Class: Fontisan::Subset::TableStrategy::Cff2

Inherits:
Object
  • Object
show all
Defined in:
lib/fontisan/subset/table_strategy/cff2.rb

Overview

CFF2 (Compact Font Format 2) subsetter strategy.

Subsets CFF2 tables by filtering the CharStrings INDEX and FDSelect to the subset glyphs, then rebuilding the table with patched offsets. The Variable Store (ItemVariationStore) is preserved unchanged — its vsindex references in charstrings remain valid because charstrings are carried as opaque bytes.

Unlike the CFF1 subsetter (which routes through UFO), CFF2 uses a standalone binary filter because CFF2 charstrings carry blend/vsindex operators for variation that the UFO round-trip would strip.

Layout of the rebuilt table:

Header (5 bytes)
Top DICT               (offsets patched to new positions)
Global Subr INDEX      (preserved as-is)
[Variable Store]       (preserved as-is, if present)
[FDArray INDEX]        (Private DICT offsets patched)
[Private DICTs]        (preserved as-is, relocated after FDArray)
[FDSelect]             (filtered to subset glyphs, if present)
CharStrings INDEX      (filtered to subset glyphs)

Private DICT offsets in Font DICTs always use the 5-byte integer encoding (operator 29 + int32). This makes the FDArray byte size independent of the offset values, breaking the layout chicken-and-egg.

See TODO #15 for the full design rationale.

Constant Summary collapse

OP_CHARSTRINGS =

CFF2 Top DICT operator codes (per Adobe TN 5177 §8.1).

17
OP_VSTORE = 
24
OP_FDARRAY = 
[12, 36].freeze
OP_FDSELECT = 
[12, 37].freeze
OP_PRIVATE = 
18
MAX_TOPDICT_ITERATIONS = 
16

Class Method Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(raw_data, mapping) ⇒ Cff2 



69
70
71
72
73
74
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 69

def initialize(raw_data, mapping)
  @raw = raw_data
  @mapping = mapping
  @reader = Tables::Cff2::TableReader.new(raw_data)
  @reader.read_top_dict
end

Class Method Details

.call(context:, tag:, table:) ⇒ String

Returns subset CFF2 table bytes.

Parameters:

  • context (SubsetContext)
  • tag (String) —

    "CFF2"

  • table (Object) —

    parsed CFF2 table (unused — we read raw bytes)

Returns:

  • (String) —

    subset CFF2 table bytes



50
51
52
53
54
55
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 50

def self.call(context:, tag:, table:)
  raw = raw_bytes_from_font(context.font) || raw_bytes_from_table(table)
  return raw unless raw && !raw.empty?

  new(raw, context.mapping).subset
end

Instance Method Details

#build_fdarray_bytes(entries, priv_offsets) ⇒ Object

Build the FDArray INDEX bytes with the given Private DICT offsets. Private offsets always use 5-byte int32 encoding for deterministic sizing across layout iterations.



163
164
165
166
167
168
169
170
171
172
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 163

def build_fdarray_bytes(entries, priv_offsets)
  rebuilt = entries.each_with_index.map do |entry, idx|
    parsed = entry[:parsed].dup
    if entry[:priv_size]
      parsed[OP_PRIVATE] = [entry[:priv_size], priv_offsets[idx] || 0]
    end
    encode_font_dict(parsed)
  end
  Tables::Cff2::IndexBuilder.build(rebuilt)
end

#build_filtered_sections(sections, retained_gids) ⇒ Object

Build all filtered/rebuilt section bytes.



116
117
118
119
120
121
122
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 116

def build_filtered_sections(sections, retained_gids)
  {
    charstrings_bytes: filter_charstrings(sections[:charstrings], retained_gids),
    fdselect_bytes: filter_fdselect(sections, retained_gids),
    fdarray_entries: sections[:fdarray] ? parse_fdarray_entries(sections[:fdarray]) : [],
  }
end

#build_top_dict_bytes(layout) ⇒ Object

Build the Top DICT bytes with updated section offsets.



