subr_hints.status)
1504      if hints.status != 2:
1505        # Check from last_check, make sure we didn't have
1506        # any operators.
1507        for i in range(hints.last_checked, index - 1):
1508          if isinstance(cs.program[i], str):
1509            hints.status = 2
1510            break;
1511        hints.last_checked = index
1512      if hints.status != 2:
1513        # Decide where to chop off from
1514        if subr_hints.status == 0:
1515          hints.last_hint = index
1516        else:
1517          hints.last_hint = index - 2 # Leave the subr call in
1518
1519@_add_method(ttLib.getTableClass('CFF '))
1520def prune_post_subset(self, options):
1521  cff = self.cff
1522  for fontname in cff.keys():
1523    font = cff[fontname]
1524    cs = font.CharStrings
1525
1526
1527    #
1528    # Drop unused FontDictionaries
1529    #
1530    if hasattr(font, "FDSelect"):
1531      sel = font.FDSelect
1532      indices = _uniq_sort(sel.gidArray)
1533      sel.gidArray = [indices.index (ss) for ss in sel.gidArray]
1534      arr = font.FDArray
1535      arr.items = [arr[i] for i in indices]
1536      arr.count = len(arr.items)
1537      del arr.file, arr.offsets
1538
1539
1540    #
1541    # Drop hints if not needed
1542    #
1543    if not options.hinting:
1544
1545      #
1546      # This can be tricky, but doesn't have to.  What we do is:
1547      #
1548      # - Run all used glyph charstrings and recurse into subroutines,
1549      # - For each charstring (including subroutines), if it has any
1550      #   of the hint stem operators, we mark it as such.  Upon returning,
1551      #   for each charstring we note all the subroutine calls it makes
1552      #   that (recursively) contain a stem,
1553      # - Dropping hinting then consists of the following two ops:
1554      #   * Drop the piece of the program in each charstring before the
1555      #     last call to a stem op or a stem-calling subroutine,
1556      #   * Drop all hintmask operations.
1557      # - It's trickier... A hintmask right after hints and a few numbers
1558      #   will act as an implicit vstemhm.  As such, we track whether
1559      #   we have seen any non-hint operators so far and do the right
1560      #   thing, recursively...  Good luck understanding that :(
1561      #
1562      css = set()
1563      for g in font.charset:
1564        c,sel = cs.getItemAndSelector(g)
1565        # Make sure it's decompiled.  We want our "decompiler" to walk
1566        # the program, not the bytecode.
1567        c.draw(basePen.NullPen())
1568        subrs = getattr(c.private, "Subrs", [])
1569        decompiler = _DehintingT2Decompiler(css, subrs, c.globalSubrs)
1570        decompiler.execute(c)
1571      for charstring in css:
1572        charstring.drop_hints()
1573
1574      # Drop font-wide hinting values
1575      all_privs = []
1576      if hasattr(font, 'FDSelect'):
1577        all_privs.extend(fd.Private for fd in font.FDArray)
1578      else:
1579        all_privs.append(font.Private)
1580      for priv in all_privs:
1581        for k in ['BlueValues', 'OtherBlues', 'FamilyBlues', 'FamilyOtherBlues',
1582                  'BlueScale', 'BlueShift', 'BlueFuzz',
1583                  'StemSnapH', 'StemSnapV', 'StdHW', 'StdVW']:
1584          if hasattr(priv, k):
1585            setattr(priv, k, None)
1586
1587
1588    #
1589    # Renumber subroutines to remove unused ones
1590    #
1591
1592    # Mark all used subroutines
1593    for g in font.charset:
