idml 0.10.6 → 0.10.7

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: f60b5f9020e44d5131e9f50e2ddfb673f9caca65bbdee8428b4670baaf96108f
-  data.tar.gz: 1f19b80511a14cea5fab0b76404be809a8f28e680deb91ee1b4caf9353a1b756
+  metadata.gz: 943028d5bf17ca306e3cce0d23c4b042ddaa9473295f2fa7f4af43009a7423a0
+  data.tar.gz: b9c3c1e5421ffa4de44396dbc94124434dda470c3f10fabd5ddd81146c3991c1
 SHA512:
-  metadata.gz: 9bea97bee388cae4ab403c3bea6cae233c3a378c569c0c678620bdb59a3ed5534ab9aa560866fad103d8a724351e396c72e9a533671fd44189719958f83d811e
-  data.tar.gz: 151917cdee9265410c788c85fb6f28d45d07ffc37cc3c4e3fc14b4578f2be57adee09393c284a53a883f1f1dbe105920ecd745a264f0311cc38c0c8f92d94a05
+  metadata.gz: 86717ea81b1ede4576d3d41e7026cd2731ec40024ea887d33a03c4ff7771384cb0ae9fd04f5948a322d48986dae2946adcebf167a7c51ae3ce7890a28726e7ee
+  data.tar.gz: 12a2089f4600a04fe5cd1d74e409c0933dc6f691156ffca4eeb87fcb25fd5d89e2a1ec58261c8a087ea00b93cf42abee0bb54d64f72032dc3b6bbe919f6cc6b4

data/TODO.pdf/139-hyphen-break-opportunities.md

@@ -0,0 +1,24 @@
+# TODO PDF 139: Hyphen break opportunities in the LineBreaker
+
+## Status: COMPLETE — implemented 2026-08-27
+
+## Problem
+
+The greedy line breaker only treated spaces as break opportunities.
+Long compound words ("state-of-the-art", "e-mail") overflowed the
+frame as one unbreakable run instead of wrapping after a hyphen.
+
+## Solution
+
+`LineBreaker#break_after?` treats hyphen-minus (U+002D), Unicode
+hyphen (U+2010), non-breaking hyphen (U+2011), and soft hyphen
+(U+00AD) as break-after points: the hyphen stays on the first line
+and the remainder wraps. Soft hyphens break without adding a
+visible hyphen. Dictionary-based hyphenation (Knuth-Liang) remains
+out of scope — the Hyphenation* PSR attributes stay parsed but
+unused.
+
+## Files
+
+- `lib/idml/text_engine/line_breaker.rb`
+- `spec/idml/text_engine/text_engine_spec.rb`

data/TODO.pdf/140-cap-xheight-first-baselines.md

@@ -0,0 +1,29 @@
+# TODO PDF 140: CapHeight / XHeight / EmboxHeight first baselines
+
+## Status: COMPLETE — implemented 2026-08-27
+
+## Problem
+
+FirstBaselineOffset honored AscentOffset and FixedHeight (TODO 128)
+but approximated CapHeight, XHeight, and EmboxHeight as leading —
+text sat noticeably lower than InDesign places it.
+
+## Solution
+
+`TextFrameRenderer#baseline_target` now handles all five modes:
+
+- **CapHeight** — the font's cap-height metric when the provider
+  fills it; pdfrb 0.7.1 never does, so the fallback is the standard
+  0.72 em proportion (Arial and Helvetica both sit within 1%).
+- **XHeight** — 0.52 em (no x-height metric exists in pdfrb).
+- **EmboxHeight** — the em box itself: the first paragraph's point
+  size.
+
+`PdfrbFontMetrics#cap_height` exposes the metric for when pdfrb
+starts populating it.
+
+## Files
+
+- `lib/idml/render/renderers/text_frame_renderer.rb`
+- `lib/idml/text_engine/pdfrb_font_metrics.rb`
+- `spec/idml/render/first_baseline_spec.rb`

