subset.py revision 26560d238f096afcafb5076cb47061f1c2855ec2 
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
10import sys
11import struct
12import time
13import array
14
15from fontTools import ttLib
16from fontTools.ttLib.tables import otTables
17from fontTools.misc import psCharStrings
18from fontTools.pens import basePen
19
20
21def _add_method(*clazzes):
22  """Returns a decorator function that adds a new method to one or
23  more classes."""
24  def wrapper(method):
25    for clazz in clazzes:
26      assert clazz.__name__ != 'DefaultTable', 'Oops, table class not found.'
27      assert not hasattr(clazz, method.func_name), \
28          "Oops, class '%s' has method '%s'." % (clazz.__name__,
29                                                 method.func_name)
30      setattr(clazz, method.func_name, method)
31    return None
32  return wrapper
33
34def _uniq_sort(l):
35  return sorted(set(l))
36
37def _set_update(s, *others):
38  # Jython's set.update only takes one other argument.
39  # Emulate real set.update...
40  for other in others:
41    s.update(other)
42
43
44@_add_method(otTables.Coverage)
45def intersect(self, glyphs):
46  "Returns ascending list of matching coverage values."
47  return [i for i,g in enumerate(self.glyphs) if g in glyphs]
48
49@_add_method(otTables.Coverage)
50def intersect_glyphs(self, glyphs):
51  "Returns set of intersecting glyphs."
52  return set(g for g in self.glyphs if g in glyphs)
53
54@_add_method(otTables.Coverage)
55def subset(self, glyphs):
56  "Returns ascending list of remaining coverage values."
57  indices = self.intersect(glyphs)
58  self.glyphs = [g for g in self.glyphs if g in glyphs]
59  return indices
60
61@_add_method(otTables.Coverage)
62def remap(self, coverage_map):
63  "Remaps coverage."
64  self.glyphs = [self.glyphs[i] for i in coverage_map]
65
66@_add_method(otTables.ClassDef)
67def intersect(self, glyphs):
68  "Returns ascending list of matching class values."
69  return _uniq_sort(
70     ([0] if any(g not in self.classDefs for g in glyphs) else []) +
71      [v for g,v in self.classDefs.iteritems() if g in glyphs])
72
73@_add_method(otTables.ClassDef)
74def intersect_class(self, glyphs, klass):
75  "Returns set of glyphs matching class."
76  if klass == 0:
77    return set(g for g in glyphs if g not in self.classDefs)
78  return set(g for g,v in self.classDefs.iteritems()
79              if v == klass and g in glyphs)
80
81@_add_method(otTables.ClassDef)
82def subset(self, glyphs, remap=False):
83  "Returns ascending list of remaining classes."
84  self.classDefs = dict((g,v) for g,v in self.classDefs.iteritems() if g in glyphs)
85  # Note: while class 0 has the special meaning of "not matched",
86  # if no glyph will ever /not match/, we can optimize class 0 out too.
87  indices = _uniq_sort(
88     ([0] if any(g not in self.classDefs for g in glyphs) else []) +
89      self.classDefs.values())
90  if remap:
91    self.remap(indices)
92  return indices
93
94@_add_method(otTables.ClassDef)
95def remap(self, class_map):
96  "Remaps classes."
97  self.classDefs = dict((g,class_map.index(v))
98                         for g,v in self.classDefs.iteritems())
99
100@_add_method(otTables.SingleSubst)
101def closure_glyphs(self, s, cur_glyphs=None):
102  if cur_glyphs == None: cur_glyphs = s.glyphs
103  if self.Format in [1, 2]:
104    s.glyphs.update(v for g,v in self.mapping.iteritems() if g in cur_glyphs)
105  else:
106    assert 0, "unknown format: %s" % self.Format
107
108@_add_method(otTables.SingleSubst)
109def subset_glyphs(self, s):
110  if self.Format in [1, 2]:
111    self.mapping = dict((g,v) for g,v in self.mapping.iteritems()
112                        if g in s.glyphs and v in s.glyphs)
113    return bool(self.mapping)
114  else:
115    assert 0, "unknown format: %s" % self.Format
116
117@_add_method(otTables.MultipleSubst)
118def closure_glyphs(self, s, cur_glyphs=None):
119  if cur_glyphs == None: cur_glyphs = s.glyphs
120  if self.Format == 1:
121    indices = self.Coverage.intersect(cur_glyphs)
122    _set_update(s.glyphs, *(self.Sequence[i].Substitute for i in indices))
123  else:
124    assert 0, "unknown format: %s" % self.Format
125
126@_add_method(otTables.MultipleSubst)
127def subset_glyphs(self, s):
128  if self.Format == 1:
129    indices = self.Coverage.subset(s.glyphs)
130    self.Sequence = [self.Sequence[i] for i in indices]
131    # Now drop rules generating glyphs we don't want
132    indices = [i for i,seq in enumerate(self.Sequence)
133         if all(sub in s.glyphs for sub in seq.Substitute)]
134    self.Sequence = [self.Sequence[i] for i in indices]
135    self.Coverage.remap(indices)
136    self.SequenceCount = len(self.Sequence)
137    return bool(self.SequenceCount)
138  else:
139    assert 0, "unknown format: %s" % self.Format
140
141@_add_method(otTables.AlternateSubst)
142def closure_glyphs(self, s, cur_glyphs=None):
143  if cur_glyphs == None: cur_glyphs = s.glyphs
144  if self.Format == 1:
145    _set_update(s.glyphs, *(vlist for g,vlist in self.alternates.iteritems()
146                            if g in cur_glyphs))
147  else:
148    assert 0, "unknown format: %s" % self.Format
149
150@_add_method(otTables.AlternateSubst)
151def subset_glyphs(self, s):
152  if self.Format == 1:
153    self.alternates = dict((g,vlist)
154                           for g,vlist in self.alternates.iteritems()
155                           if g in s.glyphs and
156                              all(v in s.glyphs for v in vlist))
157    return bool(self.alternates)
158  else:
159    assert 0, "unknown format: %s" % self.Format
160
161@_add_method(otTables.LigatureSubst)
162def closure_glyphs(self, s, cur_glyphs=None):
163  if cur_glyphs == None: cur_glyphs = s.glyphs
164  if self.Format == 1:
165    _set_update(s.glyphs, *([seq.LigGlyph for seq in seqs
166                             if all(c in s.glyphs for c in seq.Component)]
167                            for g,seqs in self.ligatures.iteritems()
168                            if g in cur_glyphs))
169  else:
170    assert 0, "unknown format: %s" % self.Format
171
172@_add_method(otTables.LigatureSubst)
173def subset_glyphs(self, s):
174  if self.Format == 1:
175    self.ligatures = dict((g,v) for g,v in self.ligatures.iteritems()
176                          if g in s.glyphs)
177    self.ligatures = dict((g,[seq for seq in seqs
178                              if seq.LigGlyph in s.glyphs and
179                                 all(c in s.glyphs for c in seq.Component)])
180                           for g,seqs in self.ligatures.iteritems())
181    self.ligatures = dict((g,v) for g,v in self.ligatures.iteritems() if v)
182    return bool(self.ligatures)
183  else:
184    assert 0, "unknown format: %s" % self.Format
185
186@_add_method(otTables.ReverseChainSingleSubst)
187def closure_glyphs(self, s, cur_glyphs=None):
188  if cur_glyphs == None: cur_glyphs = s.glyphs
189  if self.Format == 1:
190    indices = self.Coverage.intersect(cur_glyphs)
191    if(not indices or
192        not all(c.intersect(s.glyphs)
193                 for c in self.LookAheadCoverage + self.BacktrackCoverage)):
194      return
195    s.glyphs.update(self.Substitute[i] for i in indices)
196  else:
197    assert 0, "unknown format: %s" % self.Format
198
199@_add_method(otTables.ReverseChainSingleSubst)
200def subset_glyphs(self, s):
201  if self.Format == 1:
202    indices = self.Coverage.subset(s.glyphs)
203    self.Substitute = [self.Substitute[i] for i in indices]
204    # Now drop rules generating glyphs we don't want
205    indices = [i for i,sub in enumerate(self.Substitute)
206         if sub in s.glyphs]
207    self.Substitute = [self.Substitute[i] for i in indices]
208    self.Coverage.remap(indices)
209    self.GlyphCount = len(self.Substitute)
210    return bool(self.GlyphCount and
211                 all(c.subset(s.glyphs)
212                      for c in self.LookAheadCoverage+self.BacktrackCoverage))
213  else:
214    assert 0, "unknown format: %s" % self.Format
215
216@_add_method(otTables.SinglePos)
217def subset_glyphs(self, s):
218  if self.Format == 1:
219    return len(self.Coverage.subset(s.glyphs))
220  elif self.Format == 2:
221    indices = self.Coverage.subset(s.glyphs)
222    self.Value = [self.Value[i] for i in indices]
223    self.ValueCount = len(self.Value)
224    return bool(self.ValueCount)
225  else:
226    assert 0, "unknown format: %s" % self.Format
227
228@_add_method(otTables.SinglePos)
229def prune_post_subset(self, options):
230  if not options.hinting:
231    # Drop device tables
232    self.ValueFormat &= ~0x00F0
233  return True
234
235@_add_method(otTables.PairPos)
236def subset_glyphs(self, s):
237  if self.Format == 1:
238    indices = self.Coverage.subset(s.glyphs)
239    self.PairSet = [self.PairSet[i] for i in indices]
240    for p in self.PairSet:
241      p.PairValueRecord = [r for r in p.PairValueRecord
242                           if r.SecondGlyph in s.glyphs]
243      p.PairValueCount = len(p.PairValueRecord)
244    self.PairSet = [p for p in self.PairSet if p.PairValueCount]
245    self.PairSetCount = len(self.PairSet)
246    return bool(self.PairSetCount)
247  elif self.Format == 2:
248    class1_map = self.ClassDef1.subset(s.glyphs, remap=True)
249    class2_map = self.ClassDef2.subset(s.glyphs, remap=True)
250    self.Class1Record = [self.Class1Record[i] for i in class1_map]
251    for c in self.Class1Record:
252      c.Class2Record = [c.Class2Record[i] for i in class2_map]
253    self.Class1Count = len(class1_map)
254    self.Class2Count = len(class2_map)
255    return bool(self.Class1Count and