1594      c,sel = cs.getItemAndSelector(g)
1595      subrs = getattr(c.private, "Subrs", [])
1596      decompiler = _MarkingT2Decompiler(subrs, c.globalSubrs)
1597      decompiler.execute(c)
1598
1599    all_subrs = [font.GlobalSubrs]
1600    if hasattr(font, 'FDSelect'):
1601      all_subrs.extend(fd.Private.Subrs for fd in font.FDArray if hasattr(fd.Private, 'Subrs') and fd.Private.Subrs)
1602    elif hasattr(font.Private, 'Subrs') and font.Private.Subrs:
1603      all_subrs.append(font.Private.Subrs)
1604
1605    subrs = set(subrs) # Remove duplicates
1606
1607    # Prepare
1608    for subrs in all_subrs:
1609      if not hasattr(subrs, '_used'):
1610        subrs._used = set()
1611      subrs._used = _uniq_sort(subrs._used)
1612      subrs._old_bias = psCharStrings.calcSubrBias(subrs)
1613      subrs._new_bias = psCharStrings.calcSubrBias(subrs._used)
1614
1615    # Renumber glyph charstrings
1616    for g in font.charset:
1617      c,sel = cs.getItemAndSelector(g)
1618      subrs = getattr(c.private, "Subrs", [])
1619      c.subset_subroutines (subrs, font.GlobalSubrs)
1620
1621    # Renumber subroutines themselves
1622    for subrs in all_subrs:
1623
1624      if subrs == font.GlobalSubrs:
1625        if not hasattr(font, 'FDSelect') and hasattr(font.Private, 'Subrs'):
1626          local_subrs = font.Private.Subrs
1627        else:
1628          local_subrs = []
1629      else:
1630        local_subrs = subrs
1631
1632      subrs.items = [subrs.items[i] for i in subrs._used]
1633      subrs.count = len(subrs.items)
1634      del subrs.file
1635      if hasattr(subrs, 'offsets'):
1636        del subrs.offsets
1637
1638      for i in range (subrs.count):
1639        subrs[i].subset_subroutines (local_subrs, font.GlobalSubrs)
1640
1641    # Cleanup
1642    for subrs in all_subrs:
1643      del subrs._used, subrs._old_bias, subrs._new_bias
1644
1645  return True
1646
1647@_add_method(ttLib.getTableClass('cmap'))
1648def closure_glyphs(self, s):
1649  tables = [t for t in self.tables
1650            if t.platformID == 3 and t.platEncID in [1, 10]]
1651  for u in s.unicodes_requested:
1652    found = False
1653    for table in tables:
1654      if u in table.cmap:
1655        s.glyphs.add(table.cmap[u])
1656        found = True
1657        break
1658    if not found:
1659      s.log("No glyph for Unicode value %s; skipping." % u)
1660
1661@_add_method(ttLib.getTableClass('cmap'))
1662def prune_pre_subset(self, options):
1663  if not options.legacy_cmap:
1664    # Drop non-Unicode / non-Symbol cmaps
1665    self.tables = [t for t in self.tables
1666                   if t.platformID == 3 and t.platEncID in [0, 1, 10]]
1667  if not options.symbol_cmap:
1668    self.tables = [t for t in self.tables
1669                   if t.platformID == 3 and t.platEncID in [1, 10]]
1670  # TODO(behdad) Only keep one subtable?
1671  # For now, drop format=0 which can't be subset_glyphs easily?
1672  self.tables = [t for t in self.tables if t.format != 0]
1673  self.numSubTables = len(self.tables)
1674  return True # Required table
1675
1676@_add_method(ttLib.getTableClass('cmap'))
1677def subset_glyphs(self, s):
1678  s.glyphs = s.glyphs_cmaped
1679  for t in self.tables:
1680    # For reasons I don't understand I need this here
1681    # to force decompilation of the cmap format 14.
1682    try:
1683      getattr(t, "asdf")
1684    except AttributeError:
1685      pass
1686    if t.format == 14:
1687      # TODO(behdad) XXX We drop all the default-UVS mappings(g==None).
