Class: Fontisan::Variation::Optimizer

Inherits:
Object
  • Object
show all
Defined in:
lib/fontisan/variation/optimizer.rb

Overview

Optimizes CFF subroutines for variable fonts

This class analyzes CharStrings in CFF2 variable fonts and optimizes blend operations by extracting common blend sequences into subroutines, deduplicating variation regions, and minimizing ItemVariationStore data.

Optimization strategies:

  1. Blend pattern extraction - Find repeating blend sequences

  2. Region deduplication - Merge identical variation regions

  3. ItemVariationStore optimization - Compact delta storage

  4. Subroutine reordering - Place frequent blends in low IDs

Examples:

Optimizing a variable font

optimizer = Fontisan::Variation::Optimizer.new(cff2_table)
optimized = optimizer.optimize
# => Optimized CFF2 table with reduced file size

See Also:

  • docs/SUBROUTINE_ARCHITECTUREdocs/SUBROUTINE_ARCHITECTURE.md
  • docs/CFF2_ARCHITECTUREdocs/CFF2_ARCHITECTURE.md

Instance Attribute Summary collapse

Instance Method Summary collapse

Constructor Details

#initialize(cff2, options = {}) ⇒ Optimizer

Initialize optimizer

Parameters:

  • cff2 (CFF2) —

    CFF2 table with blend operators

  • options (Hash) (defaults to: {}) —

    Optimization options

Options Hash (options):

  • :max_subrs (Integer) —

    Maximum subroutines (default: 65535)

  • :region_threshold (Float) —

    Region similarity threshold (default: 0.001)

  • :deduplicate_regions (Boolean) —

    Enable region deduplication (default: true)



41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
 
# File 'lib/fontisan/variation/optimizer.rb', line 41

def initialize(cff2, options = {})
  @cff2 = cff2
  @options = {
    max_subrs: 65535,
    region_threshold: 0.001,
    deduplicate_regions: true,
  }.merge(options)

  @stats = {
    original_size: 0,
    optimized_size: 0,
    blend_patterns_found: 0,
    subroutines_created: 0,
    regions_deduplicated: 0,
  }
end

Instance Attribute Details

#cff2 ⇒ CFF2 (readonly)

Returns CFF2 table being optimized.

Returns:

  • (CFF2) —

    CFF2 table being optimized



29
30
31
 
# File 'lib/fontisan/variation/optimizer.rb', line 29

def cff2
  @cff2
end

#stats ⇒ Hash (readonly)

Returns Optimization statistics.

Returns:

  • (Hash) —

    Optimization statistics



32
33
34
 
# File 'lib/fontisan/variation/optimizer.rb', line 32

def stats
  @stats
end

Instance Method Details

#analyze_blend_patterns ⇒ Array<BlendPattern>

Analyze blend patterns in CharStrings

Scans all CharStrings to find repeating blend operator sequences that can be extracted into subroutines.

Returns:

  • (Array<BlendPattern>) —

    Identified blend patterns



93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
 
# File 'lib/fontisan/variation/optimizer.rb', line 93

def analyze_blend_patterns
  patterns = []
  glyph_count = @cff2.glyph_count

  glyph_count.times do |glyph_id|
    charstring = @cff2.charstring(glyph_id)
    next unless charstring

    # Extract blend operator sequences
    blend_sequences = extract_blend_sequences(charstring)
    patterns.concat(blend_sequences)
  end

  # Group identical patterns
  grouped = group_patterns(patterns)
  @stats[:blend_patterns_found] = grouped.length

  grouped
end

#compact_variation_data(store) ⇒ Object

Compact variation data by removing unused entries

Parameters:

  • store (ItemVariationStore) —

    Variation store



250
251
252
253
254
255
256
257
258
 
# File 'lib/fontisan/variation/optimizer.rb', line 250

def compact_variation_data(store)
  # Identify used variation indices from CharStrings
  used_indices = collect_used_variation_indices

  # Remove unused data
  store.item_variation_data.each do |data|
    data.compact_unused(used_indices)
  end
end

#coords_similar?(c1, c2) ⇒ Boolean

Check if coordinates are similar within threshold

Parameters:

  • c1 (Float) —

    First coordinate

  • c2 (Float) —

    Second coordinate

Returns:

  • (Boolean) —

    True if similar



228
229
230
 
# File 'lib/fontisan/variation/optimizer.rb', line 228

def coords_similar?(c1, c2)
  (c1 - c2).abs <= @options[:region_threshold]
end

#deduplicate_regions ⇒ Object

Deduplicate variation regions

Merges regions that are functionally identical (within threshold).



158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
 
# File 'lib/fontisan/variation/optimizer.rb', line 158

def deduplicate_regions
  return unless @cff2.variation_store

  regions = @cff2.variation_store.region_list
  original_count = regions.length

  # Find duplicate regions
  unique_regions = []
  region_mapping = {}

  regions.each_with_index do |region, index|
    # Check if region matches any existing unique region
    match_index = find_matching_region(region, unique_regions)

    if match_index
      region_mapping[index] = match_index
    else
      region_mapping[index] = unique_regions.length
      unique_regions << region
    end
  end

  # Update references in ItemVariationStore
  update_region_references(region_mapping) if regions.length > unique_regions.length

  @cff2.variation_store.region_list = unique_regions
  @stats[:regions_deduplicated] = original_count - unique_regions.length
end

#extract_blend_sequences(charstring) ⇒ Array<BlendPattern>

Extract blend sequences from CharString

Parameters:

  • charstring (String) —

    Binary CharString data

Returns:

