subset.py revision 5fd8779fd3188a550c1e73284b8d42f60c4e08b2
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_features(self):
1070  "Remove unreferenced featurs"
1071  if self.table.ScriptList:
1072    feature_indices = self.table.ScriptList.collect_features()
1073  else:
1074    feature_indices = []
1075  if self.table.FeatureList:
1076    self.table.FeatureList.subset_features(feature_indices)
1077  if self.table.ScriptList:
1078    self.table.ScriptList.subset_features(feature_indices)
1079
1080@_add_method(ttLib.getTableClass('GSUB'),
1081             ttLib.getTableClass('GPOS'))
1082def prune_pre_subset(self, options):
1083  # Drop undesired features
1084  if '*' not in options.layout_features:
1085    self.subset_feature_tags(options.layout_features)
1086  # Drop unreferenced lookups
1087  self.prune_lookups()
1088  # Prune lookups themselves
1089  if self.table.LookupList:
1090    self.table.LookupList.prune_pre_subset(options);
1091  return True
1092
1093@_add_method(ttLib.getTableClass('GSUB'),
1094             ttLib.getTableClass('GPOS'))
1095def remove_redundant_langsys(self):
1096  table = self.table
1097  if not table.ScriptList or not table.FeatureList:
1098    return
1099
1100  features = table.FeatureList.FeatureRecord
1101
1102  for s in table.ScriptList.ScriptRecord:
1103    d = s.Script.DefaultLangSys
1104    if not d:
1105      continue
1106    for lr in s.Script.LangSysRecord[:]:
1107      l = lr.LangSys
1108      # Compare d and l
1109      if len(d.FeatureIndex) != len(l.FeatureIndex):
1110        continue
1111      if (d.ReqFeatureIndex == 65535) != (l.ReqFeatureIndex == 65535):
1112        continue
1113
1114      if d.ReqFeatureIndex != 65535:
1115        if features[d.ReqFeatureIndex] != features[l.ReqFeatureIndex]:
1116          continue
1117
1118      for i in range(len(d.FeatureIndex)):
1119        if features[d.FeatureIndex[i]] != features[l.FeatureIndex[i]]:
1120          break
1121      else:
1122        # LangSys and default are equal; delete LangSys
1123        s.Script.LangSysRecord.remove(lr)
1124
1125@_add_method(ttLib.getTableClass('GSUB'),
1126             ttLib.getTableClass('GPOS'))
1127def prune_post_subset(self, options):
1128  table = self.table
1129
1130  # LookupList looks good.  Just prune lookups themselves
1131  if table.LookupList:
1132    table.LookupList.prune_post_subset(options);
1133    if not table.LookupList.Lookup:
1134      table.LookupList = None
1135
1136  if not table.LookupList:
1137    table.FeatureList = None
1138
1139  if table.FeatureList:
1140    self.remove_redundant_langsys()
1141    # Remove unreferenced features
1142    self.prune_features()
1143
1144  if table.FeatureList and not table.FeatureList.FeatureRecord:
1145    table.FeatureList = None
1146
1147  # Never drop scripts themselves as them just being available
1148  # holds semantic significance.
1149  if table.ScriptList and not table.ScriptList.ScriptRecord:
1150    table.ScriptList = None
1151
1152  return True
1153
1154@_add_method(ttLib.getTableClass('GDEF'))
1155def subset_glyphs(self, s):
1156  glyphs = s.glyphs_gsubed
1157  table = self.table
1158  if table.LigCaretList:
1159    indices = table.LigCaretList.Coverage.subset(glyphs)
1160    table.LigCaretList.LigGlyph = [table.LigCaretList.LigGlyph[i]
1161                                   for i in indices]
1162    table.LigCaretList.LigGlyphCount = len(table.LigCaretList.LigGlyph)
1163  if table.MarkAttachClassDef:
1164    table.MarkAttachClassDef.classDefs = dict((g,v) for g,v in
1165                                              table.MarkAttachClassDef.
1166                                                classDefs.items()
1167                                              if g in glyphs)
1168  if table.GlyphClassDef:
1169    table.GlyphClassDef.classDefs = dict((g,v) for g,v in
1170                                         table.GlyphClassDef.
1171                                           classDefs.items()
1172                                         if g in glyphs)
1173  if table.AttachList:
1174    indices = table.AttachList.Coverage.subset(glyphs)
1175    GlyphCount = table.AttachList.GlyphCount
1176    table.AttachList.AttachPoint = [table.AttachList.AttachPoint[i]
1177                                    for i in indices
1178                                    if i < GlyphCount]
1179    table.AttachList.GlyphCount = len(table.AttachList.AttachPoint)
1180  if hasattr(table, "MarkGlyphSetsDef") and table.MarkGlyphSetsDef:
1181    for coverage in table.MarkGlyphSetsDef.Coverage:
1182      coverage.subset(glyphs)
1183    # TODO: The following is disabled.  If enabling, we need to go fixup all
1184    # lookups that use MarkFilteringSet and map their set.
