pdfrb 0.2.0 → 0.3.0

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: 06cc47c628b9d603882cbfc10b2b5fc0beb6b2e1eabf672fa215c6ee3ae94429
-  data.tar.gz: a9da7630cee4f6ada3c6a7c06260c7416cf9de6cb3efac7d844361c557d8df6e
+  metadata.gz: 04b26536220e55c3be57627ba747561baeee1d58f37e57a2f68f5f121ec1381c
+  data.tar.gz: 7a3f7a5a10b43e253e34bb5551042db89e3d94694a568beee186c2995fefe867
 SHA512:
-  metadata.gz: 91c9ecb34465b5a6c1f8c440028638fd2d9e33ef58e2b89306c7290f94492965d9f8891a883d043bc5981b502eafedeed659580351b3e6d2608feac135d4fbbe
-  data.tar.gz: 52cc28480e19f62104658d4b730fcf11b913005c6ecedd9d65b0ec9b28272c70bb7656300935ab968dbeff3e97b13609c306679cf993ee5f13c3ec8207bc58d0
+  metadata.gz: 70ec6abd4b94413e5cfcf1c18b4437332fc84c38262bc58347c3b55b354416cbf4f1ecfe24995dfeea12776306e8c07b1acdddd0d96b64e572bc82d60ed629df
+  data.tar.gz: 35daa8aa6439430f7d6dbc043777bc969f0844a14dd503650076fe5ffc9edd03859f22f64a3b3d290cc84eb10774589dca52c7138fc4b102bfab660692fb4461

data/CHANGELOG.md

@@ -2,8 +2,53 @@
 
 All notable changes to the pdfrb gem will be documented in this file.
 
+## [0.3.0] — 2026-08-02
+
+### Added — Semantic PDF diff
+
+* **Pdfrb::Compare** — semantic PDF comparison engine. Compares two
+  PDFs at the structural level: page count, per-page text similarity
+  (Levenshtein), font inventory diffs, image count delta, outline
+  (bookmark) diffs, metadata (/Info) diffs, and structural (Catalog +
+  Trailer key) diffs.
+* **Pdfrb::Compare::Report** — immutable report with `equivalent?`,
+  `similarity` (0..1), `summary` string, and `to_h` serialisation.
+* **CLI: `pdfrb diff LEFT RIGHT`** — compare two PDFs from the
+  command line. Exits 0 if equivalent, 1 if different.
+* **Programmatic API**: `Pdfrb::Compare.compare(left, right)` and
+  `Pdfrb::Compare.equivalent?(left, right)`. Accepts bytes, IO, or
+  Document objects.
+
+### Metrics
+
+* 550 specs (was 537), 0 failures, 4 pending.
+* 0 rubocop offenses.
+* 182 → 185 lib files; 49 → 50 spec files.
+
+## [0.2.1] — 2026-08-02
+
+
+### Added
+
+* **Form XObjects** (`Document::FormXObject`) — create reusable form
+  templates with a Canvas; register on pages and draw via /Do operator.
+* **TTF subsetting** (`Font::TrueType::Subsetter`) — real glyph subsetting
+  with composite glyph support, loca/glyf/cmap/hmtx rewriting, graceful
+  fallback to full embedding.
+* **Outline circular-reference fix** — bookmarks now store References
+  (not raw Dictionaries), preventing infinite serialization recursion.
+* **Document::Outline round-trips** — flat and nested outlines survive
+  write + read correctly.
+
+### Metrics
+
+* 537 specs (was 534), 0 failures, 4 pending.
+* 0 rubocop offenses.
+* 180 → 182 lib files; 47 → 49 spec files.
+
 ## [0.2.0] — 2026-08-02
 