256                 self.Class2Count and
257                 self.Coverage.subset(s.glyphs))
258  else:
259    assert 0, "unknown format: %s" % self.Format
260
261@_add_method(otTables.PairPos)
262def prune_post_subset(self, options):
263  if not options.hinting:
264    # Drop device tables
265    self.ValueFormat1 &= ~0x00F0
266    self.ValueFormat2 &= ~0x00F0
267  return True
268
269@_add_method(otTables.CursivePos)
270def subset_glyphs(self, s):
271  if self.Format == 1:
272    indices = self.Coverage.subset(s.glyphs)
273    self.EntryExitRecord = [self.EntryExitRecord[i] for i in indices]
274    self.EntryExitCount = len(self.EntryExitRecord)
275    return bool(self.EntryExitCount)
276  else:
277    assert 0, "unknown format: %s" % self.Format
278
279@_add_method(otTables.Anchor)
280def prune_hints(self):
281  # Drop device tables / contour anchor point
282  self.Format = 1
283
284@_add_method(otTables.CursivePos)
285def prune_post_subset(self, options):
286  if not options.hinting:
287    for rec in self.EntryExitRecord:
288      if rec.EntryAnchor: rec.EntryAnchor.prune_hints()
289      if rec.ExitAnchor: rec.ExitAnchor.prune_hints()
290  return True
291
292@_add_method(otTables.MarkBasePos)
293def subset_glyphs(self, s):
294  if self.Format == 1:
295    mark_indices = self.MarkCoverage.subset(s.glyphs)
296    self.MarkArray.MarkRecord = [self.MarkArray.MarkRecord[i]
297                                 for i in mark_indices]
298    self.MarkArray.MarkCount = len(self.MarkArray.MarkRecord)
299    base_indices = self.BaseCoverage.subset(s.glyphs)
300    self.BaseArray.BaseRecord = [self.BaseArray.BaseRecord[i]
301                                 for i in base_indices]
302    self.BaseArray.BaseCount = len(self.BaseArray.BaseRecord)
303    # Prune empty classes
304    class_indices = _uniq_sort(v.Class for v in self.MarkArray.MarkRecord)
305    self.ClassCount = len(class_indices)
306    for m in self.MarkArray.MarkRecord:
307      m.Class = class_indices.index(m.Class)
308    for b in self.BaseArray.BaseRecord:
309      b.BaseAnchor = [b.BaseAnchor[i] for i in class_indices]
310    return bool(self.ClassCount and
311                 self.MarkArray.MarkCount and
312                 self.BaseArray.BaseCount)
313  else:
314    assert 0, "unknown format: %s" % self.Format
315
316@_add_method(otTables.MarkBasePos)
317def prune_post_subset(self, options):
318    if not options.hinting:
319      for m in self.MarkArray.MarkRecord:
320        if m.MarkAnchor:
321          m.MarkAnchor.prune_hints()
322      for b in self.BaseArray.BaseRecord:
323        for a in b.BaseAnchor:
324          if a:
325            a.prune_hints()
326    return True
327
328@_add_method(otTables.MarkLigPos)
329def subset_glyphs(self, s):
330  if self.Format == 1:
331    mark_indices = self.MarkCoverage.subset(s.glyphs)
332    self.MarkArray.MarkRecord = [self.MarkArray.MarkRecord[i]
333                                 for i in mark_indices]
334    self.MarkArray.MarkCount = len(self.MarkArray.MarkRecord)
335    ligature_indices = self.LigatureCoverage.subset(s.glyphs)
336    self.LigatureArray.LigatureAttach = [self.LigatureArray.LigatureAttach[i]
337                                         for i in ligature_indices]
338    self.LigatureArray.LigatureCount = len(self.LigatureArray.LigatureAttach)
339    # Prune empty classes
340    class_indices = _uniq_sort(v.Class for v in self.MarkArray.MarkRecord)
341    self.ClassCount = len(class_indices)
342    for m in self.MarkArray.MarkRecord:
343      m.Class = class_indices.index(m.Class)
344    for l in self.LigatureArray.LigatureAttach:
345      for c in l.ComponentRecord:
346        c.LigatureAnchor = [c.LigatureAnchor[i] for i in class_indices]
347    return bool(self.ClassCount and
348                 self.MarkArray.MarkCount and
349                 self.LigatureArray.LigatureCount)
350  else:
351    assert 0, "unknown format: %s" % self.Format
352
353@_add_method(otTables.MarkLigPos)
354def prune_post_subset(self, options):
355    if not options.hinting:
356      for m in self.MarkArray.MarkRecord:
357        if m.MarkAnchor:
358          m.MarkAnchor.prune_hints()
359      for l in self.LigatureArray.LigatureAttach:
360        for c in l.ComponentRecord:
361          for a in c.LigatureAnchor:
362            if a:
363              a.prune_hints()
364    return True
365
366@_add_method(otTables.MarkMarkPos)
367def subset_glyphs(self, s):
368  if self.Format == 1:
369    mark1_indices = self.Mark1Coverage.subset(s.glyphs)
370    self.Mark1Array.MarkRecord = [self.Mark1Array.MarkRecord[i]
371                                  for i in mark1_indices]
372    self.Mark1Array.MarkCount = len(self.Mark1Array.MarkRecord)
373    mark2_indices = self.Mark2Coverage.subset(s.glyphs)
374    self.Mark2Array.Mark2Record = [self.Mark2Array.Mark2Record[i]
375                                   for i in mark2_indices]
376    self.Mark2Array.MarkCount = len(self.Mark2Array.Mark2Record)
377    # Prune empty classes
378    class_indices = _uniq_sort(v.Class for v in self.Mark1Array.MarkRecord)
379    self.ClassCount = len(class_indices)
380    for m in self.Mark1Array.MarkRecord:
381      m.Class = class_indices.index(m.Class)
382    for b in self.Mark2Array.Mark2Record:
383      b.Mark2Anchor = [b.Mark2Anchor[i] for i in class_indices]
384    return bool(self.ClassCount and
385                 self.Mark1Array.MarkCount and
386                 self.Mark2Array.MarkCount)
387  else:
388    assert 0, "unknown format: %s" % self.Format
389
390@_add_method(otTables.MarkMarkPos)
391def prune_post_subset(self, options):
392    if not options.hinting:
393      # Drop device tables or contour anchor point
394      for m in self.Mark1Array.MarkRecord:
395        if m.MarkAnchor:
396          m.MarkAnchor.prune_hints()
397      for b in self.Mark2Array.Mark2Record:
398        for m in b.Mark2Anchor:
399          if m:
400            m.prune_hints()
401    return True
402
403@_add_method(otTables.SingleSubst,
404             otTables.MultipleSubst,
405             otTables.AlternateSubst,
406             otTables.LigatureSubst,
407             otTables.ReverseChainSingleSubst,
408             otTables.SinglePos,
409             otTables.PairPos,
410             otTables.CursivePos,
411             otTables.MarkBasePos,
412             otTables.MarkLigPos,
413             otTables.MarkMarkPos)
414def subset_lookups(self, lookup_indices):
415  pass
416
417@_add_method(otTables.SingleSubst,
418             otTables.MultipleSubst,
419             otTables.AlternateSubst,
420             otTables.LigatureSubst,
421             otTables.ReverseChainSingleSubst,
422             otTables.SinglePos,
423             otTables.PairPos,
424             otTables.CursivePos,
425             otTables.MarkBasePos,
426             otTables.MarkLigPos,
427             otTables.MarkMarkPos)
428def collect_lookups(self):
429  return []
430
431@_add_method(otTables.SingleSubst,
432             otTables.MultipleSubst,
433             otTables.AlternateSubst,
434             otTables.LigatureSubst,
435             otTables.ContextSubst,
436             otTables.ChainContextSubst,
437             otTables.ReverseChainSingleSubst,
438             otTables.SinglePos,
439             otTables.PairPos,
440             otTables.CursivePos,
441             otTables.MarkBasePos,
442             otTables.MarkLigPos,
443             otTables.MarkMarkPos,
444             otTables.ContextPos,
445             otTables.ChainContextPos)
446def prune_pre_subset(self, options):
447  return True
448
449@_add_method(otTables.SingleSubst,
450             otTables.MultipleSubst,
451             otTables.AlternateSubst,
452             otTables.LigatureSubst,
453             otTables.ReverseChainSingleSubst,
454             otTables.ContextSubst,
455             otTables.ChainContextSubst,
456             otTables.ContextPos,
457             otTables.ChainContextPos)
458def prune_post_subset(self, options):
459  return True
460
461@_add_method(otTables.SingleSubst,
462             otTables.AlternateSubst,
463             otTables.ReverseChainSingleSubst)
464def may_have_non_1to1(self):
465  return False
466
467@_add_method(otTables.MultipleSubst,
468             otTables.LigatureSubst,
469             otTables.ContextSubst,
470             otTables.ChainContextSubst)
471def may_have_non_1to1(self):
472  return True
473
474@_add_method(otTables.ContextSubst,
475             otTables.ChainContextSubst,
476             otTables.ContextPos,
477             otTables.ChainContextPos)
478def __classify_context(self):
479
480  class ContextHelper(object):
481    def __init__(self, klass, Format):
482      if klass.__name__.endswith('Subst'):
483        Typ = 'Sub'
484        Type = 'Subst'
485      else:
486        Typ = 'Pos'
487        Type = 'Pos'
488      if klass.__name__.startswith('Chain'):
489        Chain = 'Chain'
490      else:
491        Chain = ''
492      ChainTyp = Chain+Typ
493
494      self.Typ = Typ
495      self.Type = Type
496      self.Chain = Chain
497      self.ChainTyp = ChainTyp
498
499      self.LookupRecord = Type+'LookupRecord'
500
501      if Format == 1:
502        Coverage = lambda r: r.Coverage
503        ChainCoverage = lambda r: r.Coverage
504        ContextData = lambda r:(None,)
505        ChainContextData = lambda r:(None, None, None)
506        RuleData = lambda r:(r.Input,)
507        ChainRuleData = lambda r:(r.Backtrack, r.Input, r.LookAhead)
508        SetRuleData = None
509        ChainSetRuleData = None
510      elif Format == 2:
511        Coverage = lambda r: r.Coverage
512        ChainCoverage = lambda r: r.Coverage