246
247
248
249
250
251
252
253
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 246

def build_top_dict_bytes(layout)
  td = @reader.top_dict.dup
  td[OP_CHARSTRINGS] = layout[:charstrings_offset]
  td[OP_VSTORE] = layout[:vstore_offset] if layout[:vstore_offset]
  td[OP_FDARRAY] = layout[:fdarray_offset] if layout[:fdarray_offset]
  td[OP_FDSELECT] = layout[:fdselect_offset] if layout[:fdselect_offset]
  encode_dict(td)
end

#converge_layout(sections, filtered) ⇒ Object

Iterate Top DICT encoding to stability, then produce the final layout with FDArray patched with concrete Private DICT offsets.



180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 180

def converge_layout(sections, filtered)
  top_dict_size = @reader.header[:top_dict_length]
  placeholder_offsets = Array.new(filtered[:fdarray_entries].size, 0)
  fdarray_placeholder = if filtered[:fdarray_entries].any?
                          build_fdarray_bytes(filtered[:fdarray_entries],
                                              placeholder_offsets)
                        end
  priv_bytes = filtered[:fdarray_entries].map { |e| e[:priv_bytes] }.compact

  prev_size = nil
  layout = nil
  MAX_TOPDICT_ITERATIONS.times do
    layout = layout_offsets(sections, top_dict_size, filtered,
                            fdarray_placeholder, priv_bytes)
    new_td = build_top_dict_bytes(layout)
    top_dict_size = new_td.bytesize
    break if top_dict_size == prev_size

    prev_size = top_dict_size
  end

  finalize_layout(layout, sections, top_dict_size, filtered,
                  fdarray_placeholder, priv_bytes)
end

#encode_dict(dict) ⇒ Object

Encode a DICT hash to binary using the CFF2 DictEncoder.



354
355
356
357
358
359
360
361
362
363
364
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 354

def encode_dict(dict)
  io = +""
  dict.each do |operator, operands|
    vals = Array(operands)
    vals.each do |v|
      io << Tables::Cff2::DictEncoder.encode_integer(v.to_i)
    end
    io << Tables::Cff2::DictEncoder.encode_operator(operator)
  end
  io
end

#encode_font_dict(parsed) ⇒ Object

Encode a Font DICT. The Private operator (18) always uses 5-byte integer encoding for BOTH operands so the Font DICT byte size is deterministic across layout iterations.



369
370
371
372
373
374
375
376
377
378
379
380
381
382
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 369

def encode_font_dict(parsed)
  io = +""
  parsed.each do |operator, operands|
    vals = Array(operands)
    if operator == OP_PRIVATE && vals.size == 2
      io << encode_int32(vals[0].to_i)
      io << encode_int32(vals[1].to_i)
    else
      vals.each { |v| io << Tables::Cff2::DictEncoder.encode_integer(v.to_i) }
    end
    io << Tables::Cff2::DictEncoder.encode_operator(operator)
  end
  io
end

#encode_int32(value) ⇒ Object

5-byte integer encoding: byte 29 (marker) + 4-byte big-endian.



385
386
387
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 385

def encode_int32(value)
  [29, value].pack("CN")
end

#filter_charstrings(source_cs, retained_gids) ⇒ Object

Build a new CharStrings INDEX containing only the retained glyphs.



125
126
127
128
129
130
131
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 125

def filter_charstrings(source_cs, retained_gids)
  items = retained_gids.map do |gid|
    item = source_cs[gid]
    item.nil? || item.empty? ? "\x0e".b : item
  end
  Tables::Cff2::IndexBuilder.build(items)
end

#filter_fdselect(sections, retained_gids) ⇒ Object

Rebuild FDSelect to cover only retained glyphs in their new order.