1185    #indices = table.MarkGlyphSetsDef.Coverage = [c for c in table.MarkGlyphSetsDef.Coverage if c.glyphs]
1186  return True
1187
1188@_add_method(ttLib.getTableClass('GDEF'))
1189def prune_post_subset(self, options):
1190  table = self.table
1191  if table.LigCaretList and not table.LigCaretList.LigGlyphCount:
1192    table.LigCaretList = None
1193  if table.MarkAttachClassDef and not table.MarkAttachClassDef.classDefs:
1194    table.MarkAttachClassDef = None
1195  if table.GlyphClassDef and not table.GlyphClassDef.classDefs:
1196    table.GlyphClassDef = None
1197  if table.AttachList and not table.AttachList.GlyphCount:
1198    table.AttachList = None
1199  if hasattr(table, "MarkGlyphSetsDef") and table.MarkGlyphSetsDef and not table.MarkGlyphSetsDef.Coverage:
1200    table.MarkGlyphSetsDef = None
1201    if table.Version == 0x00010002/0x10000:
1202      table.Version = 1.0
1203  return bool(table.LigCaretList or
1204              table.MarkAttachClassDef or
1205              table.GlyphClassDef or
1206              table.AttachList or
1207              (table.Version >= 0x00010002/0x10000 and table.MarkGlyphSetsDef))
1208
1209@_add_method(ttLib.getTableClass('kern'))
1210def prune_pre_subset(self, options):
1211  # Prune unknown kern table types
1212  self.kernTables = [t for t in self.kernTables if hasattr(t, 'kernTable')]
1213  return bool(self.kernTables)
1214
1215@_add_method(ttLib.getTableClass('kern'))
1216def subset_glyphs(self, s):
1217  glyphs = s.glyphs_gsubed
1218  for t in self.kernTables:
1219    t.kernTable = dict(((a,b),v) for (a,b),v in t.kernTable.items()
1220                       if a in glyphs and b in glyphs)
1221  self.kernTables = [t for t in self.kernTables if t.kernTable]
1222  return bool(self.kernTables)
1223
1224@_add_method(ttLib.getTableClass('vmtx'))
1225def subset_glyphs(self, s):
1226  self.metrics = dict((g,v) for g,v in self.metrics.items() if g in s.glyphs)
1227  return bool(self.metrics)
1228
1229@_add_method(ttLib.getTableClass('hmtx'))
1230def subset_glyphs(self, s):
1231  self.metrics = dict((g,v) for g,v in self.metrics.items() if g in s.glyphs)
1232  return True # Required table
1233
1234@_add_method(ttLib.getTableClass('hdmx'))
1235def subset_glyphs(self, s):
1236  self.hdmx = dict((sz,dict((g,v) for g,v in l.items() if g in s.glyphs))
1237                   for sz,l in self.hdmx.items())
1238  return bool(self.hdmx)
1239
1240@_add_method(ttLib.getTableClass('VORG'))
1241def subset_glyphs(self, s):
1242  self.VOriginRecords = dict((g,v) for g,v in self.VOriginRecords.items()
1243                             if g in s.glyphs)
1244  self.numVertOriginYMetrics = len(self.VOriginRecords)
1245  return True  # Never drop; has default metrics
1246
1247@_add_method(ttLib.getTableClass('post'))
1248def prune_pre_subset(self, options):
1249  if not options.glyph_names:
1250    self.formatType = 3.0
1251  return True # Required table
1252
1253@_add_method(ttLib.getTableClass('post'))
1254def subset_glyphs(self, s):
1255  self.extraNames = []  # This seems to do it
1256  return True # Required table
1257
1258@_add_method(ttLib.getTableModule('glyf').Glyph)
1259def remapComponentsFast(self, indices):
1260  if not self.data or struct.unpack(">h", self.data[:2])[0] >= 0:
1261    return  # Not composite
1262  data = array.array("B", self.data)
1263  i = 10
1264  more = 1
1265  while more:
1266    flags =(data[i] << 8) | data[i+1]
1267    glyphID =(data[i+2] << 8) | data[i+3]
1268    # Remap
1269    glyphID = indices.index(glyphID)
1270    data[i+2] = glyphID >> 8
1271    data[i+3] = glyphID & 0xFF
1272    i += 4
1273    flags = int(flags)
1274
1275    if flags & 0x0001: i += 4  # ARG_1_AND_2_ARE_WORDS
1276    else: i += 2
1277    if flags & 0x0008: i += 2  # WE_HAVE_A_SCALE
1278    elif flags & 0x0040: i += 4  # WE_HAVE_AN_X_AND_Y_SCALE
1279    elif flags & 0x0080: i += 8  # WE_HAVE_A_TWO_BY_TWO
1280    more = flags & 0x0020  # MORE_COMPONENTS
1281
1282  self.data = data.tostring()
1283
1284@_add_method(ttLib.getTableClass('glyf'))
1285def closure_glyphs(self, s):
1286  decompose = s.glyphs
1287  while True:
1288    components = set()
1289    for g in decompose:
1290      if g not in self.glyphs:
1291        continue
1292      gl = self.glyphs[g]
1293      for c in gl.getComponentNames(self):
1294        if c not in s.glyphs:
1295          components.add(c)
1296    components = set(c for c in components if c not in s.glyphs)
1297    if not components:
1298      break
1299    decompose = components
1300    s.glyphs.update(components)
1301
1302@_add_method(ttLib.getTableClass('glyf'))
1303def prune_pre_subset(self, options):
1304  if options.notdef_glyph and not options.notdef_outline:
1305    g = self[self.glyphOrder[0]]
1306    # Yay, easy!
1307    g.__dict__.clear()
1308    g.data = ""
1309  return True
1310
1311@_add_method(ttLib.getTableClass('glyf'))
1312def subset_glyphs(self, s):
1313  self.glyphs = dict((g,v) for g,v in self.glyphs.items() if g in s.glyphs)
1314  indices = [i for i,g in enumerate(self.glyphOrder) if g in s.glyphs]
1315  for v in self.glyphs.values():
1316    if hasattr(v, "data"):
1317      v.remapComponentsFast(indices)
1318    else:
1319      pass  # No need
1320  self.glyphOrder = [g for g in self.glyphOrder if g in s.glyphs]
1321  # Don't drop empty 'glyf' tables, otherwise 'loca' doesn't get subset.