1688      t.uvsDict = dict((v,[(u,g) for u,g in l if g in s.glyphs])
1689                       for v,l in t.uvsDict.items())
1690      t.uvsDict = dict((v,l) for v,l in t.uvsDict.items() if l)
1691    else:
1692      t.cmap = dict((u,g) for u,g in t.cmap.items()
1693                    if g in s.glyphs_requested or u in s.unicodes_requested)
1694  self.tables = [t for t in self.tables
1695                 if (t.cmap if t.format != 14 else t.uvsDict)]
1696  self.numSubTables = len(self.tables)
1697  # TODO(behdad) Convert formats when needed.
1698  # In particular, if we have a format=12 without non-BMP
1699  # characters, either drop format=12 one or convert it
1700  # to format=4 if there's not one.
1701  return True # Required table
1702
1703@_add_method(ttLib.getTableClass('name'))
1704def prune_pre_subset(self, options):
1705  if '*' not in options.name_IDs:
1706    self.names = [n for n in self.names if n.nameID in options.name_IDs]
1707  if not options.name_legacy:
1708    self.names = [n for n in self.names
1709                  if n.platformID == 3 and n.platEncID == 1]
1710  if '*' not in options.name_languages:
1711    self.names = [n for n in self.names if n.langID in options.name_languages]
1712  return True  # Required table
1713
1714
1715# TODO(behdad) OS/2 ulUnicodeRange / ulCodePageRange?
1716# TODO(behdad) Drop AAT tables.
1717# TODO(behdad) Drop unneeded GSUB/GPOS Script/LangSys entries.
1718# TODO(behdad) Drop empty GSUB/GPOS, and GDEF if no GSUB/GPOS left
1719# TODO(behdad) Drop GDEF subitems if unused by lookups
1720# TODO(behdad) Avoid recursing too much (in GSUB/GPOS and in CFF)
1721# TODO(behdad) Text direction considerations.
1722# TODO(behdad) Text script / language considerations.
1723# TODO(behdad) Optionally drop 'kern' table if GPOS available
1724
1725class Options(object):
1726
1727  class UnknownOptionError(Exception):
1728    pass
1729
1730  _drop_tables_default = ['BASE', 'JSTF', 'DSIG', 'EBDT', 'EBLC', 'EBSC', 'SVG ',
1731                          'PCLT', 'LTSH']
1732  _drop_tables_default += ['Feat', 'Glat', 'Gloc', 'Silf', 'Sill']  # Graphite
1733  _drop_tables_default += ['CBLC', 'CBDT', 'sbix', 'COLR', 'CPAL']  # Color
1734  _no_subset_tables_default = ['gasp', 'head', 'hhea', 'maxp', 'vhea', 'OS/2',
1735                               'loca', 'name', 'cvt ', 'fpgm', 'prep']
1736  _hinting_tables_default = ['cvt ', 'fpgm', 'prep', 'hdmx', 'VDMX']
1737
1738  # Based on HarfBuzz shapers
1739  _layout_features_groups = {
1740    # Default shaper
1741    'common': ['ccmp', 'liga', 'locl', 'mark', 'mkmk', 'rlig'],
1742    'horizontal': ['calt', 'clig', 'curs', 'kern', 'rclt'],
1743    'vertical':  ['valt', 'vert', 'vkrn', 'vpal', 'vrt2'],
1744    'ltr': ['ltra', 'ltrm'],
1745    'rtl': ['rtla', 'rtlm'],
1746    # Complex shapers
1747    'arabic': ['init', 'medi', 'fina', 'isol', 'med2', 'fin2', 'fin3',
1748               'cswh', 'mset'],
1749    'hangul': ['ljmo', 'vjmo', 'tjmo'],
1750    'tibetan': ['abvs', 'blws', 'abvm', 'blwm'],
1751    'indic': ['nukt', 'akhn', 'rphf', 'rkrf', 'pref', 'blwf', 'half',
1752              'abvf', 'pstf', 'cfar', 'vatu', 'cjct', 'init', 'pres',
1753              'abvs', 'blws', 'psts', 'haln', 'dist', 'abvm', 'blwm'],
1754  }
1755  _layout_features_default = _uniq_sort(sum(
1756      iter(_layout_features_groups.values()), []))
1757
1758  drop_tables = _drop_tables_default
1759  no_subset_tables = _no_subset_tables_default