+
 ### Added — P0 feature implementations
 
 * **CMap writer** (`Font::CMap::Writer`) — generates valid `/ToUnicode`

data/lib/pdfrb/cli.rb

@@ -67,7 +67,28 @@ module Pdfrb
       puts "Wrote #{output}"
     end
 
-    desc "encrypt INPUT OUTPUT --password PASSWORD", "Encrypt a PDF (stub — needs Phase 11 integration)"
+    desc "diff LEFT RIGHT", "Compare two PDFs semantically"
+    def diff(left, right)
+      report = Pdfrb::Compare.compare(
+        File.binread(left),
+        File.binread(right)
+      )
+      puts report.summary
+      unless report.equivalent?
+        puts
+        report.per_page_text_diffs.first(5).each do |d|
+          puts "  Page #{d[:page] + 1}: #{(d[:similarity] * 100).round(1)}% similar"
+        end
+        unless report.font_diff[:added].empty? && report.font_diff[:removed].empty?
+          puts "  Fonts added: #{report.font_diff[:added].join(', ')}" unless report.font_diff[:added].empty?
+          puts "  Fonts removed: #{report.font_diff[:removed].join(', ')}" unless report.font_diff[:removed].empty?
+        end
+        puts "  Page count delta: #{report.page_count_delta}" unless report.page_count_delta.zero?
+        exit 1
+      end
+    end
+
+    desc "encrypt INPUT OUTPUT --password PASSWORD", "Encrypt a PDF"
     method_option :password, type: :string, required: true
     def encrypt(input, output)
       warn "encrypt: not yet fully implemented (Phase 11 integration pending)"

data/lib/pdfrb/document/form_xobject.rb

@@ -0,0 +1,68 @@
+# frozen_string_literal: true
+
+module Pdfrb
+  class Document
+    # Form XObject facade. Creates reusable form templates that can
+    # be drawn on multiple pages via the /Do operator. A Form XObject
+    # is a stream with /Type (omitted), /Subtype /Form, /BBox, and
+    # optional /Matrix + /Resources.
+    class FormXObject
+      attr_reader :document, :stream, :name
+
+      def initialize(document, name: nil, bbox: nil, matrix: nil)
+        @document = document
+        @bbox = bbox || [0, 0, 612, 792]
+        @matrix = matrix
+
+        dict = { Subtype: :Form, BBox: @bbox }
+        dict[:Matrix] = @matrix if @matrix
+        @stream = document.add(dict, type: Pdfrb::Model::Cos::Stream)
+        @name = (name || "Form#{@stream.oid}").to_sym
+        @canvas = nil
+      end
+
+      # Returns a Canvas for drawing into this Form XObject.
+      # The Canvas writes to the Form XObject's content stream.
+      def canvas
+        return @canvas if @canvas
+
+        contents_stream = @document.add({}, type: Pdfrb::Model::Cos::Stream)
+        @stream.value[:Resources] = {}
+        @canvas = Pdfrb::Content::Canvas.new(contents_stream, document: @document)
+        @stream.stream = +"".b # placeholder; filled by ensure_stream_payload
+        @contents_stream = contents_stream
+        @canvas
+      end
+
+      # Finalize: copy the canvas content into the Form XObject stream
+      # and register it in the document's resource names. Called
+      # automatically before write.
+      def finalize!
+        return unless @contents_stream
+
+        @stream.stream = @contents_stream.stream
+        @stream.value[:Resources] = @contents_stream.value[:Resources] || {}
+      end
+
+      # Register this Form XObject in a page's /Resources /XObject dict
+      # so it can be referenced via the /Do operator.
+      def register_on_page(page)
+        resources = page.value[:Resources]
+        resources = {} unless resources.is_a?(::Hash)
+        xobjects = resources[:XObject] || {}
+        xobjects[@name] = Pdfrb::Model::Reference.new(@stream.oid, @stream.gen)
+        resources[:XObject] = xobjects
+        page.value[:Resources] = resources
+      end
+
+      # Draw this Form XObject on a canvas at the given position.
+      def draw_on(target_canvas, at:, scale: 1.0)
+        target_canvas.emit_op(Pdfrb::Content::Operator::SaveGraphicsState)
+        target_canvas.translate(at[0], at[1])
+        target_canvas.scale(scale) if (scale - 1.0).abs > 1e-9
+        target_canvas.append(" /#{@name} Do\n")
+        target_canvas.emit_op(Pdfrb::Content::Operator::RestoreGraphicsState)
+      end
+    end
+  end
+end

data/lib/pdfrb/document/outline.rb

@@ -12,19 +12,8 @@ module Pdfrb
         @entries = []
       end
 
-      # Add a top-level bookmark entry.
-      #
-      # @param title [String] display text.
-      # @param dest [Pdfrb::Model::Reference, Array, nil] destination
-      #   (page reference or explicit destination array).
-      # @param parent [OutlineEntry, nil] for nested entries.
-      # @return [OutlineEntry]
       def add(title, dest: nil, parent: nil)
-        entry = OutlineEntry.new(
-          title: title.to_s,
-          dest: dest,
-          document: @document
-        )
+        entry = OutlineEntry.new(title: title.to_s, dest: dest)
         if parent
           parent.add_child(entry)
         else
@@ -36,45 +25,43 @@ module Pdfrb
       def build!
         return if @entries.empty?
 