1322  return True
1323
1324@_add_method(ttLib.getTableClass('glyf'))
1325def prune_post_subset(self, options):
1326  if not options.hinting:
1327    for v in self.glyphs.values():
1328      v.removeHinting()
1329  return True
1330
1331@_add_method(ttLib.getTableClass('CFF '))
1332def prune_pre_subset(self, options):
1333  cff = self.cff
1334  # CFF table must have one font only
1335  cff.fontNames = cff.fontNames[:1]
1336
1337  if options.notdef_glyph and not options.notdef_outline:
1338    for fontname in cff.keys():
1339      font = cff[fontname]
1340      c,_ = font.CharStrings.getItemAndSelector('.notdef')
1341      # XXX we should preserve the glyph width
1342      c.bytecode = '\x0e' # endchar
1343      c.program = None
1344
1345  return True # bool(cff.fontNames)
1346
1347@_add_method(ttLib.getTableClass('CFF '))
1348def subset_glyphs(self, s):
1349  cff = self.cff
1350  for fontname in cff.keys():
1351    font = cff[fontname]
1352    cs = font.CharStrings
1353
1354    # Load all glyphs
1355    for g in font.charset:
1356      if g not in s.glyphs: continue
1357      c,sel = cs.getItemAndSelector(g)
1358
1359    if cs.charStringsAreIndexed:
1360      indices = [i for i,g in enumerate(font.charset) if g in s.glyphs]
1361      csi = cs.charStringsIndex
1362      csi.items = [csi.items[i] for i in indices]
1363      csi.count = len(csi.items)
1364      del csi.file, csi.offsets
1365      if hasattr(font, "FDSelect"):
1366        sel = font.FDSelect
1367        sel.format = None
1368        sel.gidArray = [sel.gidArray[i] for i in indices]
1369      cs.charStrings = dict((g,indices.index(v))
1370                            for g,v in cs.charStrings.items()
1371                            if g in s.glyphs)
1372    else:
1373      cs.charStrings = dict((g,v)
1374                            for g,v in cs.charStrings.items()
1375                            if g in s.glyphs)
1376    font.charset = [g for g in font.charset if g in s.glyphs]
1377    font.numGlyphs = len(font.charset)
1378
1379  return True # any(cff[fontname].numGlyphs for fontname in cff.keys())
1380
1381@_add_method(psCharStrings.T2CharString)
1382def subset_subroutines(self, subrs, gsubrs):
1383  p = self.program
1384  assert len(p)
1385  for i in range(1, len(p)):
1386    if p[i] == 'callsubr':
1387      assert isinstance(p[i-1], int)
1388      p[i-1] = subrs._used.index(p[i-1] + subrs._old_bias) - subrs._new_bias
1389    elif p[i] == 'callgsubr':
1390      assert isinstance(p[i-1], int)
1391      p[i-1] = gsubrs._used.index(p[i-1] + gsubrs._old_bias) - gsubrs._new_bias
1392
1393@_add_method(psCharStrings.T2CharString)
1394def drop_hints(self):
1395  hints = self._hints
1396
1397  if hints.has_hint:
1398    self.program = self.program[hints.last_hint:]
1399    if hasattr(self, 'width'):
1400      # Insert width back if needed
1401      if self.width != self.private.defaultWidthX:
1402        self.program.insert(0, self.width - self.private.nominalWidthX)
1403
1404  if hints.has_hintmask:
1405    i = 0
1406    p = self.program
1407    while i < len(p):
1408      if p[i] in ['hintmask', 'cntrmask']:
1409        assert i + 1 <= len(p)
1410        del p[i:i+2]
1411        continue
1412      i += 1
1413
1414  # TODO: we currently don't drop calls to "empty" subroutines.
1415
1416  assert len(self.program)
1417
1418  del self._hints
1419
1420class _MarkingT2Decompiler(psCharStrings.SimpleT2Decompiler):
1421
1422  def __init__(self, localSubrs, globalSubrs):
1423    psCharStrings.SimpleT2Decompiler.__init__(self,
1424                                              localSubrs,
1425                                              globalSubrs)
1426    for subrs in [localSubrs, globalSubrs]:
1427      if subrs and not hasattr(subrs, "_used"):
1428        subrs._used = set()
1429
1430  def op_callsubr(self, index):
1431    self.localSubrs._used.add(self.operandStack[-1]+self.localBias)
1432    psCharStrings.SimpleT2Decompiler.op_callsubr(self, index)
1433
1434  def op_callgsubr(self, index):
1435    self.globalSubrs._used.add(self.operandStack[-1]+self.globalBias)
1436    psCharStrings.SimpleT2Decompiler.op_callgsubr(self, index)
1437
1438class _DehintingT2Decompiler(psCharStrings.SimpleT2Decompiler):
1439
1440  class Hints(object):
1441    def __init__(self):
1442      # Whether calling this charstring produces any hint stems
1443      self.has_hint = False
1444      # Index to start at to drop all hints
1445      self.last_hint = 0
1446      # Index up to which we know more hints are possible.  Only
1447      # relevant if status is 0 or 1.
1448      self.last_checked = 0
1449      # The status means:
1450      # 0: after dropping hints, this charstring is empty
1451      # 1: after dropping hints, there may be more hints continuing after this
1452      # 2: no more hints possible after this charstring
1453      self.status = 0
1454      # Has hintmask instructions; not recursive
1455      self.has_hintmask = False
1456    pass
1457
1458  def __init__(self, css, localSubrs, globalSubrs):
1459    self._css = css
1460    psCharStrings.SimpleT2Decompiler.__init__(self,
1461                                              localSubrs,
1462                                              globalSubrs)
1463
1464  def execute(self, charString):
1465    old_hints = charString._hints if hasattr(charString, '_hints') else None
1466    charString._hints = self.Hints()
1467
1468    psCharStrings.SimpleT2Decompiler.execute(self, charString)
1469
1470    hints = charString._hints
1471
1472    if hints.has_hint or hints.has_hintmask:
1473      self._css.add(charString)
1474
1475    if hints.status != 2:
1476      # Check from last_check, make sure we didn't have any operators.