134
135
136
137
138
139
140
141
142
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 134

def filter_fdselect(sections, retained_gids)
  return nil unless sections[:fdselect_offset]

  assignments = Tables::Cff2::FdSelect.read(
    @raw, sections[:fdselect_offset], sections[:charstrings].count
  )
  new_assigns = retained_gids.map { |gid| assignments[gid] || 0 }
  Tables::Cff2::FdSelect.build(new_assigns)
end

#finalize_layout(layout, sections, top_dict_size, filtered, fdarray_placeholder, priv_bytes) ⇒ Object

Produce the final layout: stabilized Top DICT bytes, header, and FDArray rebuilt with concrete Private DICT offsets.



257
258
259
260
261
262
263
264
265
266
267
268
269
270
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 257

def finalize_layout(layout, sections, top_dict_size, filtered,
                    fdarray_placeholder, priv_bytes)
  final = layout.dup
  final[:top_dict_bytes] = build_top_dict_bytes(layout)
  final[:header_bytes] = Tables::Cff2::Header.build(
    top_dict_size: top_dict_size,
  )
  if filtered[:fdarray_entries].any?
    final[:fdarray_bytes] = build_fdarray_bytes(
      filtered[:fdarray_entries], final[:private_dict_offsets] || []
    )
  end
  final
end

#gather_sections ⇒ Object

Source section gathering



92
93
94
95
96
97
98
99
100
101
102
103
104
105
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 92

def gather_sections
  td = @reader.top_dict
  hdr = @reader.header
  gsubr_off = hdr[:header_size] + hdr[:top_dict_length]

  {
    header_size: hdr[:header_size],
    charstrings: read_index(td[OP_CHARSTRINGS]),
    gsubr: read_index(gsubr_off),
    fdarray: td[OP_FDARRAY] ? read_index(td[OP_FDARRAY]) : nil,
    fdselect_offset: td[OP_FDSELECT],
    vstore_offset: td[OP_VSTORE],
  }
end

#index_bytes(index) ⇒ Object

Extract the raw bytes of a parsed INDEX from the source table.



294
295
296
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 294

def index_bytes(index)
  @raw.byteslice(index.start_offset, index.total_size)
end

#item_variation_data_end(abs_off) ⇒ Object 



330
331
332
333
334
335
336
337
338
339
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 330

def item_variation_data_end(abs_off)
  io = StringIO.new(@raw)
  io.seek(abs_off)
  item_count = read_u16(io)
  short_delta_count = read_u16(io)
  region_index_count = read_u16(io)
  long_count = region_index_count - short_delta_count
  delta_set_size = (short_delta_count * 2) + long_count
  abs_off + 6 + (region_index_count * 2) + (item_count * delta_set_size)
end

#layout_offsets(sections, top_dict_size, filtered, fdarray_bytes, priv_bytes) ⇒ Object

Assign byte offsets to every section.



206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 206

def layout_offsets(sections, top_dict_size, filtered, fdarray_bytes, priv_bytes)
  pos = sections[:header_size] + top_dict_size
  layout = { top_dict_size: top_dict_size }

  layout[:global_subrs_offset] = pos
  layout[:global_subrs_bytes] = index_bytes(sections[:gsubr])
  pos += layout[:global_subrs_bytes].bytesize

  if sections[:vstore_offset]
    layout[:vstore_offset] = pos
    layout[:vstore_bytes] = vstore_bytes(sections[:vstore_offset])
    pos += layout[:vstore_bytes].bytesize
  end

  if fdarray_bytes
    layout[:fdarray_offset] = pos
    layout[:fdarray_bytes] = fdarray_bytes
    pos += fdarray_bytes.bytesize

    priv_offsets = priv_bytes.map do |pb|
      off = pos
      pos += pb.bytesize
      off
    end
    layout[:private_dict_offsets] = priv_offsets
    layout[:private_dicts] = priv_bytes
  end

  if filtered[:fdselect_bytes]
    layout[:fdselect_offset] = pos
    layout[:fdselect_bytes] = filtered[:fdselect_bytes]
    pos += layout[:fdselect_bytes].bytesize
  end

  layout[:charstrings_offset] = pos
  layout[:charstrings_bytes] = filtered[:charstrings_bytes]
  layout
end

#parse_fdarray_entries(source_fdarray) ⇒ Object

Parse each Font DICT in the FDArray, extract Private DICT bytes, and prepare entries for relocation.