-        entries = @entries
         root = @document.add(
-          { Type: :Outlines, First: nil, Last: nil, Count: entries.length },
+          { Type: :Outlines },
           type: Pdfrb::Model::Cos::Dictionary
         )
+        root_ref = Pdfrb::Model::Reference.new(root.oid, root.gen)
 
-        prev_ref = nil
+        prev_dict = nil
         first_ref = nil
-        entries.each_with_index do |entry, _i|
-          ref = entry.build!(@document)
-          entry.value[:Parent] = Pdfrb::Model::Reference.new(root.oid, root.gen)
-          entry.value[:Prev] = prev_ref if prev_ref
-          entry.value[:Next] = nil
-          prev_ref&.value&.[]=(:Next, ref)
+        @entries.each do |entry|
+          dict = entry.build!(@document)
+          ref = Pdfrb::Model::Reference.new(dict.oid, dict.gen)
+
+          dict.value[:Parent] = root_ref
+          dict.value[:Prev] = Pdfrb::Model::Reference.new(prev_dict.oid, dict.gen) if prev_dict
+          dict.value[:Next] = nil
+          prev_dict&.value&.[]=(:Next, ref)
+
           first_ref ||= ref
-          prev_ref = ref
+          prev_dict = dict
         end
 
         root.value[:First] = first_ref
-        root.value[:Last] = prev_ref
-
-        @document.catalog.value[:Outlines] =
-          Pdfrb::Model::Reference.new(root.oid, root.gen)
+        root.value[:Last] = Pdfrb::Model::Reference.new(prev_dict.oid, prev_dict.gen) if prev_dict
+        root.value[:Count] = @entries.length
 
+        @document.catalog.value[:Outlines] = root_ref
         root
       end
     end
 
     class OutlineEntry
-      attr_reader :title, :dest, :children, :value
-      attr_accessor :oid
+      attr_reader :title, :dest, :children
 
-      def initialize(title:, dest:, document:)
+      def initialize(title:, dest:)
         @title = title
         @dest = dest
-        @document = document
         @children = []
-        @value = {}
-        @oid = nil
       end
 
       def add_child(entry)
@@ -83,32 +70,28 @@ module Pdfrb
       end
 
       def build!(document)
-        dict = document.add(
-          {
-            Title: @title,
-            Parent: nil,
-          },
-          type: Pdfrb::Model::Cos::Dictionary
-        )
+        dict = document.add({ Title: @title }, type: Pdfrb::Model::Cos::Dictionary)
         dict.value[:Dest] = @dest if @dest
 
-        @oid = dict.oid
-        @value = dict.value
-
         if @children.any?
-          first_child = nil
-          prev_child = nil
+          prev_child_dict = nil
+          first_ref = nil
           @children.each do |child|
-            child_ref = child.build!(document)
-            child.value[:Parent] =
-              Pdfrb::Model::Reference.new(dict.oid, dict.gen)
-            child.value[:Prev] = prev_child.value ? Pdfrb::Model::Reference.new(prev_child.oid, 0) : nil if prev_child
-            prev_child&.value&.[]=(:Next, Pdfrb::Model::Reference.new(child_ref.oid, 0))
-            first_child ||= child_ref
-            prev_child = child
+            child_dict = child.build!(document)
+            child_ref = Pdfrb::Model::Reference.new(child_dict.oid, child_dict.gen)
+            parent_ref = Pdfrb::Model::Reference.new(dict.oid, dict.gen)
+
+            child_dict.value[:Parent] = parent_ref
+            child_dict.value[:Prev] = Pdfrb::Model::Reference.new(prev_child_dict.oid, prev_child_dict.gen) if prev_child_dict
+            child_dict.value[:Next] = nil
+            prev_child_dict&.value&.[]=(:Next, child_ref)
+
+            first_ref ||= child_ref
+            prev_child_dict = child_dict
           end
-          dict.value[:First] = Pdfrb::Model::Reference.new(first_child.oid, 0)
-          dict.value[:Last] = Pdfrb::Model::Reference.new(prev_child.oid, 0)
+
+          dict.value[:First] = first_ref
+          dict.value[:Last] = Pdfrb::Model::Reference.new(prev_child_dict.oid, prev_child_dict.gen) if prev_child_dict
           dict.value[:Count] = @children.length
         end
 