1477      for i in range(hints.last_checked, len(charString.program) - 1):
1478        if isinstance(charString.program[i], str):
1479          hints.status = 2
1480          break;
1481        else:
1482          hints.status = 1 # There's *something* here
1483      hints.last_checked = len(charString.program)
1484
1485    if old_hints:
1486      assert hints.__dict__ == old_hints.__dict__
1487
1488  def op_callsubr(self, index):
1489    subr = self.localSubrs[self.operandStack[-1]+self.localBias]
1490    psCharStrings.SimpleT2Decompiler.op_callsubr(self, index)
1491    self.processSubr(index, subr)
1492
1493  def op_callgsubr(self, index):
1494    subr = self.globalSubrs[self.operandStack[-1]+self.globalBias]
1495    psCharStrings.SimpleT2Decompiler.op_callgsubr(self, index)
1496    self.processSubr(index, subr)
1497
1498  def op_hstem(self, index):
1499    psCharStrings.SimpleT2Decompiler.op_hstem(self, index)
1500    self.processHint(index)
1501  def op_vstem(self, index):
1502    psCharStrings.SimpleT2Decompiler.op_vstem(self, index)
1503    self.processHint(index)
1504  def op_hstemhm(self, index):
1505    psCharStrings.SimpleT2Decompiler.op_hstemhm(self, index)
1506    self.processHint(index)
1507  def op_vstemhm(self, index):
1508    psCharStrings.SimpleT2Decompiler.op_vstemhm(self, index)
1509    self.processHint(index)
1510  def op_hintmask(self, index):
1511    psCharStrings.SimpleT2Decompiler.op_hintmask(self, index)
1512    self.processHintmask(index)
1513  def op_cntrmask(self, index):
1514    psCharStrings.SimpleT2Decompiler.op_cntrmask(self, index)
1515    self.processHintmask(index)
1516
1517  def processHintmask(self, index):
1518    cs = self.callingStack[-1]
1519    hints = cs._hints
1520    hints.has_hintmask = True
1521    if hints.status != 2 and hints.has_hint:
1522      # Check from last_check, see if we may be an implicit vstem
1523      for i in range(hints.last_checked, index - 1):
1524        if isinstance(cs.program[i], str):
1525          hints.status = 2
1526          break;
1527      if hints.status != 2:
1528        # We are an implicit vstem
1529        hints.last_hint = index + 1
1530        hints.status = 0
1531    hints.last_checked = index + 1
1532
1533  def processHint(self, index):
1534    cs = self.callingStack[-1]
1535    hints = cs._hints
1536    hints.has_hint = True
1537    hints.last_hint = index
1538    hints.last_checked = index
1539
1540  def processSubr(self, index, subr):
1541    cs = self.callingStack[-1]
1542    hints = cs._hints
1543    subr_hints = subr._hints
1544
1545    if subr_hints.has_hint:
1546      if hints.status != 2:
1547        hints.has_hint = True
1548        hints.last_checked = index
1549        hints.status = subr_hints.status
1550        # Decide where to chop off from
1551        if subr_hints.status == 0:
1552          hints.last_hint = index
1553        else:
1554          hints.last_hint = index - 2 # Leave the subr call in
1555      else:
1556        # In my understanding, this is a font bug.  Ie. it has hint stems
1557        # *after* path construction.  I've seen this in widespread fonts.
1558        # Best to ignore the hints I suppose...
1559        pass
1560        #assert 0
1561    else:
1562      hints.status = max(hints.status, subr_hints.status)
1563      if hints.status != 2:
1564        # Check from last_check, make sure we didn't have
1565        # any operators.
1566        for i in range(hints.last_checked, index - 1):
1567          if isinstance(cs.program[i], str):
1568            hints.status = 2
1569            break;
1570        hints.last_checked = index
1571      if hints.status != 2:
1572        # Decide where to chop off from
1573        if subr_hints.status == 0:
1574          hints.last_hint = index
1575        else:
1576          hints.last_hint = index - 2 # Leave the subr call in
1577
1578@_add_method(ttLib.getTableClass('CFF '))
1579def prune_post_subset(self, options):
1580  cff = self.cff
1581  for fontname in cff.keys():
1582    font = cff[fontname]
1583    cs = font.CharStrings
1584
1585
1586    #
1587    # Drop unused FontDictionaries
1588    #
1589    if hasattr(font, "FDSelect"):
1590      sel = font.FDSelect
1591      indices = _uniq_sort(sel.gidArray)
1592      sel.gidArray = [indices.index (ss) for ss in sel.gidArray]
1593      arr = font.FDArray
1594      arr.items = [arr[i] for i in indices]
1595      arr.count = len(arr.items)
1596      del arr.file, arr.offsets
1597
1598
1599    #
1600    # Drop hints if not needed
1601    #
1602    if not options.hinting:
1603
1604      #
1605      # This can be tricky, but doesn't have to.  What we do is:
1606      #
1607      # - Run all used glyph charstrings and recurse into subroutines,
1608      # - For each charstring (including subroutines), if it has any
1609      #   of the hint stem operators, we mark it as such.  Upon returning,
1610      #   for each charstring we note all the subroutine calls it makes
1611      #   that (recursively) contain a stem,
1612      # - Dropping hinting then consists of the following two ops:
1613      #   * Drop the piece of the program in each charstring before the
1614      #     last call to a stem op or a stem-calling subroutine,
1615      #   * Drop all hintmask operations.
1616      # - It's trickier... A hintmask right after hints and a few numbers
1617      #   will act as an implicit vstemhm.  As such, we track whether
1618      #   we have seen any non-hint operators so far and do the right
1619      #   thing, recursively...  Good luck understanding that :(
1620      #
1621      css = set()
1622      for g in font.charset:
1623        c,sel = cs.getItemAndSelector(g)
1624        # Make sure it's decompiled.  We want our "decompiler" to walk
1625        # the program, not the bytecode.