513        ContextData = lambda r:(r.ClassDef,)
514        ChainContextData = lambda r:(r.LookAheadClassDef,
515                                      r.InputClassDef,
516                                      r.BacktrackClassDef)
517        RuleData = lambda r:(r.Class,)
518        ChainRuleData = lambda r:(r.LookAhead, r.Input, r.Backtrack)
519        def SetRuleData(r, d):(r.Class,) = d
520        def ChainSetRuleData(r, d):(r.LookAhead, r.Input, r.Backtrack) = d
521      elif Format == 3:
522        Coverage = lambda r: r.Coverage[0]
523        ChainCoverage = lambda r: r.InputCoverage[0]
524        ContextData = None
525        ChainContextData = None
526        RuleData = lambda r: r.Coverage
527        ChainRuleData = lambda r:(r.LookAheadCoverage +
528                                   r.InputCoverage +
529                                   r.BacktrackCoverage)
530        SetRuleData = None
531        ChainSetRuleData = None
532      else:
533        assert 0, "unknown format: %s" % Format
534
535      if Chain:
536        self.Coverage = ChainCoverage
537        self.ContextData = ChainContextData
538        self.RuleData = ChainRuleData
539        self.SetRuleData = ChainSetRuleData
540      else:
541        self.Coverage = Coverage
542        self.ContextData = ContextData
543        self.RuleData = RuleData
544        self.SetRuleData = SetRuleData
545
546      if Format == 1:
547        self.Rule = ChainTyp+'Rule'
548        self.RuleCount = ChainTyp+'RuleCount'
549        self.RuleSet = ChainTyp+'RuleSet'
550        self.RuleSetCount = ChainTyp+'RuleSetCount'
551        self.Intersect = lambda glyphs, c, r: [r] if r in glyphs else []
552      elif Format == 2:
553        self.Rule = ChainTyp+'ClassRule'
554        self.RuleCount = ChainTyp+'ClassRuleCount'
555        self.RuleSet = ChainTyp+'ClassSet'
556        self.RuleSetCount = ChainTyp+'ClassSetCount'
557        self.Intersect = lambda glyphs, c, r: c.intersect_class(glyphs, r)
558
559        self.ClassDef = 'InputClassDef' if Chain else 'ClassDef'
560        self.ClassDefIndex = 1 if Chain else 0
561        self.Input = 'Input' if Chain else 'Class'
562
563  if self.Format not in [1, 2, 3]:
564    return None  # Don't shoot the messenger; let it go
565  if not hasattr(self.__class__, "__ContextHelpers"):
566    self.__class__.__ContextHelpers = {}
567  if self.Format not in self.__class__.__ContextHelpers:
568    helper = ContextHelper(self.__class__, self.Format)
569    self.__class__.__ContextHelpers[self.Format] = helper
570  return self.__class__.__ContextHelpers[self.Format]
571
572@_add_method(otTables.ContextSubst,
573             otTables.ChainContextSubst)
574def closure_glyphs(self, s, cur_glyphs=None):
575  if cur_glyphs == None: cur_glyphs = s.glyphs
576  c = self.__classify_context()
577
578  indices = c.Coverage(self).intersect(s.glyphs)
579  if not indices:
580    return []
581  cur_glyphs = c.Coverage(self).intersect_glyphs(s.glyphs);
582
583  if self.Format == 1:
584    ContextData = c.ContextData(self)
585    rss = getattr(self, c.RuleSet)
586    for i in indices:
587      if not rss[i]: continue
588      for r in getattr(rss[i], c.Rule):
589        if not r: continue
590        if all(all(c.Intersect(s.glyphs, cd, k) for k in klist)
591          for cd,klist in zip(ContextData, c.RuleData(r))):
592          chaos = False
593          for ll in getattr(r, c.LookupRecord):
594            if not ll: continue
595            seqi = ll.SequenceIndex
596            if chaos:
597              pos_glyphs = s.glyphs
598            else:
599              if seqi == 0:
600                pos_glyphs = set([c.Coverage(self).glyphs[i]])
601              else:
602                pos_glyphs = set([r.Input[seqi - 1]])
603            lookup = s.table.LookupList.Lookup[ll.LookupListIndex]
604            chaos = chaos or lookup.may_have_non_1to1()
605            lookup.closure_glyphs(s, cur_glyphs=pos_glyphs)
606  elif self.Format == 2:
607    ClassDef = getattr(self, c.ClassDef)
608    indices = ClassDef.intersect(cur_glyphs)
609    ContextData = c.ContextData(self)
610    rss = getattr(self, c.RuleSet)
611    for i in indices:
612      if not rss[i]: continue
613      for r in getattr(rss[i], c.Rule):
614        if not r: continue
615        if all(all(c.Intersect(s.glyphs, cd, k) for k in klist)
616          for cd,klist in zip(ContextData, c.RuleData(r))):
617          chaos = False
618          for ll in getattr(r, c.LookupRecord):
619            if not ll: continue
620            seqi = ll.SequenceIndex
621            if chaos:
622              pos_glyphs = s.glyphs
623            else:
624              if seqi == 0:
625                pos_glyphs = ClassDef.intersect_class(cur_glyphs, i)
626              else:
627                pos_glyphs = ClassDef.intersect_class(s.glyphs,
628                                                      getattr(r, c.Input)[seqi - 1])
629            lookup = s.table.LookupList.Lookup[ll.LookupListIndex]
630            chaos = chaos or lookup.may_have_non_1to1()
631            lookup.closure_glyphs(s, cur_glyphs=pos_glyphs)
632  elif self.Format == 3:
633    if not all(x.intersect(s.glyphs) for x in c.RuleData(self)):
634      return []
635    r = self
636    chaos = False
637    for ll in getattr(r, c.LookupRecord):
638      if not ll: continue
639      seqi = ll.SequenceIndex
640      if chaos:
641        pos_glyphs = s.glyphs
642      else:
643        if seqi == 0:
644          pos_glyphs = cur_glyphs
645        else:
646          pos_glyphs = r.InputCoverage[seqi].intersect_glyphs(s.glyphs)
647      lookup = s.table.LookupList.Lookup[ll.LookupListIndex]
648      chaos = chaos or lookup.may_have_non_1to1()
649      lookup.closure_glyphs(s, cur_glyphs=pos_glyphs)
650  else:
651    assert 0, "unknown format: %s" % self.Format
652
653@_add_method(otTables.ContextSubst,
654             otTables.ContextPos,
655             otTables.ChainContextSubst,
656             otTables.ChainContextPos)
657def subset_glyphs(self, s):
658  c = self.__classify_context()
659
660  if self.Format == 1:
661    indices = self.Coverage.subset(s.glyphs)
662    rss = getattr(self, c.RuleSet)
663    rss = [rss[i] for i in indices]
664    for rs in rss:
665      if not rs: continue
666      ss = getattr(rs, c.Rule)
667      ss = [r for r in ss
668            if r and all(all(g in s.glyphs for g in glist)
669              for glist in c.RuleData(r))]
670      setattr(rs, c.Rule, ss)
671      setattr(rs, c.RuleCount, len(ss))
672    # Prune empty subrulesets
673    rss = [rs for rs in rss if rs and getattr(rs, c.Rule)]
674    setattr(self, c.RuleSet, rss)
675    setattr(self, c.RuleSetCount, len(rss))
676    return bool(rss)
677  elif self.Format == 2:
678    if not self.Coverage.subset(s.glyphs):
679      return False
680    ContextData = c.ContextData(self)
681    klass_maps = [x.subset(s.glyphs, remap=True) for x in ContextData]
682
683    # Keep rulesets for class numbers that survived.
684    indices = klass_maps[c.ClassDefIndex]
685    rss = getattr(self, c.RuleSet)
686    rssCount = getattr(self, c.RuleSetCount)
687    rss = [rss[i] for i in indices if i < rssCount]
688    del rssCount
689    # Delete, but not renumber, unreachable rulesets.
690    indices = getattr(self, c.ClassDef).intersect(self.Coverage.glyphs)
691    rss = [rss if i in indices else None for i,rss in enumerate(rss)]
692    while rss and rss[-1] == None:
693      del rss[-1]
694
695    for rs in rss:
696      if not rs: continue
697      ss = getattr(rs, c.Rule)
698      ss = [r for r in ss
699            if r and all(all(k in klass_map for k in klist)
700              for klass_map,klist in zip(klass_maps, c.RuleData(r)))]
701      setattr(rs, c.Rule, ss)
702      setattr(rs, c.RuleCount, len(ss))
703
704      # Remap rule classes
705      for r in ss:
706        c.SetRuleData(r, [[klass_map.index(k) for k in klist]
707               for klass_map,klist in zip(klass_maps, c.RuleData(r))])
708    return bool(rss)
709  elif self.Format == 3:
710    return all(x.subset(s.glyphs) for x in c.RuleData(self))
711  else:
712    assert 0, "unknown format: %s" % self.Format
713
714@_add_method(otTables.ContextSubst,
715             otTables.ChainContextSubst,
716             otTables.ContextPos,
717             otTables.ChainContextPos)
718def subset_lookups(self, lookup_indices):
719  c = self.__classify_context()
720
721  if self.Format in [1, 2]:
722    for rs in getattr(self, c.RuleSet):
723      if not rs: continue
724      for r in getattr(rs, c.Rule):
725        if not r: continue
726        setattr(r, c.LookupRecord,
727                 [ll for ll in getattr(r, c.LookupRecord)
728                  if ll and ll.LookupListIndex in lookup_indices])
729        for ll in getattr(r, c.LookupRecord):
730          if not ll: continue
731          ll.LookupListIndex = lookup_indices.index(ll.LookupListIndex)
732  elif self.Format == 3:
733    setattr(self, c.LookupRecord,
734             [ll for ll in getattr(self, c.LookupRecord)
735              if ll and ll.LookupListIndex in lookup_indices])
736    for ll in getattr(self, c.LookupRecord):
737      if not ll: continue
738      ll.LookupListIndex = lookup_indices.index(ll.LookupListIndex)
739  else:
740    assert 0, "unknown format: %s" % self.Format
741
742@_add_method(otTables.ContextSubst,
743             otTables.ChainContextSubst,
744             otTables.ContextPos,
745             otTables.ChainContextPos)
746def collect_lookups(self):
747  c = self.__classify_context()
748
749  if self.Format in [1, 2]:
750    return [ll.LookupListIndex
751      for rs in getattr(self, c.RuleSet) if rs