  • (Array<BlendPattern>) —

    Blend patterns found



117
118
119
120
121
122
123
124
125
126
127
128
129
130
 
# File 'lib/fontisan/variation/optimizer.rb', line 117

def extract_blend_sequences(charstring)
  patterns = []
  operators = parse_charstring_operators(charstring)

  operators.each_with_index do |op, index|
    next unless blend_operator?(op)

    # Extract blend and surrounding context
    pattern = extract_pattern_context(operators, index)
    patterns << pattern if pattern
  end

  patterns
end

#extract_blend_subroutines(patterns) ⇒ Array<Subroutine>

Extract common blend sequences into subroutines

Parameters:

  • patterns (Array<BlendPattern>) —

    Blend patterns

Returns:

  • (Array<Subroutine>) —

    Created subroutines



136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
 
# File 'lib/fontisan/variation/optimizer.rb', line 136

def extract_blend_subroutines(patterns)
  # Filter patterns by frequency and savings
  candidates = patterns.select do |pattern|
    pattern[:frequency] >= 2 && pattern[:savings].positive?
  end

  # Convert patterns to format expected by SubroutineOptimizer
  # The optimizer expects objects with methods, but we have hashes
  # For now, just select and order them directly
  selected = candidates.sort_by { |p| -p[:savings] }
    .take(@options[:max_subrs])

  # Order by frequency for efficient encoding
  ordered = selected.sort_by { |p| -p[:frequency] }

  @stats[:subroutines_created] = ordered.length
  ordered
end

#find_matching_region(region, unique_regions) ⇒ Integer?

Find matching region within threshold

Parameters:

  • region (RegionAxisCoordinates) —

    Region to match

  • unique_regions (Array<RegionAxisCoordinates>) —

    Existing unique regions

Returns:

  • (Integer, nil) —

    Index of matching region or nil



192
193
194
195
196
197
 
# File 'lib/fontisan/variation/optimizer.rb', line 192

def find_matching_region(region, unique_regions)
  unique_regions.each_with_index do |unique, index|
    return index if regions_match?(region, unique)
  end
  nil
end

#optimize ⇒ CFF2

Optimize CFF2 table

Performs all optimization passes and returns optimized table.

Returns:

  • (CFF2) —

    Optimized CFF2 table



63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
 
# File 'lib/fontisan/variation/optimizer.rb', line 63

def optimize
  @stats[:original_size] = estimate_table_size(@cff2)

  # Step 1: Analyze blend patterns across all CharStrings
  blend_patterns = analyze_blend_patterns

  # Step 2: Extract common blend sequences into subroutines
  subroutines = extract_blend_subroutines(blend_patterns)

  # Step 3: Deduplicate variation regions
  deduplicate_regions if @options[:deduplicate_regions]

  # Step 4: Optimize ItemVariationStore
  optimize_item_variation_store

  # Step 5: Rebuild CharStrings with subroutine calls
  rebuild_charstrings(subroutines)

  @stats[:optimized_size] = estimate_table_size(@cff2)
  @stats[:savings_percent] = calculate_savings_percent

  @cff2
end

#optimize_delta_encoding(store) ⇒ Object

Optimize delta encoding for efficiency

Parameters:

  • store (ItemVariationStore) —

    Variation store



263
264
265
 
# File 'lib/fontisan/variation/optimizer.rb', line 263

def optimize_delta_encoding(store)
  store.item_variation_data.each(&:optimize_encoding)
end

#optimize_item_variation_store ⇒ Object

Optimize ItemVariationStore

Compacts delta storage by removing unused data and optimizing encoding.



235
236
237
238
239
240
241
242
243
244
245
 
# File 'lib/fontisan/variation/optimizer.rb', line 235

def optimize_item_variation_store
  return unless @cff2.variation_store

  store = @cff2.variation_store

  # Remove unused variation data
  compact_variation_data(store)

  # Optimize delta encoding (use shortest representation)
  optimize_delta_encoding(store)
end

#rebuild_charstrings(subroutines) ⇒ Object

Rebuild CharStrings with subroutine calls

Parameters:

  • subroutines (Array<Subroutine>) —

    Subroutines to use



270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
 
# File 'lib/fontisan/variation/optimizer.rb', line 270

def rebuild_charstrings(subroutines)
  return if subroutines.empty?

  glyph_count = @cff2.glyph_count

  glyph_count.times do |glyph_id|
    charstring = @cff2.charstring(glyph_id)
    next unless charstring

    # Rewrite CharString to use subroutines
    optimized = rewrite_with_subroutines(charstring, subroutines)
    @cff2.set_charstring(glyph_id, optimized)
  end

  # Update subroutine index in CFF2
  @cff2.local_subr_index = subroutines
end

#regions_match?(r1, r2) ⇒ Boolean

Check if two regions match within threshold

Parameters:

  • r1 (RegionAxisCoordinates) —

    First region

  • r2 (RegionAxisCoordinates) —

    Second region

Returns:

  • (Boolean) —

    True if regions match



204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
 
# File 'lib/fontisan/variation/optimizer.rb', line 204

def regions_match?(r1, r2)
  return false unless r1.axis_count == r2.axis_count

  r1.axis_count.times do |i|
    coords1 = r1.region_axes[i]
    coords2 = r2.region_axes[i]

    # Compare start, peak, end coordinates
    return false unless coords_similar?(coords1.start_coord,
                                        coords2.start_coord)
    return false unless coords_similar?(coords1.peak_coord,
                                        coords2.peak_coord)
    return false unless coords_similar?(coords1.end_coord,
                                        coords2.end_coord)
  end

  true
end

#statistics ⇒ Hash

Get optimization statistics

Returns:

  • (Hash) —

    Statistics about optimization



291
292
293
 
# File 'lib/fontisan/variation/optimizer.rb', line 291

def statistics
  @stats
end