1626        c.draw(basePen.NullPen())
1627        subrs = getattr(c.private, "Subrs", [])
1628        decompiler = _DehintingT2Decompiler(css, subrs, c.globalSubrs)
1629        decompiler.execute(c)
1630      for charstring in css:
1631        charstring.drop_hints()
1632
1633      # Drop font-wide hinting values
1634      all_privs = []
1635      if hasattr(font, 'FDSelect'):
1636        all_privs.extend(fd.Private for fd in font.FDArray)
1637      else:
1638        all_privs.append(font.Private)
1639      for priv in all_privs:
1640        for k in ['BlueValues', 'OtherBlues', 'FamilyBlues', 'FamilyOtherBlues',
1641                  'BlueScale', 'BlueShift', 'BlueFuzz',
1642                  'StemSnapH', 'StemSnapV', 'StdHW', 'StdVW']:
1643          if hasattr(priv, k):
1644            setattr(priv, k, None)
1645
1646
1647    #
1648    # Renumber subroutines to remove unused ones
1649    #
1650
1651    # Mark all used subroutines
1652    for g in font.charset:
1653      c,sel = cs.getItemAndSelector(g)
1654      subrs = getattr(c.private, "Subrs", [])
1655      decompiler = _MarkingT2Decompiler(subrs, c.globalSubrs)
1656      decompiler.execute(c)
1657
1658    all_subrs = [font.GlobalSubrs]
1659    if hasattr(font, 'FDSelect'):
1660      all_subrs.extend(fd.Private.Subrs for fd in font.FDArray if hasattr(fd.Private, 'Subrs') and fd.Private.Subrs)
1661    elif hasattr(font.Private, 'Subrs') and font.Private.Subrs:
1662      all_subrs.append(font.Private.Subrs)
1663
1664    subrs = set(subrs) # Remove duplicates
1665
1666    # Prepare
1667    for subrs in all_subrs:
1668      if not hasattr(subrs, '_used'):
1669        subrs._used = set()
1670      subrs._used = _uniq_sort(subrs._used)
1671      subrs._old_bias = psCharStrings.calcSubrBias(subrs)
1672      subrs._new_bias = psCharStrings.calcSubrBias(subrs._used)
1673
1674    # Renumber glyph charstrings
1675    for g in font.charset:
1676      c,sel = cs.getItemAndSelector(g)
1677      subrs = getattr(c.private, "Subrs", [])
1678      c.subset_subroutines (subrs, font.GlobalSubrs)
1679
1680    # Renumber subroutines themselves
1681    for subrs in all_subrs:
1682
1683      if subrs == font.GlobalSubrs:
1684        if not hasattr(font, 'FDSelect') and hasattr(font.Private, 'Subrs'):
1685          local_subrs = font.Private.Subrs
1686        else:
1687          local_subrs = []
1688      else:
1689        local_subrs = subrs
1690
1691      subrs.items = [subrs.items[i] for i in subrs._used]
1692      subrs.count = len(subrs.items)
1693      del subrs.file
1694      if hasattr(subrs, 'offsets'):
1695        del subrs.offsets
1696
1697      for i in range (subrs.count):
1698        subrs[i].subset_subroutines (local_subrs, font.GlobalSubrs)
1699
1700    # Cleanup
1701    for subrs in all_subrs:
1702      del subrs._used, subrs._old_bias, subrs._new_bias
1703
1704  return True
1705
1706@_add_method(ttLib.getTableClass('cmap'))
1707def closure_glyphs(self, s):
1708  tables = [t for t in self.tables
1709            if t.platformID == 3 and t.platEncID in [1, 10]]
1710  for u in s.unicodes_requested:
1711    found = False
1712    for table in tables:
1713      if u in table.cmap:
1714        s.glyphs.add(table.cmap[u])
1715        found = True
1716        break
1717    if not found:
1718      s.log("No glyph for Unicode value %s; skipping." % u)
1719
1720@_add_method(ttLib.getTableClass('cmap'))
1721def prune_pre_subset(self, options):
1722  if not options.legacy_cmap:
1723    # Drop non-Unicode / non-Symbol cmaps
1724    self.tables = [t for t in self.tables
1725                   if t.platformID == 3 and t.platEncID in [0, 1, 10]]
1726  if not options.symbol_cmap:
1727    self.tables = [t for t in self.tables
1728                   if t.platformID == 3 and t.platEncID in [1, 10]]
1729  # TODO(behdad) Only keep one subtable?
1730  # For now, drop format=0 which can't be subset_glyphs easily?
1731  self.tables = [t for t in self.tables if t.format != 0]
1732  self.numSubTables = len(self.tables)
1733  return True # Required table
1734
1735@_add_method(ttLib.getTableClass('cmap'))
1736def subset_glyphs(self, s):
1737  s.glyphs = s.glyphs_cmaped
1738  for t in self.tables:
1739    # For reasons I don't understand I need this here
1740    # to force decompilation of the cmap format 14.
1741    try:
1742      getattr(t, "asdf")
1743    except AttributeError:
1744      pass
1745    if t.format == 14:
1746      # TODO(behdad) XXX We drop all the default-UVS mappings(g==None).
1747      t.uvsDict = dict((v,[(u,g) for u,g in l if g in s.glyphs])
1748                       for v,l in t.uvsDict.items())
1749      t.uvsDict = dict((v,l) for v,l in t.uvsDict.items() if l)
1750    else:
1751      t.cmap = dict((u,g) for u,g in t.cmap.items()
1752                    if g in s.glyphs_requested or u in s.unicodes_requested)
1753  self.tables = [t for t in self.tables
1754                 if (t.cmap if t.format != 14 else t.uvsDict)]
1755  self.numSubTables = len(self.tables)
1756  # TODO(behdad) Convert formats when needed.
1757  # In particular, if we have a format=12 without non-BMP
1758  # characters, either drop format=12 one or convert it
1759  # to format=4 if there's not one.
1760  return True # Required table
1761
1762@_add_method(ttLib.getTableClass('name'))
1763def prune_pre_subset(self, options):
1764  if '*' not in options.name_IDs:
1765    self.names = [n for n in self.names if n.nameID in options.name_IDs]
1766  if not options.name_legacy:
1767    self.names = [n for n in self.names
1768                  if n.platformID == 3 and n.platEncID == 1]
1769  if '*' not in options.name_languages:
1770    self.names = [n for n in self.names if n.langID in options.name_languages]
1771  return True  # Required table
1772
1773
1774# TODO(behdad) OS/2 ulUnicodeRange / ulCodePageRange?