data/TODO.pdf/141-perf-hundred-pages.md

@@ -0,0 +1,22 @@
+# TODO PDF 141: 100+ page performance regression guard
+
+## Status: COMPLETE — implemented 2026-08-27
+
+## Problem
+
+No test guarded against per-page complexity blowups — a regression
+to O(document) work per page would only surface as slow user
+renders.
+
+## Solution
+
+`perf_hundred_pages_spec.rb` builds a synthetic 120-spread package
+(one page + one text frame each, all sharing a wordy story) and
+renders it end-to-end through `Idml::Render.render`. Asserts all
+120 PDF pages emit and the render completes under a generous 60s
+ceiling — sized to trip only on pathological blowups, not machine
+noise. Baseline on this branch: ~2s for 120 pages.
+
+## Files
+
+- `spec/idml/render/perf_hundred_pages_spec.rb`

data/TODO.pdf/61-known-limitations.md

@@ -1,6 +1,6 @@
 # TODO PDF 61: Known limitations and future work
 
-## Status: DOCUMENTED — refreshed 2026-08-26
+## Status: DOCUMENTED — refreshed 2026-08-27
 
 Previously listed limitations (font embedding, tagged PDF, per-run
 styling, dead code) were resolved by TODOs 52/53/55/67/76 long ago;
@@ -9,8 +9,9 @@ this list reflects the current state.
 ## Known limitations
 
 ### Text layout
-- **Hyphenation** is not implemented (Hyphenation* attributes are
-  parsed but unused); a hyphenation dictionary would be required.
+- **Hyphenation**: dictionary-based hyphenation is not implemented
+  (Hyphenation* attributes are parsed but unused); compound words
+  wrap after explicit hyphens (TODO 139).
 - **Justification**: word/letter spacing caps apply (TODO 119),
   MaximumGlyphScaling stretches (TODO 129) and MinimumGlyphScaling
   compresses overlong lines (TODO 135); ToBinding / RTL binding
@@ -18,8 +19,9 @@ this list reflects the current state.
 - **Keep options**: KeepAllLinesTogether (TODO 123), KeepWithNext
   (TODO 127), and the KeepFirstLines / KeepLastLines windows
   (TODO 133, line-count approximation) are honored.
-- **First baseline**: AscentOffset and FixedHeight are honored
-  (TODO 128); CapHeight / XHeight approximate as leading.
+- **First baseline**: all five modes honor their metrics (TODOs
+  128/140); CapHeight / XHeight use standard font proportions
+  (0.72 / 0.52 em) until a metrics provider fills them.
 - **StartParagraph** breaks act at frame/column granularity — no
   odd/even page parity.
 - **Text wrap**: BoundingBoxTextWrap, Contour, and Inverse modes
@@ -54,4 +56,6 @@ this list reflects the current state.
 
 - Hyphenation dictionary integration
 - Shape-mode text wrap contours (side-aware)
-- Performance benchmark for 100+ page documents
+- Named mojikumi sets (詳細 etc.)
+- ToBinding / RTL binding alignment
+- Vertical-mode keep options and run-grouped Latin rotation

data/lib/idml/render/renderers/text_frame_renderer.rb

@@ -629,12 +629,19 @@ module Idml
         # the first line's baseline sits. The layout's default is
         # leading-based; AscentOffset shifts by (ascent − leading)
         # and FixedHeight by (MinimumFirstBaselineOffset − leading).
-        # CapHeight / XHeight / EmboxHeight approximate as leading.
-        # Returns the extra shift to subtract from the cursor.
+        # CapHeight uses the font's cap-height metric (leading
+        # fallback when the font reports none); XHeight approximates
+        # as 70% of cap height; EmboxHeight uses the em box (the
+        # first paragraph's point size). Returns the extra shift to
+        # subtract from the cursor.
+        FIRST_BASELINE_MODES = %w[
+          AscentOffset FixedHeight CapHeight XHeight EmboxHeight
+        ].freeze
+
         def self.first_baseline_offset(context, state, font)
           pref = context.item&.text_frame_preference&.first
           mode = pref&.first_baseline_offset
-          return 0.0 unless %w[AscentOffset FixedHeight].include?(mode)
+          return 0.0 unless FIRST_BASELINE_MODES.include?(mode)
 
           baseline_target(pref, mode, state, font) -
             first_paragraph_leading(state)
@@ -643,21 +650,52 @@ module Idml
 