752      for r in getattr(rs, c.Rule) if r
753      for ll in getattr(r, c.LookupRecord) if ll]
754  elif self.Format == 3:
755    return [ll.LookupListIndex
756      for ll in getattr(self, c.LookupRecord) if ll]
757  else:
758    assert 0, "unknown format: %s" % self.Format
759
760@_add_method(otTables.ExtensionSubst)
761def closure_glyphs(self, s, cur_glyphs=None):
762  if self.Format == 1:
763    self.ExtSubTable.closure_glyphs(s, cur_glyphs)
764  else:
765    assert 0, "unknown format: %s" % self.Format
766
767@_add_method(otTables.ExtensionSubst)
768def may_have_non_1to1(self):
769  if self.Format == 1:
770    return self.ExtSubTable.may_have_non_1to1()
771  else:
772    assert 0, "unknown format: %s" % self.Format
773
774@_add_method(otTables.ExtensionSubst,
775             otTables.ExtensionPos)
776def prune_pre_subset(self, options):
777  if self.Format == 1:
778    return self.ExtSubTable.prune_pre_subset(options)
779  else:
780    assert 0, "unknown format: %s" % self.Format
781
782@_add_method(otTables.ExtensionSubst,
783             otTables.ExtensionPos)
784def subset_glyphs(self, s):
785  if self.Format == 1:
786    return self.ExtSubTable.subset_glyphs(s)
787  else:
788    assert 0, "unknown format: %s" % self.Format
789
790@_add_method(otTables.ExtensionSubst,
791             otTables.ExtensionPos)
792def prune_post_subset(self, options):
793  if self.Format == 1:
794    return self.ExtSubTable.prune_post_subset(options)
795  else:
796    assert 0, "unknown format: %s" % self.Format
797
798@_add_method(otTables.ExtensionSubst,
799             otTables.ExtensionPos)
800def subset_lookups(self, lookup_indices):
801  if self.Format == 1:
802    return self.ExtSubTable.subset_lookups(lookup_indices)
803  else:
804    assert 0, "unknown format: %s" % self.Format
805
806@_add_method(otTables.ExtensionSubst,
807             otTables.ExtensionPos)
808def collect_lookups(self):
809  if self.Format == 1:
810    return self.ExtSubTable.collect_lookups()
811  else:
812    assert 0, "unknown format: %s" % self.Format
813
814@_add_method(otTables.Lookup)
815def closure_glyphs(self, s, cur_glyphs=None):
816  for st in self.SubTable:
817    if not st: continue
818    st.closure_glyphs(s, cur_glyphs)
819
820@_add_method(otTables.Lookup)
821def prune_pre_subset(self, options):
822  ret = False
823  for st in self.SubTable:
824    if not st: continue
825    if st.prune_pre_subset(options): ret = True
826  return ret
827
828@_add_method(otTables.Lookup)
829def subset_glyphs(self, s):
830  self.SubTable = [st for st in self.SubTable if st and st.subset_glyphs(s)]
831  self.SubTableCount = len(self.SubTable)
832  return bool(self.SubTableCount)
833
834@_add_method(otTables.Lookup)
835def prune_post_subset(self, options):
836  ret = False
837  for st in self.SubTable:
838    if not st: continue
839    if st.prune_post_subset(options): ret = True
840  return ret
841
842@_add_method(otTables.Lookup)
843def subset_lookups(self, lookup_indices):
844  for s in self.SubTable:
845    s.subset_lookups(lookup_indices)
846
847@_add_method(otTables.Lookup)
848def collect_lookups(self):
849  return _uniq_sort(sum((st.collect_lookups() for st in self.SubTable
850                         if st), []))
851
852@_add_method(otTables.Lookup)
853def may_have_non_1to1(self):
854  return any(st.may_have_non_1to1() for st in self.SubTable if st)
855
856@_add_method(otTables.LookupList)
857def prune_pre_subset(self, options):
858  ret = False
859  for l in self.Lookup:
860    if not l: continue
861    if l.prune_pre_subset(options): ret = True
862  return ret
863
864@_add_method(otTables.LookupList)
865def subset_glyphs(self, s):
866  "Returns the indices of nonempty lookups."
867  return [i for i,l in enumerate(self.Lookup) if l and l.subset_glyphs(s)]
868
869@_add_method(otTables.LookupList)
870def prune_post_subset(self, options):
871  ret = False
872  for l in self.Lookup:
873    if not l: continue
874    if l.prune_post_subset(options): ret = True
875  return ret
876
877@_add_method(otTables.LookupList)
878def subset_lookups(self, lookup_indices):
879  self.Lookup = [self.Lookup[i] for i in lookup_indices
880                 if i < self.LookupCount]
881  self.LookupCount = len(self.Lookup)
882  for l in self.Lookup:
883    l.subset_lookups(lookup_indices)
884
885@_add_method(otTables.LookupList)
886def closure_lookups(self, lookup_indices):
887  lookup_indices = _uniq_sort(lookup_indices)
888  recurse = lookup_indices
889  while True:
890    recurse_lookups = sum((self.Lookup[i].collect_lookups()
891                            for i in recurse if i < self.LookupCount), [])
892    recurse_lookups = [l for l in recurse_lookups
893                       if l not in lookup_indices and l < self.LookupCount]
894    if not recurse_lookups:
895      return _uniq_sort(lookup_indices)
896    recurse_lookups = _uniq_sort(recurse_lookups)
897    lookup_indices.extend(recurse_lookups)
898    recurse = recurse_lookups
899
900@_add_method(otTables.Feature)
901def subset_lookups(self, lookup_indices):
902  self.LookupListIndex = [l for l in self.LookupListIndex
903                          if l in lookup_indices]
904  # Now map them.
905  self.LookupListIndex = [lookup_indices.index(l)
906                          for l in self.LookupListIndex]
907  self.LookupCount = len(self.LookupListIndex)
908  return self.LookupCount
909
910@_add_method(otTables.Feature)
911def collect_lookups(self):
912  return self.LookupListIndex[:]
913
914@_add_method(otTables.FeatureList)
915def subset_lookups(self, lookup_indices):
916  "Returns the indices of nonempty features."
917  feature_indices = [i for i,f in enumerate(self.FeatureRecord)
918                     if f.Feature.subset_lookups(lookup_indices)]
919  self.subset_features(feature_indices)
920  return feature_indices
921
922@_add_method(otTables.FeatureList)
923def collect_lookups(self, feature_indices):
924  return _uniq_sort(sum((self.FeatureRecord[i].Feature.collect_lookups()
925                         for i in feature_indices
926                          if i < self.FeatureCount), []))
927
928@_add_method(otTables.FeatureList)
929def subset_features(self, feature_indices):
930  self.FeatureRecord = [self.FeatureRecord[i] for i in feature_indices]
931  self.FeatureCount = len(self.FeatureRecord)
932  return bool(self.FeatureCount)
933
934@_add_method(otTables.DefaultLangSys,
935             otTables.LangSys)
936def subset_features(self, feature_indices):
937  if self.ReqFeatureIndex in feature_indices:
938    self.ReqFeatureIndex = feature_indices.index(self.ReqFeatureIndex)
939  else:
940    self.ReqFeatureIndex = 65535
941  self.FeatureIndex = [f for f in self.FeatureIndex if f in feature_indices]
942  # Now map them.
943  self.FeatureIndex = [feature_indices.index(f) for f in self.FeatureIndex
944                       if f in feature_indices]
945  self.FeatureCount = len(self.FeatureIndex)
946  return bool(self.FeatureCount or self.ReqFeatureIndex != 65535)
947
948@_add_method(otTables.DefaultLangSys,
949             otTables.LangSys)
950def collect_features(self):
951  feature_indices = self.FeatureIndex[:]
952  if self.ReqFeatureIndex != 65535:
953    feature_indices.append(self.ReqFeatureIndex)
954  return _uniq_sort(feature_indices)
955
956@_add_method(otTables.Script)
957def subset_features(self, feature_indices):
958  if(self.DefaultLangSys and
959      not self.DefaultLangSys.subset_features(feature_indices)):
960    self.DefaultLangSys = None
961  self.LangSysRecord = [l for l in self.LangSysRecord
962                        if l.LangSys.subset_features(feature_indices)]
963  self.LangSysCount = len(self.LangSysRecord)
964  return bool(self.LangSysCount or self.DefaultLangSys)
965
966@_add_method(otTables.Script)
967def collect_features(self):
968  feature_indices = [l.LangSys.collect_features() for l in self.LangSysRecord]
969  if self.DefaultLangSys:
970    feature_indices.append(self.DefaultLangSys.collect_features())
971  return _uniq_sort(sum(feature_indices, []))
972
973@_add_method(otTables.ScriptList)
974def subset_features(self, feature_indices):
975  self.ScriptRecord = [s for s in self.ScriptRecord
976                       if s.Script.subset_features(feature_indices)]
977  self.ScriptCount = len(self.ScriptRecord)
978  return bool(self.ScriptCount)
979
980@_add_method(otTables.ScriptList)
981def collect_features(self):
982  return _uniq_sort(sum((s.Script.collect_features()
983                         for s in self.ScriptRecord), []))
984
985@_add_method(ttLib.getTableClass('GSUB'))
986def closure_glyphs(self, s):
987  s.table = self.table
988  feature_indices = self.table.ScriptList.collect_features()
989  lookup_indices = self.table.FeatureList.collect_lookups(feature_indices)
990  while True:
991    orig_glyphs = s.glyphs.copy()
992    for i in lookup_indices:
993      if i >= self.table.LookupList.LookupCount: continue
994      if not self.table.LookupList.Lookup[i]: continue
995      self.table.LookupList.Lookup[i].closure_glyphs(s)
996    if orig_glyphs == s.glyphs:
997      break
998  del s.table
999
1000@_add_method(ttLib.getTableClass('GSUB'),
1001             ttLib.getTableClass('GPOS'))
1002def subset_glyphs(self, s):
1003  s.glyphs = s.glyphs_gsubed
1004  lookup_indices = self.table.LookupList.subset_glyphs(s)
1005  self.subset_lookups(lookup_indices)
1006  self.prune_lookups()
1007  return True
1008
1009@_add_method(ttLib.getTableClass('GSUB'),
1010             ttLib.getTableClass('GPOS'))
1011def subset_lookups(self, lookup_indices):
1012  """Retrains specified lookups, then removes empty features, language
1013     systems, and scripts."""