1775# TODO(behdad) Drop AAT tables.
1776# TODO(behdad) Drop unneeded GSUB/GPOS Script/LangSys entries.
1777# TODO(behdad) Drop empty GSUB/GPOS, and GDEF if no GSUB/GPOS left
1778# TODO(behdad) Drop GDEF subitems if unused by lookups
1779# TODO(behdad) Avoid recursing too much (in GSUB/GPOS and in CFF)
1780# TODO(behdad) Text direction considerations.
1781# TODO(behdad) Text script / language considerations.
1782# TODO(behdad) Optionally drop 'kern' table if GPOS available
1783
1784class Options(object):
1785
1786  class UnknownOptionError(Exception):
1787    pass
1788
1789  _drop_tables_default = ['BASE', 'JSTF', 'DSIG', 'EBDT', 'EBLC', 'EBSC', 'SVG ',
1790                          'PCLT', 'LTSH']
1791  _drop_tables_default += ['Feat', 'Glat', 'Gloc', 'Silf', 'Sill']  # Graphite
1792  _drop_tables_default += ['CBLC', 'CBDT', 'sbix', 'COLR', 'CPAL']  # Color
1793  _no_subset_tables_default = ['gasp', 'head', 'hhea', 'maxp', 'vhea', 'OS/2',
1794                               'loca', 'name', 'cvt ', 'fpgm', 'prep']
1795  _hinting_tables_default = ['cvt ', 'fpgm', 'prep', 'hdmx', 'VDMX']
1796
1797  # Based on HarfBuzz shapers
1798  _layout_features_groups = {
1799    # Default shaper
1800    'common': ['ccmp', 'liga', 'locl', 'mark', 'mkmk', 'rlig'],
1801    'horizontal': ['calt', 'clig', 'curs', 'kern', 'rclt'],
1802    'vertical':  ['valt', 'vert', 'vkrn', 'vpal', 'vrt2'],
1803    'ltr': ['ltra', 'ltrm'],
1804    'rtl': ['rtla', 'rtlm'],
1805    # Complex shapers
1806    'arabic': ['init', 'medi', 'fina', 'isol', 'med2', 'fin2', 'fin3',
1807               'cswh', 'mset'],
1808    'hangul': ['ljmo', 'vjmo', 'tjmo'],
1809    'tibetan': ['abvs', 'blws', 'abvm', 'blwm'],
1810    'indic': ['nukt', 'akhn', 'rphf', 'rkrf', 'pref', 'blwf', 'half',
1811              'abvf', 'pstf', 'cfar', 'vatu', 'cjct', 'init', 'pres',
1812              'abvs', 'blws', 'psts', 'haln', 'dist', 'abvm', 'blwm'],
1813  }
1814  _layout_features_default = _uniq_sort(sum(
1815      iter(_layout_features_groups.values()), []))
1816
1817  drop_tables = _drop_tables_default
1818  no_subset_tables = _no_subset_tables_default
1819  hinting_tables = _hinting_tables_default
1820  layout_features = _layout_features_default
1821  hinting = True
1822  glyph_names = False
1823  legacy_cmap = False
1824  symbol_cmap = False
1825  name_IDs = [1, 2]  # Family and Style
1826  name_legacy = False
1827  name_languages = [0x0409]  # English
1828  notdef_glyph = True # gid0 for TrueType / .notdef for CFF
1829  notdef_outline = False # No need for notdef to have an outline really
1830  recommended_glyphs = False  # gid1, gid2, gid3 for TrueType
1831  recalc_bounds = False # Recalculate font bounding boxes
1832  canonical_order = False # Order tables as recommended
1833  flavor = None # May be 'woff'
1834
1835  def __init__(self, **kwargs):
1836
1837    self.set(**kwargs)
1838
1839  def set(self, **kwargs):
1840    for k,v in kwargs.items():
1841      if not hasattr(self, k):
1842        raise self.UnknownOptionError("Unknown option '%s'" % k)
1843      setattr(self, k, v)
1844
1845  def parse_opts(self, argv, ignore_unknown=False):
1846    ret = []
1847    opts = {}
1848    for a in argv:
1849      orig_a = a
1850      if not a.startswith('--'):
1851        ret.append(a)
1852        continue
1853      a = a[2:]
1854      i = a.find('=')
1855      op = '='
1856      if i == -1:
1857        if a.startswith("no-"):
1858          k = a[3:]
1859          v = False
1860        else:
1861          k = a
1862          v = True
1863      else:
1864        k = a[:i]
1865        if k[-1] in "-+":
1866          op = k[-1]+'='  # Ops is '-=' or '+=' now.