1760  hinting_tables = _hinting_tables_default
1761  layout_features = _layout_features_default
1762  hinting = True
1763  glyph_names = False
1764  legacy_cmap = False
1765  symbol_cmap = False
1766  name_IDs = [1, 2]  # Family and Style
1767  name_legacy = False
1768  name_languages = [0x0409]  # English
1769  notdef_glyph = True # gid0 for TrueType / .notdef for CFF
1770  notdef_outline = False # No need for notdef to have an outline really
1771  recommended_glyphs = False  # gid1, gid2, gid3 for TrueType
1772  recalc_bounds = False # Recalculate font bounding boxes
1773  canonical_order = False # Order tables as recommended
1774  flavor = None # May be 'woff'
1775
1776  def __init__(self, **kwargs):
1777
1778    self.set(**kwargs)
1779
1780  def set(self, **kwargs):
1781    for k,v in kwargs.items():
1782      if not hasattr(self, k):
1783        raise self.UnknownOptionError("Unknown option '%s'" % k)
1784      setattr(self, k, v)
1785
1786  def parse_opts(self, argv, ignore_unknown=False):
1787    ret = []
1788    opts = {}
1789    for a in argv:
1790      orig_a = a
1791      if not a.startswith('--'):
1792        ret.append(a)
1793        continue
1794      a = a[2:]
1795      i = a.find('=')
1796      op = '='
1797      if i == -1:
1798        if a.startswith("no-"):
1799          k = a[3:]
1800          v = False
1801        else:
1802          k = a
1803          v = True
1804      else:
1805        k = a[:i]
1806        if k[-1] in "-+":
1807          op = k[-1]+'='  # Ops is '-=' or '+=' now.
1808          k = k[:-1]
1809        v = a[i+1:]
1810      k = k.replace('-', '_')
1811      if not hasattr(self, k):
1812        if ignore_unknown is True or k in ignore_unknown:
1813          ret.append(orig_a)
1814          continue
1815        else:
1816          raise self.UnknownOptionError("Unknown option '%s'" % a)
1817
1818      ov = getattr(self, k)
1819      if isinstance(ov, bool):
1820        v = bool(v)
1821      elif isinstance(ov, int):
1822        v = int(v)
1823      elif isinstance(ov, list):
1824        vv = v.split(',')
1825        if vv == ['']:
1826          vv = []
1827        vv = [int(x, 0) if len(x) and x[0] in "0123456789" else x for x in vv]
1828        if op == '=':
1829          v = vv
1830        elif op == '+=':
1831          v = ov
1832          v.extend(vv)
1833        elif op == '-=':
1834          v = ov
1835          for x in vv:
1836            if x in v:
1837              v.remove(x)
1838        else:
1839          assert False
1840
1841      opts[k] = v
1842    self.set(**opts)
1843
1844    return ret
1845
1846
1847class Subsetter(object):
1848
1849  def __init__(self, options=None, log=None):
1850
1851    if not log:
1852      log = Logger()
1853    if not options:
1854      options = Options()
1855
1856    self.options = options
1857    self.log = log
1858    self.unicodes_requested = set()
1859    self.glyphs_requested = set()
1860    self.glyphs = set()
1861
1862  def populate(self, glyphs=[], unicodes=[], text=""):
1863    self.unicodes_requested.update(unicodes)
1864    if isinstance(text, bytes):
1865      text = text.decode("utf8")
1866    for u in text:
1867      self.unicodes_requested.add(ord(u))
1868    self.glyphs_requested.update(glyphs)
1869    self.glyphs.update(glyphs)
1870
1871  def _prune_pre_subset(self, font):
1872
1873    for tag in font.keys():
1874      if tag == 'GlyphOrder': continue