1014  self.table.LookupList.subset_lookups(lookup_indices)
1015  feature_indices = self.table.FeatureList.subset_lookups(lookup_indices)
1016  self.table.ScriptList.subset_features(feature_indices)
1017
1018@_add_method(ttLib.getTableClass('GSUB'),
1019             ttLib.getTableClass('GPOS'))
1020def prune_lookups(self):
1021  "Remove unreferenced lookups"
1022  feature_indices = self.table.ScriptList.collect_features()
1023  lookup_indices = self.table.FeatureList.collect_lookups(feature_indices)
1024  lookup_indices = self.table.LookupList.closure_lookups(lookup_indices)
1025  self.subset_lookups(lookup_indices)
1026
1027@_add_method(ttLib.getTableClass('GSUB'),
1028             ttLib.getTableClass('GPOS'))
1029def subset_feature_tags(self, feature_tags):
1030  feature_indices = [i for i,f in
1031                     enumerate(self.table.FeatureList.FeatureRecord)
1032                     if f.FeatureTag in feature_tags]
1033  self.table.FeatureList.subset_features(feature_indices)
1034  self.table.ScriptList.subset_features(feature_indices)
1035
1036@_add_method(ttLib.getTableClass('GSUB'),
1037             ttLib.getTableClass('GPOS'))
1038def prune_pre_subset(self, options):
1039  if '*' not in options.layout_features:
1040    self.subset_feature_tags(options.layout_features)
1041  self.prune_lookups()
1042  self.table.LookupList.prune_pre_subset(options);
1043  return True
1044
1045@_add_method(ttLib.getTableClass('GSUB'),
1046             ttLib.getTableClass('GPOS'))
1047def prune_post_subset(self, options):
1048  self.table.LookupList.prune_post_subset(options);
1049  return True
1050
1051@_add_method(ttLib.getTableClass('GDEF'))
1052def subset_glyphs(self, s):
1053  glyphs = s.glyphs_gsubed
1054  table = self.table
1055  if table.LigCaretList:
1056    indices = table.LigCaretList.Coverage.subset(glyphs)
1057    table.LigCaretList.LigGlyph = [table.LigCaretList.LigGlyph[i]
1058                                   for i in indices]
1059    table.LigCaretList.LigGlyphCount = len(table.LigCaretList.LigGlyph)
1060    if not table.LigCaretList.LigGlyphCount:
1061      table.LigCaretList = None
1062  if table.MarkAttachClassDef:
1063    table.MarkAttachClassDef.classDefs = dict((g,v) for g,v in
1064                                              table.MarkAttachClassDef.
1065                                                classDefs.iteritems()
1066                                              if g in glyphs)
1067    if not table.MarkAttachClassDef.classDefs:
1068      table.MarkAttachClassDef = None
1069  if table.GlyphClassDef:
1070    table.GlyphClassDef.classDefs = dict((g,v) for g,v in
1071                                         table.GlyphClassDef.
1072                                           classDefs.iteritems()
1073                                         if g in glyphs)
1074    if not table.GlyphClassDef.classDefs:
1075      table.GlyphClassDef = None
1076  if table.AttachList:
1077    indices = table.AttachList.Coverage.subset(glyphs)
1078    table.AttachList.AttachPoint = [table.AttachList.AttachPoint[i]
1079                                    for i in indices]
1080    table.AttachList.GlyphCount = len(table.AttachList.AttachPoint)
1081    if not table.AttachList.GlyphCount:
1082      table.AttachList = None
1083  return bool(table.LigCaretList or
1084               table.MarkAttachClassDef or
1085               table.GlyphClassDef or
1086               table.AttachList)
1087
1088@_add_method(ttLib.getTableClass('kern'))
1089def prune_pre_subset(self, options):
1090  # Prune unknown kern table types
1091  self.kernTables = [t for t in self.kernTables if hasattr(t, 'kernTable')]
1092  return bool(self.kernTables)
1093
1094@_add_method(ttLib.getTableClass('kern'))
1095def subset_glyphs(self, s):
1096  glyphs = s.glyphs_gsubed
1097  for t in self.kernTables:
1098    t.kernTable = dict(((a,b),v) for (a,b),v in t.kernTable.iteritems()
1099                       if a in glyphs and b in glyphs)
1100  self.kernTables = [t for t in self.kernTables if t.kernTable]
1101  return bool(self.kernTables)
1102
1103@_add_method(ttLib.getTableClass('vmtx'),
1104             ttLib.getTableClass('hmtx'))
1105def subset_glyphs(self, s):
1106  self.metrics = dict((g,v) for g,v in self.metrics.iteritems() if g in s.glyphs)
1107  return bool(self.metrics)
1108
1109@_add_method(ttLib.getTableClass('hdmx'))
1110def subset_glyphs(self, s):
1111  self.hdmx = dict((sz,dict((g,v) for g,v in l.iteritems() if g in s.glyphs))
1112                   for sz,l in self.hdmx.iteritems())
1113  return bool(self.hdmx)
1114
1115@_add_method(ttLib.getTableClass('VORG'))
1116def subset_glyphs(self, s):
1117  self.VOriginRecords = dict((g,v) for g,v in self.VOriginRecords.iteritems()
1118                             if g in s.glyphs)
1119  self.numVertOriginYMetrics = len(self.VOriginRecords)
1120  return True  # Never drop; has default metrics
1121
1122@_add_method(ttLib.getTableClass('post'))
1123def prune_pre_subset(self, options):
1124  if not options.glyph_names:
1125    self.formatType = 3.0
1126  return True
1127
1128@_add_method(ttLib.getTableClass('post'))
1129def subset_glyphs(self, s):
1130  self.extraNames = []  # This seems to do it
1131  return True
1132
1133@_add_method(ttLib.getTableModule('glyf').Glyph)
1134def remapComponentsFast(self, indices):
1135  if not self.data or struct.unpack(">h", self.data[:2])[0] >= 0:
1136    return  # Not composite
1137  data = array.array("B", self.data)
1138  i = 10
1139  more = 1
1140  while more:
1141    flags =(data[i] << 8) | data[i+1]
1142    glyphID =(data[i+2] << 8) | data[i+3]
1143    # Remap
1144    glyphID = indices.index(glyphID)
1145    data[i+2] = glyphID >> 8
1146    data[i+3] = glyphID & 0xFF
1147    i += 4
1148    flags = int(flags)
1149
1150    if flags & 0x0001: i += 4  # ARG_1_AND_2_ARE_WORDS
1151    else: i += 2
1152    if flags & 0x0008: i += 2  # WE_HAVE_A_SCALE
1153    elif flags & 0x0040: i += 4  # WE_HAVE_AN_X_AND_Y_SCALE
1154    elif flags & 0x0080: i += 8  # WE_HAVE_A_TWO_BY_TWO
1155    more = flags & 0x0020  # MORE_COMPONENTS
1156
1157  self.data = data.tostring()
1158
1159@_add_method(ttLib.getTableClass('glyf'))
1160def closure_glyphs(self, s):
1161  decompose = s.glyphs
1162  while True:
1163    components = set()
1164    for g in decompose:
1165      if g not in self.glyphs:
1166        continue
1167      gl = self.glyphs[g]
1168      for c in gl.getComponentNames(self):
1169        if c not in s.glyphs:
1170          components.add(c)
1171    components = set(c for c in components if c not in s.glyphs)
1172    if not components:
1173      break
1174    decompose = components
1175    s.glyphs.update(components)
1176
1177@_add_method(ttLib.getTableClass('glyf'))
1178def prune_pre_subset(self, options):
1179  if options.notdef_glyph and not options.notdef_outline:
1180    g = self[self.glyphOrder[0]]
1181    # Yay, easy!
1182    g.__dict__.clear()
1183    g.data = ""
1184  return True
1185
1186@_add_method(ttLib.getTableClass('glyf'))
1187def subset_glyphs(self, s):
1188  self.glyphs = dict((g,v) for g,v in self.glyphs.iteritems() if g in s.glyphs)
1189  indices = [i for i,g in enumerate(self.glyphOrder) if g in s.glyphs]
1190  for v in self.glyphs.itervalues():
1191    if hasattr(v, "data"):
1192      v.remapComponentsFast(indices)
1193    else:
1194      pass  # No need
1195  self.glyphOrder = [g for g in self.glyphOrder if g in s.glyphs]
1196  # Don't drop empty 'glyf' tables, otherwise 'loca' doesn't get subset.
1197  return True
1198
1199@_add_method(ttLib.getTableClass('glyf'))
1200def prune_post_subset(self, options):
1201  if not options.hinting:
1202    for v in self.glyphs.itervalues():
1203      v.removeHinting()
1204  return True
1205
1206@_add_method(ttLib.getTableClass('CFF '))
1207def prune_pre_subset(self, options):
1208  cff = self.cff
1209  # CFF table must have one font only
1210  cff.fontNames = cff.fontNames[:1]
1211
1212  if options.notdef_glyph and not options.notdef_outline:
1213    for fontname in cff.keys():
1214      font = cff[fontname]
1215      c,_ = font.CharStrings.getItemAndSelector('.notdef')
1216      # XXX we should preserve the glyph width
1217      c.bytecode = '\x0e' # endchar
1218      c.program = None
1219
1220  return True # bool(cff.fontNames)
1221
1222@_add_method(ttLib.getTableClass('CFF '))
1223def subset_glyphs(self, s):
1224  cff = self.cff
1225  for fontname in cff.keys():
1226    font = cff[fontname]
1227    cs = font.CharStrings
1228
1229    # Load all glyphs
1230    for g in font.charset:
1231      if g not in s.glyphs: continue
1232      c,sel = cs.getItemAndSelector(g)
1233
1234    if cs.charStringsAreIndexed:
1235      indices = [i for i,g in enumerate(font.charset) if g in s.glyphs]
1236      csi = cs.charStringsIndex
1237      csi.items = [csi.items[i] for i in indices]
1238      csi.count = len(csi.items)
1239      del csi.file, csi.offsets
1240      if hasattr(font, "FDSelect"):
1241        sel = font.FDSelect
1242        sel.format = None
1243        sel.gidArray = [sel.gidArray[i] for i in indices]
1244      cs.charStrings = dict((g,indices.index(v))
1245                            for g,v in cs.charStrings.iteritems()
1246                            if g in s.glyphs)
1247    else:
1248      cs.charStrings = dict((g,v)
1249                            for g,v in cs.charStrings.iteritems()
1250                            if g in s.glyphs)
1251    font.charset = [g for g in font.charset if g in s.glyphs]
1252    font.numGlyphs = len(font.charset)
1253
1254  return True # any(cff[fontname].numGlyphs for fontname in cff.keys())
1255
1256@_add_method(psCharStrings.T2CharString)
1257def subset_subroutines(self, subrs, gsubrs):
1258  p = self.program
1259  assert len(p)
1260  for i in xrange(1, len(p)):
1261    if p[i] == 'callsubr':
1262      assert type(p[i-1]) is int
1263      p[i-1] = subrs._used.index(p[i-1] + subrs._old_bias) - subrs._new_bias
1264    elif p[i] == 'callgsubr':
1265      assert type(p[i-1]) is int
1266      p[i-1] = gsubrs._used.index(p[i-1] + gsubrs._old_bias) - gsubrs._new_bias
1267
1268@_add_method(psCharStrings.T2CharString)
1269def drop_hints(self):
1270  hints = self._hints
1271
1272  if hints.has_hint:
1273    self.program = self.program[hints.last_hint:]
1274    if hasattr(self, 'width'):
1275      # Insert width back if needed
1276      if self.width != self.private.defaultWidthX:
1277        self.program.insert(0, self.width - self.private.nominalWidthX)
1278
1279  if hints.has_hintmask:
1280    i = 0
1281    p = self.program
1282    while i < len(p):
1283      if p[i] in ['hintmask', 'cntrmask']:
1284        assert i + 1 <= len(p)
1285        del p[i:i+2]
1286        continue
1287      i += 1
1288
1289  assert len(self.program)
1290
1291  del self._hints
1292
1293class _MarkingT2Decompiler(psCharStrings.SimpleT2Decompiler):
1294
1295  def __init__(self, localSubrs, globalSubrs):
1296    psCharStrings.SimpleT2Decompiler.__init__(self,
1297                                              localSubrs,
1298                                              globalSubrs)
1299    for subrs in [localSubrs, globalSubrs]:
1300      if subrs and not hasattr(subrs, "_used"):
1301        subrs._used = set()
1302
1303  def op_callsubr(self, index):
1304    self.localSubrs._used.add(self.operandStack[-1]+self.localBias)
1305    psCharStrings.SimpleT2Decompiler.op_callsubr(self, index)
1306
1307  def op_callgsubr(self, index):
1308    self.globalSubrs._used.add(self.operandStack[-1]+self.globalBias)
1309    psCharStrings.SimpleT2Decompiler.op_callgsubr(self, index)
1310
1311class _DehintingT2Decompiler(psCharStrings.SimpleT2Decompiler):
1312
1313  class Hints:
1314    def __init__(self):
1315      # Whether calling this charstring produces any hint stems
1316      self.has_hint = False
1317      # Index to start at to drop all hints
1318      self.last_hint = 0
1319      # Index up to which we know more hints are possible.  Only
1320      # relevant if status is 0 or 1.