1867          k = k[:-1]
1868        v = a[i+1:]
1869      k = k.replace('-', '_')
1870      if not hasattr(self, k):
1871        if ignore_unknown is True or k in ignore_unknown:
1872          ret.append(orig_a)
1873          continue
1874        else:
1875          raise self.UnknownOptionError("Unknown option '%s'" % a)
1876
1877      ov = getattr(self, k)
1878      if isinstance(ov, bool):
1879        v = bool(v)
1880      elif isinstance(ov, int):
1881        v = int(v)
1882      elif isinstance(ov, list):
1883        vv = v.split(',')
1884        if vv == ['']:
1885          vv = []
1886        vv = [int(x, 0) if len(x) and x[0] in "0123456789" else x for x in vv]
1887        if op == '=':
1888          v = vv
1889        elif op == '+=':
1890          v = ov
1891          v.extend(vv)
1892        elif op == '-=':
1893          v = ov
1894          for x in vv:
1895            if x in v:
1896              v.remove(x)
1897        else:
1898          assert False
1899
1900      opts[k] = v
1901    self.set(**opts)
1902
1903    return ret
1904
1905
1906class Subsetter(object):
1907
1908  def __init__(self, options=None, log=None):
1909
1910    if not log:
1911      log = Logger()
1912    if not options:
1913      options = Options()
1914
1915    self.options = options
1916    self.log = log
1917    self.unicodes_requested = set()
1918    self.glyphs_requested = set()
1919    self.glyphs = set()
1920
1921  def populate(self, glyphs=[], unicodes=[], text=""):
1922    self.unicodes_requested.update(unicodes)
1923    if isinstance(text, bytes):
1924      text = text.decode("utf8")
1925    for u in text:
1926      self.unicodes_requested.add(ord(u))
1927    self.glyphs_requested.update(glyphs)
1928    self.glyphs.update(glyphs)
1929
1930  def _prune_pre_subset(self, font):
1931
1932    for tag in font.keys():
1933      if tag == 'GlyphOrder': continue
1934
1935      if(tag in self.options.drop_tables or
1936         (tag in self.options.hinting_tables and not self.options.hinting)):
1937        self.log(tag, "dropped")
1938        del font[tag]
1939        continue
1940
1941      clazz = ttLib.getTableClass(tag)
1942
1943      if hasattr(clazz, 'prune_pre_subset'):
1944        table = font[tag]
1945        self.log.lapse("load '%s'" % tag)
1946        retain = table.prune_pre_subset(self.options)
1947        self.log.lapse("prune  '%s'" % tag)
1948        if not retain:
1949          self.log(tag, "pruned to empty; dropped")
1950          del font[tag]
1951          continue
1952        else:
1953          self.log(tag, "pruned")
1954
1955  def _closure_glyphs(self, font):
1956
1957    realGlyphs = set(font.getGlyphOrder())
1958
1959    self.glyphs = self.glyphs_requested.copy()
1960
1961    if 'cmap' in font:
1962      font['cmap'].closure_glyphs(self)
1963      self.glyphs.intersection_update(realGlyphs)
1964    self.glyphs_cmaped = self.glyphs
1965
1966    if self.options.notdef_glyph:
1967      if 'glyf' in font:
1968        self.glyphs.add(font.getGlyphName(0))
1969        self.log("Added gid0 to subset")
1970      else:
1971        self.glyphs.add('.notdef')
1972        self.log("Added .notdef to subset")
1973    if self.options.recommended_glyphs:
1974      if 'glyf' in font:
1975        for i in range(min(4, len(font.getGlyphOrder()))):
1976          self.glyphs.add(font.getGlyphName(i))
1977        self.log("Added first four glyphs to subset")
1978
1979    if 'GSUB' in font:
1980      self.log("Closing glyph list over 'GSUB': %d glyphs before" %
1981                len(self.glyphs))
1982      self.log.glyphs(self.glyphs, font=font)
1983      font['GSUB'].closure_glyphs(self)
1984      self.glyphs.intersection_update(realGlyphs)
1985      self.log("Closed  glyph list over 'GSUB': %d glyphs after" %
1986                len(self.glyphs))
1987      self.log.glyphs(self.glyphs, font=font)
1988      self.log.lapse("close glyph list over 'GSUB'")
1989    self.glyphs_gsubed = self.glyphs.copy()
1990
1991    if 'glyf' in font:
1992      self.log("Closing glyph list over 'glyf': %d glyphs before" %
1993                len(self.glyphs))
1994      self.log.glyphs(self.glyphs, font=font)
1995      font['glyf'].closure_glyphs(self)
1996      self.glyphs.intersection_update(realGlyphs)
1997      self.log("Closed  glyph list over 'glyf': %d glyphs after" %
1998                len(self.glyphs))
1999      self.log.glyphs(self.glyphs, font=font)
2000      self.log.lapse("close glyph list over 'glyf'")
2001    self.glyphs_glyfed = self.glyphs.copy()
2002
2003    self.glyphs_all = self.glyphs.copy()
2004
2005    self.log("Retaining %d glyphs: " % len(self.glyphs_all))
2006
2007    del self.glyphs
2008
2009
2010  def _subset_glyphs(self, font):
2011    for tag in font.keys():
2012      if tag == 'GlyphOrder': continue
2013      clazz = ttLib.getTableClass(tag)
2014
2015      if tag in self.options.no_subset_tables:
2016        self.log(tag, "subsetting not needed")
2017      elif hasattr(clazz, 'subset_glyphs'):
2018        table = font[tag]
2019        self.glyphs = self.glyphs_all
2020        retain = table.subset_glyphs(self)
2021        del self.glyphs
2022        self.log.lapse("subset '%s'" % tag)
2023        if not retain:
2024          self.log(tag, "subsetted to empty; dropped")
2025          del font[tag]
2026        else:
2027          self.log(tag, "subsetted")
2028      else:
2029        self.log(tag, "NOT subset; don't know how to subset; dropped")
2030        del font[tag]
2031
2032    glyphOrder = font.getGlyphOrder()
2033    glyphOrder = [g for g in glyphOrder if g in self.glyphs_all]
2034    font.setGlyphOrder(glyphOrder)
2035    font._buildReverseGlyphOrderDict()
2036    self.log.lapse("subset GlyphOrder")
2037
2038  def _prune_post_subset(self, font):
2039    for tag in font.keys():
2040      if tag == 'GlyphOrder': continue
2041      clazz = ttLib.getTableClass(tag)
2042      if hasattr(clazz, 'prune_post_subset'):
2043        table = font[tag]
2044        retain = table.prune_post_subset(self.options)