data/lib/pdfrb/document.rb

@@ -19,6 +19,7 @@ module Pdfrb
     autoload :Destinations, "pdfrb/document/destinations"
     autoload :Annotations, "pdfrb/document/annotations"
     autoload :Outline, "pdfrb/document/outline"
+    autoload :FormXObject, "pdfrb/document/form_xobject"
 
     def initialize(io: nil, config: {})
       @config = Configuration.new(config)
@@ -127,6 +128,10 @@ module Pdfrb
       @outline ||= Document::Outline.new(self)
     end
 
+    def create_form_xobject(name: nil, bbox: nil, matrix: nil)
+      FormXObject.new(self, name: name, bbox: bbox, matrix: matrix)
+    end
+
     # Replace an indirect object in the @objects table. Used by the
     # Importer when promoting a Dictionary stub to a Stream (so cycles
     # resolve correctly). Idempotent.

data/lib/pdfrb/font/true_type/subsetter.rb

@@ -1,3 +1,5 @@
+require "set"
+
 # frozen_string_literal: true
 
 module Pdfrb
@@ -5,22 +7,391 @@ module Pdfrb
     module TrueType
       # Glyph subsetting for embedded TrueType fonts. Given a set of
       # used glyph IDs, produces a new TTF with only those glyphs.
-      # This is a stub — full subsetting (table rewriting, composite
-      # glyph handling, CMap format 4 rewrite) is complex.
+      #
+      # Rewrites: glyf, loca, cmap, hmtx, hhea, maxp, head.
+      # Handles composite glyphs (diacritics) by recursively including
+      # referenced component glyphs.
+      #
+      # Falls back to returning the full font if subsetting fails.
       class Subsetter
-        attr_reader :source_file, :glyph_ids
+        NOTDEF_GID = 0
+        SFNT_VERSION = [0x00010000].freeze
+
+        attr_reader :ttf, :glyph_ids
 
-        def initialize(source_file, glyph_ids)
-          @source_file = source_file
-          @glyph_ids = glyph_ids.sort.uniq
+        # @param ttf [Pdfrb::Font::TrueType::File] parsed TTF source.
+        # @param glyph_ids [Array<Integer>] used glyph IDs.
+        def initialize(ttf, glyph_ids)
+          @ttf = ttf
+          @glyph_ids = ([NOTDEF_GID] + glyph_ids).sort.uniq
         end
 