1321      self.last_checked = 0
1322      # The status means:
1323      # 0: after dropping hints, this charstring is empty
1324      # 1: after dropping hints, there may be more hints continuing after this
1325      # 2: no more hints possible after this charstring
1326      self.status = 0
1327      # Has hintmask instructions; not recursive
1328      self.has_hintmask = False
1329    pass
1330
1331  def __init__(self, css, localSubrs, globalSubrs):
1332    self._css = css
1333    psCharStrings.SimpleT2Decompiler.__init__(self,
1334                                              localSubrs,
1335                                              globalSubrs)
1336
1337  def execute(self, charString):
1338    old_hints = charString._hints if hasattr(charString, '_hints') else None
1339    charString._hints = self.Hints()
1340
1341    psCharStrings.SimpleT2Decompiler.execute(self, charString)
1342
1343    hints = charString._hints
1344
1345    if hints.has_hint or hints.has_hintmask:
1346      self._css.add(charString)
1347
1348    if hints.status != 2:
1349      # Check from last_check, make sure we didn't have any operators.
1350      for i in xrange(hints.last_checked, len(charString.program) - 1):
1351        if type(charString.program[i]) == str:
1352          hints.status = 2
1353          break;
1354        else:
1355          hints.status = 1 # There's *something* here
1356      hints.last_checked = len(charString.program)
1357
1358    if old_hints:
1359      assert hints.__dict__ == old_hints.__dict__
1360
1361  def op_callsubr(self, index):
1362    subr = self.localSubrs[self.operandStack[-1]+self.localBias]
1363    psCharStrings.SimpleT2Decompiler.op_callsubr(self, index)
1364    self.processSubr(index, subr)
1365
1366  def op_callgsubr(self, index):
1367    subr = self.globalSubrs[self.operandStack[-1]+self.globalBias]
1368    psCharStrings.SimpleT2Decompiler.op_callgsubr(self, index)
1369    self.processSubr(index, subr)
1370
1371  def op_hstem(self, index):
1372    psCharStrings.SimpleT2Decompiler.op_hstem(self, index)
1373    self.processHint(index)
1374  def op_vstem(self, index):
1375    psCharStrings.SimpleT2Decompiler.op_vstem(self, index)
1376    self.processHint(index)
1377  def op_hstemhm(self, index):
1378    psCharStrings.SimpleT2Decompiler.op_hstemhm(self, index)
1379    self.processHint(index)
1380  def op_vstemhm(self, index):
1381    psCharStrings.SimpleT2Decompiler.op_vstemhm(self, index)
1382    self.processHint(index)
1383  def op_hintmask(self, index):
1384    psCharStrings.SimpleT2Decompiler.op_hintmask(self, index)
1385    self.processHintmask(index)
1386  def op_cntrmask(self, index):
1387    psCharStrings.SimpleT2Decompiler.op_cntrmask(self, index)
1388    self.processHintmask(index)
1389
1390  def processHintmask(self, index):
1391    cs = self.callingStack[-1]
1392    hints = cs._hints
1393    hints.has_hintmask = True
1394    if hints.status != 2 and hints.has_hint:
1395      # Check from last_check, see if we may be an implicit vstem
1396      for i in xrange(hints.last_checked, index - 1):
1397        if type(cs.program[i]) == str:
1398          hints.status = 2
1399          break;
1400      if hints.status != 2:
1401        # We are an implicit vstem
1402        hints.last_hint = index + 1
1403        hints.status = 0
1404      hints.last_checked = index + 1
1405
1406  def processHint(self, index):
1407    cs = self.callingStack[-1]
1408    hints = cs._hints
1409    hints.has_hint = True
1410    hints.last_hint = index
1411    hints.last_checked = index
1412
1413  def processSubr(self, index, subr):
1414    cs = self.callingStack[-1]
1415    hints = cs._hints
1416    subr_hints = subr._hints
1417
1418    if subr_hints.has_hint:
1419      if hints.status != 2:
1420        hints.has_hint = True
1421        hints.last_checked = index
1422        hints.status = subr_hints.status
1423        # Decide where to chop off from
1424        if subr_hints.status == 0:
1425          hints.last_hint = index
1426        else:
1427          hints.last_hint = index - 2 # Leave the subr call in
1428      else:
1429        # In my understanding, this is a font bug.  Ie. it has hint stems
1430        # *after* path construction.  I've seen this in widespread fonts.
1431        # Best to ignore the hints I suppose...
1432        pass
1433        #assert 0
1434    else:
1435      hints.status = max(hints.status, subr_hints.status)
1436      if hints.status != 2:
1437        # Check from last_check, make sure we didn't have
1438        # any operators.
1439        for i in xrange(hints.last_checked, index - 1):
1440          if type(cs.program[i]) == str:
1441            hints.status = 2
1442            break;
1443        hints.last_checked = index
1444
1445@_add_method(ttLib.getTableClass('CFF '))
1446def prune_post_subset(self, options):
1447  cff = self.cff
1448  for fontname in cff.keys():
1449    font = cff[fontname]
1450    cs = font.CharStrings
1451
1452
1453    #
1454    # Drop unused FontDictionaries
1455    #
1456    if hasattr(font, "FDSelect"):
1457      sel = font.FDSelect
1458      indices = _uniq_sort(sel.gidArray)
1459      sel.gidArray = [indices.index (ss) for ss in sel.gidArray]
1460      arr = font.FDArray
1461      arr.items = [arr[i] for i in indices]
1462      arr.count = len(arr.items)
1463      del arr.file, arr.offsets
1464
1465
1466    #
1467    # Drop hints if not needed
1468    #
1469    if not options.hinting:
1470
1471      #
1472      # This can be tricky, but doesn't have to.  What we do is:
1473      #
1474      # - Run all used glyph charstrings and recurse into subroutines,
1475      # - For each charstring (including subroutines), if it has any
1476      #   of the hint stem operators, we mark it as such.  Upon returning,
1477      #   for each charstring we note all the subroutine calls it makes
1478      #   that (recursively) contain a stem,
1479      # - Dropping hinting then consists of the following two ops:
1480      #   * Drop the piece of the program in each charstring before the
1481      #     last call to a stem op or a stem-calling subroutine,
1482      #   * Drop all hintmask operations.
1483      # - It's trickier... A hintmask right after hints and a few numbers
1484      #   will act as an implicit vstemhm.  As such, we track whether
1485      #   we have seen any non-hint operators so far and do the right
1486      #   thing, recursively...  Good luck understanding that :(
1487      #
1488      css = set()
1489      for g in font.charset:
1490        c,sel = cs.getItemAndSelector(g)
1491        # Make sure it's decompiled.  We want our "decompiler" to walk
1492        # the program, not the bytecode.
1493        c.draw(basePen.NullPen())
1494        subrs = getattr(c.private, "Subrs", [])
1495        decompiler = _DehintingT2Decompiler(css, subrs, c.globalSubrs)
1496        decompiler.execute(c)
1497      for charstring in css:
1498        charstring.drop_hints()
1499
1500      # Drop font-wide hinting values
1501      all_privs = []
1502      if hasattr(font, 'FDSelect'):
1503        all_privs.extend(fd.Private for fd in font.FDArray)
1504      else:
1505        all_privs.append(font.Private)
1506      for priv in all_privs:
1507        priv.BlueValues = []
1508        for k in ['OtherBlues', 'StemSnapH', 'StemSnapV', 'StdHW', 'StdVW', \
1509                  'FamilyBlues', 'FamilyOtherBlues']:
1510          if hasattr(priv, k):
1511            setattr(priv, k, None)
1512
1513
1514    #
1515    # Renumber subroutines to remove unused ones
1516    #
1517
1518    # Mark all used subroutines
1519    for g in font.charset:
1520      c,sel = cs.getItemAndSelector(g)
1521      subrs = getattr(c.private, "Subrs", [])
1522      decompiler = _MarkingT2Decompiler(subrs, c.globalSubrs)
1523      decompiler.execute(c)
1524
1525    all_subrs = [font.GlobalSubrs]
1526    if hasattr(font, 'FDSelect'):
1527      all_subrs.extend(fd.Private.Subrs for fd in font.FDArray if hasattr(fd.Private, 'Subrs') and fd.Private.Subrs)
1528    elif hasattr(font.Private, 'Subrs') and font.Private.Subrs:
1529      all_subrs.append(font.Private.Subrs)
1530
1531    subrs = set(subrs) # Remove duplicates
1532
1533    # Prepare
1534    for subrs in all_subrs:
1535      if not hasattr(subrs, '_used'):
1536        subrs._used = set()
1537      subrs._used = _uniq_sort(subrs._used)
1538      subrs._old_bias = psCharStrings.calcSubrBias(subrs)
1539      subrs._new_bias = psCharStrings.calcSubrBias(subrs._used)
1540
1541    # Renumber glyph charstrings
1542    for g in font.charset:
1543      c,sel = cs.getItemAndSelector(g)
1544      subrs = getattr(c.private, "Subrs", [])
1545      c.subset_subroutines (subrs, font.GlobalSubrs)
1546
1547    # Renumber subroutines themselves
1548    for subrs in all_subrs:
1549
1550      if subrs == font.GlobalSubrs:
1551        if not hasattr(font, 'FDSelect') and hasattr(font.Private, 'Subrs'):
1552          local_subrs = font.Private.Subrs
1553        else:
1554          local_subrs = []
1555      else:
1556        local_subrs = subrs
1557
1558      subrs.items = [subrs.items[i] for i in subrs._used]
1559      subrs.count = len(subrs.items)
1560      del subrs.file
1561      if hasattr(subrs, 'offsets'):
1562        del subrs.offsets
1563
1564      for i in xrange (subrs.count):
1565        subrs[i].subset_subroutines (local_subrs, font.GlobalSubrs)
1566
1567    # Cleanup
1568    for subrs in all_subrs:
1569      del subrs._used, subrs._old_bias, subrs._new_bias
1570
1571  return True
1572
1573@_add_method(ttLib.getTableClass('cmap'))
1574def closure_glyphs(self, s):
1575  tables = [t for t in self.tables
1576            if t.platformID == 3 and t.platEncID in [1, 10]]
1577  for u in s.unicodes_requested:
1578    found = False
1579    for table in tables:
1580      if u in table.cmap:
1581        s.glyphs.add(table.cmap[u])
1582        found = True
1583        break
1584    if not found:
1585      s.log("No glyph for Unicode value %s; skipping." % u)
1586
1587@_add_method(ttLib.getTableClass('cmap'))
1588def prune_pre_subset(self, options):
1589  if not options.legacy_cmap:
1590    # Drop non-Unicode / non-Symbol cmaps
1591    self.tables = [t for t in self.tables
1592                   if t.platformID == 3 and t.platEncID in [0, 1, 10]]
1593  if not options.symbol_cmap:
1594    self.tables = [t for t in self.tables
1595                   if t.platformID == 3 and t.platEncID in [1, 10]]
1596  # TODO(behdad) Only keep one subtable?