1875
1876      if(tag in self.options.drop_tables or
1877         (tag in self.options.hinting_tables and not self.options.hinting)):
1878        self.log(tag, "dropped")
1879        del font[tag]
1880        continue
1881
1882      clazz = ttLib.getTableClass(tag)
1883
1884      if hasattr(clazz, 'prune_pre_subset'):
1885        table = font[tag]
1886        self.log.lapse("load '%s'" % tag)
1887        retain = table.prune_pre_subset(self.options)
1888        self.log.lapse("prune  '%s'" % tag)
1889        if not retain:
1890          self.log(tag, "pruned to empty; dropped")
1891          del font[tag]
1892          continue
1893        else:
1894          self.log(tag, "pruned")
1895
1896  def _closure_glyphs(self, font):
1897
1898    realGlyphs = set(font.getGlyphOrder())
1899
1900    self.glyphs = self.glyphs_requested.copy()
1901
1902    if 'cmap' in font:
1903      font['cmap'].closure_glyphs(self)
1904      self.glyphs.intersection_update(realGlyphs)
1905    self.glyphs_cmaped = self.glyphs
1906
1907    if self.options.notdef_glyph:
1908      if 'glyf' in font:
1909        self.glyphs.add(font.getGlyphName(0))
1910        self.log("Added gid0 to subset")
1911      else:
1912        self.glyphs.add('.notdef')
1913        self.log("Added .notdef to subset")
1914    if self.options.recommended_glyphs:
1915      if 'glyf' in font:
1916        for i in min(range(4), len(font.getGlyphOrder())):
1917          self.glyphs.add(font.getGlyphName(i))
1918        self.log("Added first four glyphs to subset")
1919
1920    if 'GSUB' in font:
1921      self.log("Closing glyph list over 'GSUB': %d glyphs before" %
1922                len(self.glyphs))
1923      self.log.glyphs(self.glyphs, font=font)
1924      font['GSUB'].closure_glyphs(self)
1925      self.glyphs.intersection_update(realGlyphs)
1926      self.log("Closed  glyph list over 'GSUB': %d glyphs after" %
1927                len(self.glyphs))
1928      self.log.glyphs(self.glyphs, font=font)
1929      self.log.lapse("close glyph list over 'GSUB'")
1930    self.glyphs_gsubed = self.glyphs.copy()
1931
1932    if 'glyf' in font:
1933      self.log("Closing glyph list over 'glyf': %d glyphs before" %
1934                len(self.glyphs))
1935      self.log.glyphs(self.glyphs, font=font)
1936      font['glyf'].closure_glyphs(self)
1937      self.glyphs.intersection_update(realGlyphs)
1938      self.log("Closed  glyph list over 'glyf': %d glyphs after" %
1939                len(self.glyphs))
1940      self.log.glyphs(self.glyphs, font=font)
1941      self.log.lapse("close glyph list over 'glyf'")
1942    self.glyphs_glyfed = self.glyphs.copy()
1943
1944    self.glyphs_all = self.glyphs.copy()
1945
1946    self.log("Retaining %d glyphs: " % len(self.glyphs_all))
1947
1948    del self.glyphs
1949
1950
1951  def _subset_glyphs(self, font):
1952    for tag in font.keys():
1953      if tag == 'GlyphOrder': continue
1954      clazz = ttLib.getTableClass(tag)
1955
1956      if tag in self.options.no_subset_tables:
1957        self.log(tag, "subsetting not needed")
1958      elif hasattr(clazz, 'subset_glyphs'):
1959        table = font[tag]
1960        self.glyphs = self.glyphs_all
1961        retain = table.subset_glyphs(self)
1962        del self.glyphs
1963        self.log.lapse("subset '%s'" % tag)
1964        if not retain:
1965          self.log(tag, "subsetted to empty; dropped")
1966          del font[tag]
1967        else:
1968          self.log(tag, "subsetted")
1969      else:
1970        self.log(tag, "NOT subset; don't know how to subset; dropped")
1971        del font[tag]
1972
1973    glyphOrder = font.getGlyphOrder()
1974    glyphOrder = [g for g in glyphOrder if g in self.glyphs_all]
1975    font.setGlyphOrder(glyphOrder)
1976    font._buildReverseGlyphOrderDict()
1977    self.log.lapse("subset GlyphOrder")
1978
1979  def _prune_post_subset(self, font):
1980    for tag in font.keys():
1981      if tag == 'GlyphOrder': continue
1982      clazz = ttLib.getTableClass(tag)
1983      if hasattr(clazz, 'prune_post_subset'):
1984        table = font[tag]
1985        retain = table.prune_post_subset(self.options)
1986        self.log.lapse("prune  '%s'" % tag)
1987        if not retain:
1988          self.log(tag, "pruned to empty; dropped")
1989          del font[tag]
1990        else:
1991          self.log(tag, "pruned")
1992
1993  def subset(self, font):
1994
1995    self._prune_pre_subset(font)
1996    self._closure_glyphs(font)
1997    self._subset_glyphs(font)
1998    self._prune_post_subset(font)
1999
2000
2001class Logger(object):
2002
2003  def __init__(self, verbose=False, xml=False, timing=False):
2004    self.verbose = verbose
2005    self.xml = xml
2006    self.timing = timing
2007    self.last_time = self.start_time = time.time()
2008
2009  def parse_opts(self, argv):
2010    argv = argv[:]
2011    for v in ['verbose', 'xml', 'timing']:
2012      if "--"+v in argv:
2013        setattr(self, v, True)
2014        argv.remove("--"+v)
2015    return argv
2016
2017  def __call__(self, *things):
2018    if not self.verbose:
2019      return
2020    print(' '.join(str(x) for x in things))
2021
2022  def lapse(self, *things):
2023    if not self.timing:
2024      return
2025    new_time = time.time()
2026    print("Took %0.3fs to %s" %(new_time - self.last_time,
2027                                 ' '.join(str(x) for x in things)))
2028    self.last_time = new_time
2029
2030  def glyphs(self, glyphs, font=None):
2031    if not self.verbose:
2032      return
2033    self("Names: ", sorted(glyphs))
2034    if font:
2035      reverseGlyphMap = font.getReverseGlyphMap()
2036      self("Gids : ", sorted(reverseGlyphMap[g] for g in glyphs))
2037
2038  def font(self, font, file=sys.stdout):
2039    if not self.xml:
2040      return
2041    from fontTools.misc import xmlWriter
2042    writer = xmlWriter.XMLWriter(file)
2043    font.disassembleInstructions = False  # Work around ttLib bug
2044    for tag in font.keys():
2045      writer.begintag(tag)
2046      writer.newline()
2047      font[tag].toXML(writer, font)
2048      writer.endtag(tag)
2049      writer.newline()
2050
2051
2052def load_font(fontFile,
2053              options,
2054              allowVID=False,
2055              checkChecksums=False,
2056              dontLoadGlyphNames=False):
2057
2058  font = ttLib.TTFont(fontFile,
2059                      allowVID=allowVID,
2060                      checkChecksums=checkChecksums,
2061                      recalcBBoxes=options.recalc_bounds)
2062
2063  # Hack:
2064  #
2065  # If we don't need glyph names, change 'post' class to not try to
2066  # load them.  It avoid lots of headache with broken fonts as well
2067  # as loading time.
2068  #
2069  # Ideally ttLib should provide a way to ask it to skip loading
2070  # glyph names.  But it currently doesn't provide such a thing.