146
147
148
149
150
151
152
153
154
155
156
157
158
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 146

def parse_fdarray_entries(source_fdarray)
  source_fdarray.to_a.map do |fd_bytes|
    parsed = @reader.parse_dict(fd_bytes)
    priv = parsed[OP_PRIVATE]
    entry = { parsed: parsed, priv_bytes: nil, priv_size: nil }
    next entry unless priv.is_a?(Array) && priv.size == 2

    size, offset = priv
    entry[:priv_bytes] = @raw.byteslice(offset, size)
    entry[:priv_size] = size
    entry
  end
end

#read_index(offset) ⇒ Object 



107
108
109
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 107

def read_index(offset)
  Tables::Cff2::Index.read(@raw, offset)
end

#read_u16(io) ⇒ Object 



341
342
343
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 341

def read_u16(io)
  io.read(2).unpack1("n")
end

#read_u32(io) ⇒ Object 



345
346
347
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 345

def read_u32(io)
  io.read(4).unpack1("N")
end

#region_list_end(abs_off) ⇒ Object 



322
323
324
325
326
327
328
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 322

def region_list_end(abs_off)
  io = StringIO.new(@raw)
  io.seek(abs_off)
  axis_count = read_u16(io)
  region_count = read_u16(io)
  abs_off + 4 + (region_count * axis_count * 6)
end

#subset ⇒ String

Build the subset CFF2 table.

Returns:

  • (String) 


79
80
81
82
83
84
85
86
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 79

def subset
  sections = gather_sections
  retained_gids = @mapping.old_ids.sort

  filtered = build_filtered_sections(sections, retained_gids)
  layout = converge_layout(sections, filtered)
  write_table(layout, filtered)
end

#vstore_bytes(vstore_offset) ⇒ Object

Extract VStore bytes by computing its byte extent.



299
300
301
302
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 299

def vstore_bytes(vstore_offset)
  size = vstore_extent(vstore_offset)
  @raw.byteslice(vstore_offset, size)
end

#vstore_extent(vs_off) ⇒ Object

Compute the byte size of an ItemVariationStore.



305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 305

def vstore_extent(vs_off)
  io = StringIO.new(@raw)
  io.seek(vs_off)
  _format = read_u16(io)
  region_list_off = read_u32(io)
  data_count = read_u16(io)
  data_offsets = Array.new(data_count) { read_u32(io) }

  extents = []
  extents << region_list_end(vs_off + region_list_off) if region_list_off.positive?
  data_offsets.each do |rel_off|
    extents << item_variation_data_end(vs_off + rel_off) if rel_off.positive?
  end
  end_off = extents.max || (8 + (data_count * 4))
  end_off - vs_off
end

#write_table(layout, filtered) ⇒ Object

Table serialization



276
277
278
279
280
281
282
283
284
285
286
287
 
# File 'lib/fontisan/subset/table_strategy/cff2.rb', line 276

def write_table(layout, filtered)
  io = StringIO.new(+"")
  io << layout[:header_bytes]
  io << layout[:top_dict_bytes]
  io << layout[:global_subrs_bytes]
  io << layout[:vstore_bytes] if layout[:vstore_bytes]
  io << layout[:fdarray_bytes] if layout[:fdarray_bytes]
  (layout[:private_dicts] || []).each { |pd| io << pd }
  io << layout[:fdselect_bytes] if layout[:fdselect_bytes]
  io << layout[:charstrings_bytes]
  io.string
end