1597  # For now, drop format=0 which can't be subset_glyphs easily?
1598  self.tables = [t for t in self.tables if t.format != 0]
1599  self.numSubTables = len(self.tables)
1600  return bool(self.tables)
1601
1602@_add_method(ttLib.getTableClass('cmap'))
1603def subset_glyphs(self, s):
1604  s.glyphs = s.glyphs_cmaped
1605  for t in self.tables:
1606    # For reasons I don't understand I need this here
1607    # to force decompilation of the cmap format 14.
1608    try:
1609      getattr(t, "asdf")
1610    except AttributeError:
1611      pass
1612    if t.format == 14:
1613      # TODO(behdad) XXX We drop all the default-UVS mappings(g==None).
1614      t.uvsDict = dict((v,[(u,g) for u,g in l if g in s.glyphs])
1615                       for v,l in t.uvsDict.iteritems())
1616      t.uvsDict = dict((v,l) for v,l in t.uvsDict.iteritems() if l)
1617    else:
1618      t.cmap = dict((u,g) for u,g in t.cmap.iteritems()
1619                    if g in s.glyphs_requested or u in s.unicodes_requested)
1620  self.tables = [t for t in self.tables
1621                 if (t.cmap if t.format != 14 else t.uvsDict)]
1622  self.numSubTables = len(self.tables)
1623  # TODO(behdad) Convert formats when needed.
1624  # In particular, if we have a format=12 without non-BMP
1625  # characters, either drop format=12 one or convert it
1626  # to format=4 if there's not one.
1627  return bool(self.tables)
1628
1629@_add_method(ttLib.getTableClass('name'))
1630def prune_pre_subset(self, options):
1631  if '*' not in options.name_IDs:
1632    self.names = [n for n in self.names if n.nameID in options.name_IDs]
1633  if not options.name_legacy:
1634    self.names = [n for n in self.names
1635                  if n.platformID == 3 and n.platEncID == 1]
1636  if '*' not in options.name_languages:
1637    self.names = [n for n in self.names if n.langID in options.name_languages]
1638  return True  # Retain even if empty
1639
1640
1641# TODO(behdad) OS/2 ulUnicodeRange / ulCodePageRange?
1642# TODO(behdad) Drop AAT tables.
1643# TODO(behdad) Drop unneeded GSUB/GPOS Script/LangSys entries.
1644# TODO(behdad) Drop empty GSUB/GPOS, and GDEF if no GSUB/GPOS left
1645# TODO(behdad) Drop GDEF subitems if unused by lookups
1646# TODO(behdad) Avoid recursing too much (in GSUB/GPOS and in CFF)
1647# TODO(behdad) Text direction considerations.
1648# TODO(behdad) Text script / language considerations.
1649
1650class Options(object):
1651
1652  class UnknownOptionError(Exception):
1653    pass
1654
1655  _drop_tables_default = ['BASE', 'JSTF', 'DSIG', 'EBDT', 'EBLC', 'EBSC', 'SVG ',
1656                          'PCLT', 'LTSH']
1657  _drop_tables_default += ['Feat', 'Glat', 'Gloc', 'Silf', 'Sill']  # Graphite
1658  _drop_tables_default += ['CBLC', 'CBDT', 'sbix', 'COLR', 'CPAL']  # Color
1659  _no_subset_tables_default = ['gasp', 'head', 'hhea', 'maxp', 'vhea', 'OS/2',
1660                               'loca', 'name', 'cvt ', 'fpgm', 'prep']
1661  _hinting_tables_default = ['cvt ', 'fpgm', 'prep', 'hdmx', 'VDMX']
1662
1663  # Based on HarfBuzz shapers
1664  _layout_features_groups = {
1665    # Default shaper
1666    'common': ['ccmp', 'liga', 'locl', 'mark', 'mkmk', 'rlig'],
1667    'horizontal': ['calt', 'clig', 'curs', 'kern', 'rclt'],
1668    'vertical':  ['valt', 'vert', 'vkrn', 'vpal', 'vrt2'],
1669    'ltr': ['ltra', 'ltrm'],
1670    'rtl': ['rtla', 'rtlm'],
1671    # Complex shapers
1672    'arabic': ['init', 'medi', 'fina', 'isol', 'med2', 'fin2', 'fin3',
1673               'cswh', 'mset'],
1674    'hangul': ['ljmo', 'vjmo', 'tjmo'],
1675    'tibetan': ['abvs', 'blws', 'abvm', 'blwm'],
1676    'indic': ['nukt', 'akhn', 'rphf', 'rkrf', 'pref', 'blwf', 'half',
1677              'abvf', 'pstf', 'cfar', 'vatu', 'cjct', 'init', 'pres',
1678              'abvs', 'blws', 'psts', 'haln', 'dist', 'abvm', 'blwm'],
1679  }
1680  _layout_features_default = _uniq_sort(sum(
1681      _layout_features_groups.itervalues(), []))
1682
1683  drop_tables = _drop_tables_default
1684  no_subset_tables = _no_subset_tables_default
1685  hinting_tables = _hinting_tables_default
1686  layout_features = _layout_features_default
1687  hinting = False
1688  glyph_names = False
1689  legacy_cmap = False
1690  symbol_cmap = False
1691  name_IDs = [1, 2]  # Family and Style
1692  name_legacy = False
1693  name_languages = [0x0409]  # English
1694  notdef_glyph = True # gid0 for TrueType / .notdef for CFF
1695  notdef_outline = False # No need for notdef to have an outline really
1696  recommended_glyphs = False  # gid1, gid2, gid3 for TrueType
1697  recalc_bounds = False # Recalculate font bounding boxes
1698  canonical_order = False # Order tables as recommended
1699  flavor = None # May be 'woff'
1700
1701  def __init__(self, **kwargs):
1702
1703    self.set(**kwargs)
1704
1705  def set(self, **kwargs):
1706    for k,v in kwargs.iteritems():
1707      if not hasattr(self, k):
1708        raise self.UnknownOptionError("Unknown option '%s'" % k)
1709      setattr(self, k, v)
1710
1711  def parse_opts(self, argv, ignore_unknown=False):
1712    ret = []
1713    opts = {}
1714    for a in argv:
1715      orig_a = a
1716      if not a.startswith('--'):
1717        ret.append(a)
1718        continue
1719      a = a[2:]
1720      i = a.find('=')
1721      op = '='
1722      if i == -1:
1723        if a.startswith("no-"):
1724          k = a[3:]
1725          v = False
1726        else:
1727          k = a
1728          v = True
1729      else:
1730        k = a[:i]
1731        if k[-1] in "-+":
1732          op = k[-1]+'='  # Ops is '-=' or '+=' now.