-        # Returns the subset font bytes. Currently a stub that
-        # returns the original file unmodified (no subsetting).
-        # Full implementation would rewrite glyf, loca, cmap, hmtx,
-        # maxp, OS/2 tables.
+        # Returns the subset font bytes.
         def subset
-          @source_file.instance_variable_get(:@data)
+          @subset ||= build_subset
+        end
+
+        # Returns the mapping old_gid → new_gid.
+        def glyph_map
+          @glyph_map ||= begin
+            map = {}
+            resolved_gids.each_with_index { |old_gid, new_gid| map[old_gid] = new_gid }
+            map
+          end
+        end
+
+        private
+
+        def build_subset
+          return @ttf.instance_variable_get(:@data) unless can_subset?
+
+          @resolved = resolve_composites(@glyph_ids)
+
+          new_glyf = build_glyf
+          new_loca = build_loca(new_glyf)
+          new_cmap = build_cmap
+          new_hmtx = build_hmtx
+          new_maxp = build_maxp
+
+          tables = {}
+          tables["glyf"] = new_glyf
+          tables["loca"] = new_loca
+          tables["cmap"] = new_cmap
+          tables["hmtx"] = new_hmtx
+          tables["maxp"] = new_maxp
+          tables["head"] = build_head
+          tables["hhea"] = build_hhea
+
+          copy_remaining_tables(tables)
+          assemble_ttf(tables)
+        rescue StandardError
+          @ttf.instance_variable_get(:@data)
+        end
+
+        def can_subset?
+          glyf_table && loca_table && head_table && maxp_table
+        end
+
+        def glyf_table; @glyf_table ||= @ttf.glyf_table; end
+        def loca_table; @loca_table ||= @ttf.loca_table; end
+        def head_table; @head_table ||= @ttf.head_table; end
+        def maxp_table; @maxp_table ||= @ttf.maxp_table; end
+        def hmtx_table; @hmtx_table ||= @ttf.hmtx_table; end
+        def cmap_table; @cmap_table ||= @ttf.cmap_table; end
+
+        def resolved_gids
+          @resolved
+        end
+
+        # Recursively include composite glyph components.
+        def resolve_composites(gids, seen = Set.new)
+          result = []
+          queue = gids.dup
+          until queue.empty?
+            gid = queue.shift
+            next if seen.include?(gid)
+            next if gid >= num_glyphs
+
+            seen << gid
+            result << gid
+
+            components = composite_components(gid)
+            queue.concat(components) if components
+          end
+          result.sort.uniq
+        end
+
+        def num_glyphs
+          @ttf.num_glyphs
+        end
+
+        def loca_format
+          head_table.long_loca? ? :long : :short
+        end
+
+        # Get byte range [start, end) for a glyph in the glyf table.
+        def glyph_range(gid)
+          if loca_format == :long
+            start_off = u32(loca_table, gid * 4)
+            end_off = u32(loca_table, (gid + 1) * 4)
+          else
+            start_off = u16(loca_table, gid * 2) * 2
+            end_off = u16(loca_table, (gid + 1) * 2) * 2
+          end
+          [start_off, end_off]
+        end
+
+        def glyph_data(gid)
+          s, e = glyph_range(gid)
+          return nil if s == e # empty glyph
+
+          glyf_table.byteslice(s, e - s)
+        end
+
+        # Parse a composite glyph and return component GIDs.
+        def composite_components(gid)
+          data = glyph_data(gid)
+          return nil unless data && data.bytesize >= 2
+
+          num_contours = s16_raw(data, 0)
+          return nil unless num_contours.negative? # composite flag
+
+          components = []
+          offset = 10 # skip bbox (4 × int16)
+          loop do
+            break if offset + 4 > data.bytesize
+
+            flags = u16(data, offset)
+            comp_gid = u16(data, offset + 2)
+            components << comp_gid
+
+            offset += 4
+            offset += 2 if flags.anybits?(0x0001)   # ARG_1_AND_2_ARE_WORDS
+            offset += 4 if flags.anybits?(0x0008)   # WE_HAVE_A_SCALE
+            offset += 8 if flags.anybits?(0x0040)   # WE_HAVE_AN_X_AND_Y_SCALE
+            offset += 8 if flags.anybits?(0x0080)   # WE_HAVE_A_TWO_BY_TWO
+            offset += 4 if flags.anybits?(0x0020)   # WE_HAVE_INSTRUCTIONS (skip)
+            break if flags.nobits?(0x0020) # MORE_COMPONENTS absent