2071  #
2072  if dontLoadGlyphNames:
2073    post = ttLib.getTableClass('post')
2074    saved = post.decode_format_2_0
2075    post.decode_format_2_0 = post.decode_format_3_0
2076    f = font['post']
2077    if f.formatType == 2.0:
2078      f.formatType = 3.0
2079    post.decode_format_2_0 = saved
2080
2081  return font
2082
2083def save_font(font, outfile, options):
2084  if options.flavor and not hasattr(font, 'flavor'):
2085    raise Exception("fonttools version does not support flavors.")
2086  font.flavor = options.flavor
2087  font.save(outfile, reorderTables=options.canonical_order)
2088
2089def main(args):
2090
2091  log = Logger()
2092  args = log.parse_opts(args)
2093
2094  options = Options()
2095  args = options.parse_opts(args, ignore_unknown=['text'])
2096
2097  if len(args) < 2:
2098    print("usage: pyftsubset font-file glyph... [--text=ABC]... [--option=value]...", file=sys.stderr)
2099    sys.exit(1)
2100
2101  fontfile = args[0]
2102  args = args[1:]
2103
2104  dontLoadGlyphNames =(not options.glyph_names and
2105         all(any(g.startswith(p)
2106             for p in ['gid', 'glyph', 'uni', 'U+'])
2107              for g in args))
2108
2109  font = load_font(fontfile, options, dontLoadGlyphNames=dontLoadGlyphNames)
2110  subsetter = Subsetter(options=options, log=log)
2111  log.lapse("load font")
2112
2113  names = font.getGlyphNames()
2114  log.lapse("loading glyph names")
2115
2116  glyphs = []
2117  unicodes = []
2118  text = ""
2119  for g in args:
2120    if g == '*':
2121      glyphs.extend(font.getGlyphOrder())
2122      continue
2123    if g in names:
2124      glyphs.append(g)
2125      continue
2126    if g.startswith('--text='):
2127      text += g[7:]
2128      continue
2129    if g.startswith('uni') or g.startswith('U+'):
2130      if g.startswith('uni') and len(g) > 3:
2131        g = g[3:]
2132      elif g.startswith('U+') and len(g) > 2:
2133        g = g[2:]
2134      u = int(g, 16)
2135      unicodes.append(u)
2136      continue
2137    if g.startswith('gid') or g.startswith('glyph'):
2138      if g.startswith('gid') and len(g) > 3:
2139        g = g[3:]
2140      elif g.startswith('glyph') and len(g) > 5:
2141        g = g[5:]
2142      try:
2143        glyphs.append(font.getGlyphName(int(g), requireReal=True))
2144      except ValueError:
2145        raise Exception("Invalid glyph identifier: %s" % g)
2146      continue
2147    raise Exception("Invalid glyph identifier: %s" % g)
2148  log.lapse("compile glyph list")
2149  log("Unicodes:", unicodes)
2150  log("Glyphs:", glyphs)
2151
2152  subsetter.populate(glyphs=glyphs, unicodes=unicodes, text=text)
2153  subsetter.subset(font)
2154
2155  outfile = fontfile + '.subset'
2156
2157  save_font (font, outfile, options)
2158  log.lapse("compile and save font")
2159
2160  log.last_time = log.start_time
2161  log.lapse("make one with everything(TOTAL TIME)")
2162
2163  if log.verbose:
2164    import os
2165    log("Input  font: %d bytes" % os.path.getsize(fontfile))
2166    log("Subset font: %d bytes" % os.path.getsize(outfile))
2167
2168  log.font(font)
2169
2170  font.close()
2171
2172
2173__all__ = [
2174  'Options',
2175  'Subsetter',
2176  'Logger',
2177  'load_font',
2178  'save_font',
2179  'main'
2180]
2181
2182if __name__ == '__main__':
2183  main(sys.argv[1:])
2184