1733          k = k[:-1]
1734        v = a[i+1:]
1735      k = k.replace('-', '_')
1736      if not hasattr(self, k):
1737        if ignore_unknown == True or k in ignore_unknown:
1738          ret.append(orig_a)
1739          continue
1740        else:
1741          raise self.UnknownOptionError("Unknown option '%s'" % a)
1742
1743      ov = getattr(self, k)
1744      if isinstance(ov, bool):
1745        v = bool(v)
1746      elif isinstance(ov, int):
1747        v = int(v)
1748      elif isinstance(ov, list):
1749        vv = v.split(',')
1750        if vv == ['']:
1751          vv = []
1752        vv = [int(x, 0) if len(x) and x[0] in "0123456789" else x for x in vv]
1753        if op == '=':
1754          v = vv
1755        elif op == '+=':
1756          v = ov
1757          v.extend(vv)
1758        elif op == '-=':
1759          v = ov
1760          for x in vv:
1761            if x in v:
1762              v.remove(x)
1763        else:
1764          assert 0
1765
1766      opts[k] = v
1767    self.set(**opts)
1768
1769    return ret
1770
1771
1772class Subsetter(object):
1773
1774  def __init__(self, options=None, log=None):
1775
1776    if not log:
1777      log = Logger()
1778    if not options:
1779      options = Options()
1780
1781    self.options = options
1782    self.log = log
1783    self.unicodes_requested = set()
1784    self.glyphs_requested = set()
1785    self.glyphs = set()
1786
1787  def populate(self, glyphs=[], unicodes=[], text=""):
1788    self.unicodes_requested.update(unicodes)
1789    if isinstance(text, str):
1790      text = text.decode("utf8")
1791    for u in text:
1792      self.unicodes_requested.add(ord(u))
1793    self.glyphs_requested.update(glyphs)
1794    self.glyphs.update(glyphs)
1795
1796  def _prune_pre_subset(self, font):
1797
1798    for tag in font.keys():
1799      if tag == 'GlyphOrder': continue
1800
1801      if(tag in self.options.drop_tables or
1802         (tag in self.options.hinting_tables and not self.options.hinting)):
1803        self.log(tag, "dropped")
1804        del font[tag]
1805        continue
1806
1807      clazz = ttLib.getTableClass(tag)
1808
1809      if hasattr(clazz, 'prune_pre_subset'):
1810        table = font[tag]
1811        self.log.lapse("load '%s'" % tag)
1812        retain = table.prune_pre_subset(self.options)
1813        self.log.lapse("prune  '%s'" % tag)
1814        if not retain:
1815          self.log(tag, "pruned to empty; dropped")
1816          del font[tag]
1817          continue
1818        else:
1819          self.log(tag, "pruned")
1820
1821  def _closure_glyphs(self, font):
1822
1823    self.glyphs = self.glyphs_requested.copy()
1824
1825    if 'cmap' in font:
1826      font['cmap'].closure_glyphs(self)
1827    self.glyphs_cmaped = self.glyphs
1828
1829    if self.options.notdef_glyph:
1830      if 'glyf' in font:
1831        self.glyphs.add(font.getGlyphName(0))
1832        self.log("Added gid0 to subset")
1833      else:
1834        self.glyphs.add('.notdef')
1835        self.log("Added .notdef to subset")
1836    if self.options.recommended_glyphs:
1837      if 'glyf' in font:
1838        for i in range(4):
1839          self.glyphs.add(font.getGlyphName(i))
1840        self.log("Added first four glyphs to subset")
1841
1842    if 'GSUB' in font:
1843      self.log("Closing glyph list over 'GSUB': %d glyphs before" %
1844                len(self.glyphs))
1845      self.log.glyphs(self.glyphs, font=font)
1846      font['GSUB'].closure_glyphs(self)
1847      self.log("Closed  glyph list over 'GSUB': %d glyphs after" %
1848                len(self.glyphs))
1849      self.log.glyphs(self.glyphs, font=font)
1850      self.log.lapse("close glyph list over 'GSUB'")
1851    self.glyphs_gsubed = self.glyphs.copy()
1852
1853    if 'glyf' in font:
1854      self.log("Closing glyph list over 'glyf': %d glyphs before" %
1855                len(self.glyphs))
1856      self.log.glyphs(self.glyphs, font=font)
1857      font['glyf'].closure_glyphs(self)
1858      self.log("Closed  glyph list over 'glyf': %d glyphs after" %
1859                len(self.glyphs))
1860      self.log.glyphs(self.glyphs, font=font)
1861      self.log.lapse("close glyph list over 'glyf'")
1862    self.glyphs_glyfed = self.glyphs.copy()
1863
1864    self.glyphs_all = self.glyphs.copy()
1865
1866    self.log("Retaining %d glyphs: " % len(self.glyphs_all))
1867
1868  def _subset_glyphs(self, font):
1869    for tag in font.keys():
1870      if tag == 'GlyphOrder': continue
1871      clazz = ttLib.getTableClass(tag)
1872
1873      if tag in self.options.no_subset_tables:
1874        self.log(tag, "subsetting not needed")
1875      elif hasattr(clazz, 'subset_glyphs'):
1876        table = font[tag]
1877        self.glyphs = self.glyphs_all
1878        retain = table.subset_glyphs(self)
1879        self.glyphs = self.glyphs_all
1880        self.log.lapse("subset '%s'" % tag)
1881        if not retain:
1882          self.log(tag, "subsetted to empty; dropped")
1883          del font[tag]
1884        else:
1885          self.log(tag, "subsetted")
1886      else:
1887        self.log(tag, "NOT subset; don't know how to subset; dropped")
1888        del font[tag]
1889
1890    glyphOrder = font.getGlyphOrder()
1891    glyphOrder = [g for g in glyphOrder if g in self.glyphs_all]
1892    font.setGlyphOrder(glyphOrder)
1893    font._buildReverseGlyphOrderDict()
1894    self.log.lapse("subset GlyphOrder")
1895
1896  def _prune_post_subset(self, font):
1897    for tag in font.keys():
1898      if tag == 'GlyphOrder': continue
1899      clazz = ttLib.getTableClass(tag)
1900      if hasattr(clazz, 'prune_post_subset'):
1901        table = font[tag]
1902        retain = table.prune_post_subset(self.options)
1903        self.log.lapse("prune  '%s'" % tag)
1904        if not retain:
1905          self.log(tag, "pruned to empty; dropped")
1906          del font[tag]
1907        else:
1908          self.log(tag, "pruned")
1909
1910  def subset(self, font):
1911
1912    self._prune_pre_subset(font)
1913    self._closure_glyphs(font)
1914    self._subset_glyphs(font)
1915    self._prune_post_subset(font)
1916
1917
1918class Logger(object):
1919
1920  def __init__(self, verbose=False, xml=False, timing=False):
1921    self.verbose = verbose
1922    self.xml = xml
1923    self.timing = timing
1924    self.last_time = self.start_time = time.time()
1925
1926  def parse_opts(self, argv):
1927    argv = argv[:]
1928    for v in ['verbose', 'xml', 'timing']:
1929      if "--"+v in argv:
1930        setattr(self, v, True)
1931        argv.remove("--"+v)
1932    return argv
1933
1934  def __call__(self, *things):
1935    if not self.verbose:
1936      return
1937    print ' '.join(str(x) for x in things)
1938
1939  def lapse(self, *things):
1940    if not self.timing:
1941      return
1942    new_time = time.time()
1943    print "Took %0.3fs to %s" %(new_time - self.last_time,
1944                                 ' '.join(str(x) for x in things))
1945    self.last_time = new_time
1946
1947  def glyphs(self, glyphs, font=None):
1948    self("Names: ", sorted(glyphs))
1949    if font:
1950      reverseGlyphMap = font.getReverseGlyphMap()
1951      self("Gids : ", sorted(reverseGlyphMap[g] for g in glyphs))
1952
1953  def font(self, font, file=sys.stdout):
1954    if not self.xml:
1955      return
1956    from fontTools.misc import xmlWriter
1957    writer = xmlWriter.XMLWriter(file)
1958    font.disassembleInstructions = False  # Work around ttLib bug
1959    for tag in font.keys():
1960      writer.begintag(tag)
1961      writer.newline()
1962      font[tag].toXML(writer, font)
1963      writer.endtag(tag)
1964      writer.newline()
1965
1966
1967def load_font(fontFile,
1968              options,
1969              checkChecksums=False,
1970              dontLoadGlyphNames=False):
1971
1972  font = ttLib.TTFont(fontFile,
1973                                checkChecksums=checkChecksums,
1974                                recalcBBoxes=options.recalc_bounds)
1975
1976  # Hack:
1977  #
1978  # If we don't need glyph names, change 'post' class to not try to
1979  # load them.  It avoid lots of headache with broken fonts as well
1980  # as loading time.
1981  #
1982  # Ideally ttLib should provide a way to ask it to skip loading
1983  # glyph names.  But it currently doesn't provide such a thing.
1984  #
1985  if dontLoadGlyphNames:
1986    post = ttLib.getTableClass('post')
1987    saved = post.decode_format_2_0
1988    post.decode_format_2_0 = post.decode_format_3_0
1989    f = font['post']
1990    if f.formatType == 2.0:
1991      f.formatType = 3.0
1992    post.decode_format_2_0 = saved
1993
1994  return font
1995
1996def save_font(font, outfile, options):
1997  if options.flavor and not hasattr(font, 'flavor'):
1998    raise Exception("fonttools version does not support flavors.")
1999  font.flavor = options.flavor
2000  font.save(outfile, reorderTables=options.canonical_order)
2001
2002def main(args):
2003
2004  log = Logger()
2005  args = log.parse_opts(args)
2006
2007  options = Options()
2008  args = options.parse_opts(args, ignore_unknown=['text'])
2009
2010  if len(args) < 2:
2011    print >>sys.stderr, "usage: pyftsubset font-file glyph... [--text=ABC]... [--option=value]..."
2012    sys.exit(1)
2013
2014  fontfile = args[0]
2015  args = args[1:]
2016
2017  dontLoadGlyphNames =(not options.glyph_names and
2018         all(any(g.startswith(p)
2019             for p in ['gid', 'glyph', 'uni', 'U+'])
2020              for g in args))
2021
2022  font = load_font(fontfile, options, dontLoadGlyphNames=dontLoadGlyphNames)
2023  subsetter = Subsetter(options=options, log=log)
2024  log.lapse("load font")
2025
2026  names = font.getGlyphNames()
2027  log.lapse("loading glyph names")
2028
2029  glyphs = []
2030  unicodes = []
2031  text = ""
2032  for g in args:
2033    if g == '*':
2034      glyphs.extend(font.getGlyphOrder())
2035      continue
2036    if g in names:
2037      glyphs.append(g)
2038      continue
2039    if g.startswith('--text='):
2040      text += g[7:]
2041      continue
2042    if g.startswith('uni') or g.startswith('U+'):
2043      if g.startswith('uni') and len(g) > 3:
2044        g = g[3:]
2045      elif g.startswith('U+') and len(g) > 2:
2046        g = g[2:]
2047      u = int(g, 16)
2048      unicodes.append(u)
2049      continue
2050    if g.startswith('gid') or g.startswith('glyph'):
2051      if g.startswith('gid') and len(g) > 3:
2052        g = g[3:]
2053      elif g.startswith('glyph') and len(g) > 5:
2054        g = g[5:]
2055      try:
2056        glyphs.append(font.getGlyphName(int(g), requireReal=1))
2057      except ValueError:
2058        raise Exception("Invalid glyph identifier: %s" % g)
2059      continue
2060    raise Exception("Invalid glyph identifier: %s" % g)
2061  log.lapse("compile glyph list")
2062  log("Unicodes:", unicodes)
2063  log("Glyphs:", glyphs)
2064
2065  subsetter.populate(glyphs=glyphs, unicodes=unicodes, text=text)
2066  subsetter.subset(font)
2067
2068  outfile = fontfile + '.subset'
2069
2070  save_font (font, outfile, options)
2071  log.lapse("compile and save font")
2072
2073  log.last_time = log.start_time
2074  log.lapse("make one with everything(TOTAL TIME)")
2075
2076  if log.verbose:
2077    import os
2078    log("Input  font: %d bytes" % os.path.getsize(fontfile))
2079    log("Subset font: %d bytes" % os.path.getsize(outfile))
2080
2081  log.font(font)
2082
2083  font.close()
2084
2085
2086__all__ = [
2087  'Options',
2088  'Subsetter',
2089  'Logger',
2090  'load_font',
2091  'save_font',
2092  'main'
2093]
2094
2095if __name__ == '__main__':
2096  main(sys.argv[1:])
2097