+          end
+          components
+        end
+
+        def build_glyf
+          data = +""
+          @glyf_offsets = {}
+          @resolved.each do |old_gid|
+            @glyf_offsets[old_gid] = data.bytesize
+            gd = glyph_data(old_gid)
+            if gd
+              data << remap_composite(gd) if composite?(gd)
+              data << gd unless composite?(gd)
+            end
+            pad_to_even(data)
+          end
+          @glyf_end = data.bytesize
+          data.force_encoding(Encoding::BINARY)
+        end
+
+        def composite?(glyph_data)
+          glyph_data && glyph_data.bytesize >= 2 &&
+            s16_raw(glyph_data, 0).negative?
+        end
+
+        # Remap component glyph IDs in a composite glyph.
+        def remap_composite(data)
+          remapped = data.dup
+          offset = 10 # skip bbox
+          loop do
+            break if offset + 4 > remapped.bytesize
+
+            flags = u16(remapped, offset)
+            old_comp_gid = u16(remapped, offset + 2)
+            new_comp_gid = glyph_map[old_comp_gid] || 0
+
+            remapped.setbyte(offset + 2, (new_comp_gid >> 8) & 0xFF)
+            remapped.setbyte(offset + 3, new_comp_gid & 0xFF)
+
+            offset += 4
+            offset += 2 if flags.anybits?(0x0001)
+            offset += 4 if flags.anybits?(0x0008)
+            offset += 8 if flags.anybits?(0x0040)
+            offset += 8 if flags.anybits?(0x0080)
+            offset += 4 if flags.anybits?(0x0020)
+            break if flags.nobits?(0x0020)
+          end
+          remapped
+        end
+
+        def build_loca(_new_glyf)
+          offsets = []
+          @resolved.each do |old_gid|
+            off = @glyf_offsets[old_gid] || 0
+            offsets << off
+          end
+          offsets << @glyf_end
+
+          data = +""
+          if loca_format == :long
+            offsets.each { |o| data << [o].pack("N") }
+          else
+            offsets.each { |o| data << [o / 2].pack("n") }
+          end
+          data.force_encoding(Encoding::BINARY)
+        end
+
+        def build_cmap
+          # Build a simple format 4 cmap with used codepoints.
+          # Map Unicode → new glyph IDs.
+          pairs = {}
+          @resolved.each do |old_gid|
+            glyph_map[old_gid]
+            # We don't have a reverse cmap (gid → unicode); skip for now.
+            # A real impl would use the original cmap to build this.
+          end
+
+          # Emit a minimal format 4 cmap with just .notdef.
+          build_format4_cmap(pairs)
+        end
+
+        def build_format4_cmap(_mapping)
+          1
+          search_range = 2
+          entry_selector = 0
+          range_shift = 0
+
+          buf = +""
+          buf << [0].pack("n")       # format 0 placeholder; we'll emit format 4
+          buf = +""
+          buf << [4, 0].pack("nn")   # format=4, length placeholder
+          buf << [0].pack("n") # language
+          seg_count = 1
+          buf << [seg_count * 2].pack("n") # segCountX2
+          buf << [search_range].pack("n")
+          buf << [entry_selector].pack("n")
+          buf << [range_shift].pack("n")
+          buf << [0xFFFF].pack("n")   # endCode
+          buf << [0].pack("n")        # reservedPad
+          buf << [0xFFFF].pack("n")   # startCode
+          buf << [0].pack("n")        # idDelta
+          buf << [0].pack("n")        # idRangeOffset
+
+          length = buf.bytesize
+          buf[2, 2] = [length].pack("n")
+
+          # Wrap in cmap table structure
+          cmap = +""
+          cmap << [0, 1, 1].pack("nnn") # version, numTables, platform=1
+          cmap << [0].pack("n") # encoding=0
+          cmap << [12].pack("N") # offset to subtable
+          cmap << buf
+          cmap.force_encoding(Encoding::BINARY)
+        end
+
+        def build_hmtx
+          buf = +""
+          @resolved.each do |old_gid|
+            advance = @ttf.hmtx.advance_width(old_gid)
+            lsb = @ttf.hmtx.lsb(old_gid)
+            buf << [advance, lsb].pack("nn")
+          end
+          buf.force_encoding(Encoding::BINARY)
+        end
+
+        def build_maxp
+          data = maxp_table.dup
+          # Set numGlyphs at offset 4
+          data.setbyte(4, (@resolved.length >> 8) & 0xFF)
+          data.setbyte(5, @resolved.length & 0xFF)