2045        self.log.lapse("prune  '%s'" % tag)
2046        if not retain:
2047          self.log(tag, "pruned to empty; dropped")
2048          del font[tag]
2049        else:
2050          self.log(tag, "pruned")
2051
2052  def subset(self, font):
2053
2054    self._prune_pre_subset(font)
2055    self._closure_glyphs(font)
2056    self._subset_glyphs(font)
2057    self._prune_post_subset(font)
2058
2059
2060class Logger(object):
2061
2062  def __init__(self, verbose=False, xml=False, timing=False):
2063    self.verbose = verbose
2064    self.xml = xml
2065    self.timing = timing
2066    self.last_time = self.start_time = time.time()
2067
2068  def parse_opts(self, argv):
2069    argv = argv[:]
2070    for v in ['verbose', 'xml', 'timing']:
2071      if "--"+v in argv:
2072        setattr(self, v, True)
2073        argv.remove("--"+v)
2074    return argv
2075
2076  def __call__(self, *things):
2077    if not self.verbose:
2078      return
2079    print(' '.join(str(x) for x in things))
2080
2081  def lapse(self, *things):
2082    if not self.timing:
2083      return
2084    new_time = time.time()
2085    print("Took %0.3fs to %s" %(new_time - self.last_time,
2086                                 ' '.join(str(x) for x in things)))
2087    self.last_time = new_time
2088
2089  def glyphs(self, glyphs, font=None):
2090    if not self.verbose:
2091      return
2092    self("Names: ", sorted(glyphs))
2093    if font:
2094      reverseGlyphMap = font.getReverseGlyphMap()
2095      self("Gids : ", sorted(reverseGlyphMap[g] for g in glyphs))
2096
2097  def font(self, font, file=sys.stdout):
2098    if not self.xml:
2099      return
2100    from fontTools.misc import xmlWriter
2101    writer = xmlWriter.XMLWriter(file)
2102    for tag in font.keys():
2103      writer.begintag(tag)
2104      writer.newline()
2105      font[tag].toXML(writer, font)
2106      writer.endtag(tag)
2107      writer.newline()
2108
2109
2110def load_font(fontFile,
2111              options,
2112              allowVID=False,
2113              checkChecksums=False,
2114              dontLoadGlyphNames=False):
2115
2116  font = ttLib.TTFont(fontFile,
2117                      allowVID=allowVID,
2118                      checkChecksums=checkChecksums,
2119                      recalcBBoxes=options.recalc_bounds)
2120
2121  # Hack:
2122  #
2123  # If we don't need glyph names, change 'post' class to not try to
2124  # load them.  It avoid lots of headache with broken fonts as well
2125  # as loading time.
2126  #
2127  # Ideally ttLib should provide a way to ask it to skip loading
2128  # glyph names.  But it currently doesn't provide such a thing.
2129  #
2130  if dontLoadGlyphNames:
2131    post = ttLib.getTableClass('post')
2132    saved = post.decode_format_2_0
2133    post.decode_format_2_0 = post.decode_format_3_0
2134    f = font['post']
2135    if f.formatType == 2.0:
2136      f.formatType = 3.0
2137    post.decode_format_2_0 = saved
2138
2139  return font
2140
2141def save_font(font, outfile, options):
2142  if options.flavor and not hasattr(font, 'flavor'):
2143    raise Exception("fonttools version does not support flavors.")
2144  font.flavor = options.flavor
2145  font.save(outfile, reorderTables=options.canonical_order)
2146
2147def main(args):
2148
2149  log = Logger()
2150  args = log.parse_opts(args)
2151
2152  options = Options()
2153  args = options.parse_opts(args, ignore_unknown=['text'])
2154
2155  if len(args) < 2:
2156    print("usage: pyftsubset font-file glyph... [--text=ABC]... [--option=value]...", file=sys.stderr)
2157    sys.exit(1)
2158
2159  fontfile = args[0]
2160  args = args[1:]
2161
2162  dontLoadGlyphNames =(not options.glyph_names and
2163         all(any(g.startswith(p)
2164             for p in ['gid', 'glyph', 'uni', 'U+'])
2165              for g in args))
2166
2167  font = load_font(fontfile, options, dontLoadGlyphNames=dontLoadGlyphNames)
2168  subsetter = Subsetter(options=options, log=log)
2169  log.lapse("load font")
2170
2171  names = font.getGlyphNames()
2172  log.lapse("loading glyph names")
2173
2174  glyphs = []
2175  unicodes = []
2176  text = ""
2177  for g in args:
2178    if g == '*':
2179      glyphs.extend(font.getGlyphOrder())
2180      continue
2181    if g in names:
2182      glyphs.append(g)
2183      continue
2184    if g.startswith('--text='):
2185      text += g[7:]
2186      continue
2187    if g.startswith('uni') or g.startswith('U+'):
2188      if g.startswith('uni') and len(g) > 3:
2189        g = g[3:]
2190      elif g.startswith('U+') and len(g) > 2:
2191        g = g[2:]
2192      u = int(g, 16)
2193      unicodes.append(u)
2194      continue
2195    if g.startswith('gid') or g.startswith('glyph'):
2196      if g.startswith('gid') and len(g) > 3:
2197        g = g[3:]
2198      elif g.startswith('glyph') and len(g) > 5:
2199        g = g[5:]
2200      try:
2201        glyphs.append(font.getGlyphName(int(g), requireReal=True))
2202      except ValueError:
2203        raise Exception("Invalid glyph identifier: %s" % g)
2204      continue
2205    raise Exception("Invalid glyph identifier: %s" % g)
2206  log.lapse("compile glyph list")
2207  log("Unicodes:", unicodes)
2208  log("Glyphs:", glyphs)
2209
2210  subsetter.populate(glyphs=glyphs, unicodes=unicodes, text=text)
2211  subsetter.subset(font)
2212
2213  outfile = fontfile + '.subset'
2214
2215  save_font (font, outfile, options)
2216  log.lapse("compile and save font")
2217
2218  log.last_time = log.start_time
2219  log.lapse("make one with everything(TOTAL TIME)")
2220
2221  if log.verbose:
2222    import os
2223    log("Input  font: %d bytes" % os.path.getsize(fontfile))
2224    log("Subset font: %d bytes" % os.path.getsize(outfile))
2225
2226  log.font(font)
2227
2228  font.close()
2229
2230
2231__all__ = [
2232  'Options',
2233  'Subsetter',
2234  'Logger',
2235  'load_font',
2236  'save_font',
2237  'main'
2238]
2239
2240if __name__ == '__main__':
2241  main(sys.argv[1:])
2242