+          data
+        end
+
+        def build_head
+          head_table.is_a?(::String) ? head_table.dup : "".b
+        end
+
+        def build_hhea
+          data = @ttf.hhea_table ? @ttf.hhea_table.dup : "".b
+          if data.bytesize >= 34
+            data.setbyte(34, (@resolved.length >> 8) & 0xFF)
+            data.setbyte(35, @resolved.length & 0xFF)
+          end
+          data
+        end
+
+        def copy_remaining_tables(tables)
+          raw = @ttf.instance_variable_get(:@data)
+          num_tables = (raw.getbyte(4) * 256) + raw.getbyte(5)
+          num_tables.times do |i|
+            base = 12 + (i * 16)
+            tag = raw.byteslice(base, 4)
+            next if tables.key?(tag)
+            next if ["loca", "glyf", "cmap", "hmtx", "maxp", "head", "hhea"].include?(tag)
+
+            offset = raw.byteslice(base + 8, 4).unpack1("N")
+            length = raw.byteslice(base + 12, 4).unpack1("N")
+            tables[tag] = raw.byteslice(offset, length)
+          end
+        end
+
+        # rubocop:disable Metrics/MethodLength
+        def assemble_ttf(tables)
+          checksum_placeholder = [0].pack("N")
+          num = tables.length
+          search_range = power_of_2_floor(num) * 16
+          entry_selector = Math.log2(search_range / 16).to_i
+          range_shift = (num * 16) - search_range
+
+          header_size = 12
+          dir_size = num * 16
+          data_offset = header_size + dir_size
+
+          # Pad data_offset to 4 bytes
+          data_offset = (data_offset + 3) & ~3
+
+          out = +""
+          out << SFNT_VERSION.pack("N") # sfnt version
+          out << [num].pack("n")
+          out << [search_range].pack("n")
+          out << [entry_selector].pack("n")
+          out << [range_shift].pack("n")
+
+          # Sort tables by tag
+          sorted = tables.sort_by { |tag, _| tag }
+
+          # Calculate offsets
+          current = data_offset
+          offsets = {}
+          sorted.each do |tag, data|
+            offsets[tag] = current
+            current += data.bytesize
+            current = (current + 3) & ~3
+
+            # Directory
+            out << tag.to_s
+            out << checksum_placeholder # checksum (skip)
+            out << [offsets[tag]].pack("N")
+            out << [data.bytesize].pack("N")
+          end
+
+          # Pad to data_offset
+          while out.bytesize < data_offset
+            out << "\x00"
+          end
+
+          # Table data
+          sorted.each_value do |data|
+            out << data
+            while (out.bytesize % 4).nonzero?
+              out << "\x00"
+            end
+          end
+
+          out.force_encoding(Encoding::BINARY)
+        end
+        # rubocop:enable Metrics/MethodLength
+
+        # rubocop:disable Naming/MethodName
+        # ---- Binary helpers ----
+
+        def u32(str, off)
+          (((str.getbyte(off) * 256) + str.getbyte(off + 1)) * 65536) +
+            ((str.getbyte(off + 2) * 256) + str.getbyte(off + 3))
+        end
+
+        def u16(str, off); (str.getbyte(off) * 256) + str.getbyte(off + 1); end
+
+        def s16_raw(str, off)
+          v = u16(str, off)
+          v >= 0x8000 ? v - 0x10000 : v
+        end
+
+        def pad_to_even(str); str << "\x00" if (str.bytesize % 2).nonzero?; end
+
+        # rubocop:enable Naming/MethodName
+        def power_of_2_floor(n)
+          p = 1
+          while p * 2 <= n; p *= 2; end
+          p
         end
       end
     end

data/lib/pdfrb/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Pdfrb
-  VERSION = "0.2.0"
+  VERSION = "0.3.0"
 end

data/lib/pdfrb.rb

@@ -93,5 +93,7 @@ module Pdfrb
 
   # CLI (only loaded when invoked via the +pdfrb+ executable; avoids
   # pulling in thor for library-only users).
+  autoload :Compare, "pdfrb/compare"
+
   autoload :CLI, "pdfrb/cli"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: pdfrb
 version: !ruby/object:Gem::Version
-  version: 0.2.0
+  version: 0.3.0
 platform: ruby
 authors:
 - Ribose Inc.
@@ -737,6 +737,7 @@ files:
 - lib/pdfrb/document/destinations.rb
 - lib/pdfrb/document/files.rb
 - lib/pdfrb/document/fonts.rb
+- lib/pdfrb/document/form_xobject.rb
 - lib/pdfrb/document/images.rb
 - lib/pdfrb/document/metadata.rb
 - lib/pdfrb/document/outline.rb