         # Distance from the top inset to the first baseline under
         # the given mode.
+        # pdfrb 0.7.1 never fills the cap_height metric, so
+        # CapHeight / XHeight fall back to standard font
+        # proportions (cap ≈ 0.72 em, x ≈ 0.52 em — Arial and
+        # Helvetica both sit within 1% of these).
+        CAP_HEIGHT_RATIO = 0.72
+        X_HEIGHT_RATIO = 0.52
+
         def self.baseline_target(pref, mode, state, font)
-          if mode == "AscentOffset"
-            ascent_scaled(font) * first_paragraph_leading(state)
+          leading = first_paragraph_leading(state)
+          case mode
+          when "AscentOffset"
+            ratio_scaled(font.ascent, font) * leading
+          when "CapHeight"
+            cap_ratio(font) * leading
+          when "XHeight"
+            X_HEIGHT_RATIO * leading
+          when "EmboxHeight"
+            first_paragraph_size(state)
           else
-            pref.minimum_first_baseline_offset ||
-              first_paragraph_leading(state)
+            pref.minimum_first_baseline_offset || leading
           end
         end
         private_class_method :baseline_target
 
-        def self.ascent_scaled(font)
-          upem = font.units_per_em.zero? ? 1 : font.units_per_em
-          font.ascent / upem
+        def self.cap_ratio(font)
+          cap = font.cap_height
+          return CAP_HEIGHT_RATIO unless cap
+
+          ratio_scaled(cap, font)
+        end
+        private_class_method :cap_ratio
+
+        def self.ratio_scaled(metric, font)
+          upem = font.units_per_em
+          return 1.0 if upem.zero?
+
+          metric / upem
+        end
+        private_class_method :ratio_scaled
+
+        def self.first_paragraph_size(state)
+          paragraph = state.current_paragraph || state.paragraphs.first
+          first_run = paragraph&.runs&.first
+          first_run&.point_size || DEFAULT_SIZE
         end
-        private_class_method :ascent_scaled
+        private_class_method :first_paragraph_size
 
         def self.first_paragraph_leading(state)
           paragraph = state.current_paragraph || state.paragraphs.first

data/lib/idml/text_engine/line_breaker.rb

@@ -6,10 +6,16 @@ module Idml
     Line = Struct.new(:glyphs, :width, :x_offset)
 
     # Greedy word-wrap line breaker. Accumulates glyphs until
-    # exceeding the frame width, then breaks at the last space.
-    # CJK runs get a kinsoku shori post-pass (no line starts or
-    # ends with a forbidden character).
+    # exceeding the frame width, then breaks at the last break
+    # opportunity: a space, or a hyphen (compound words wrap after
+    # the hyphen, which stays on the first line). CJK runs get a
+    # kinsoku shori post-pass (no line starts or ends with a
+    # forbidden character).
     class LineBreaker
+      # Break-after opportunities besides spaces: hyphen-minus,
+      # the Unicode hyphens, and the soft hyphen (which breaks
+      # without adding a visible hyphen).
+      HYPHEN_CODEPOINTS = [0x2D, 0x2010, 0x2011, 0x00AD].freeze
       def self.break(glyphs:, frame_width:)
         lines = new(frame_width).break(glyphs)
         return lines unless cjk_run?(glyphs)
@@ -44,26 +50,26 @@ module Idml
         lines = []
         current = []
         current_width = 0
-        last_space_idx = -1
-        width_at_space = 0
+        break_idx = -1
+        width_at_break = 0
 
         glyphs.each_with_index do |glyph, _idx|
           current << glyph
           current_width += glyph.width
 
-          if glyph.is_space
-            last_space_idx = current.length - 1
-            width_at_space = current_width
+          if break_after?(glyph)
+            break_idx = current.length - 1
+            width_at_break = current_width
           end
 
           next unless current_width > @frame_width && current.length > 1
 
-          if last_space_idx >= 0
-            line_glyphs = current[0..last_space_idx]
-            lines << Line.new(line_glyphs, width_at_space, 0)
-            current = current[(last_space_idx + 1)..]
+          if break_idx >= 0
+            line_glyphs = current[0..break_idx]
+            lines << Line.new(line_glyphs, width_at_break, 0)
+            current = current[(break_idx + 1)..]
             current_width = current.sum(&:width)
-            last_space_idx = -1
+            break_idx = -1
           elsif cjk_break?(current)
             lines << Line.new(current[0..-2], current_width - glyph.width, 0)
             current = [glyph]
@@ -78,6 +84,11 @@ module Idml
         lines << Line.new(current, current_width, 0) if current.any?
         lines
       end
+
+      def break_after?(glyph)
+        glyph.is_space || HYPHEN_CODEPOINTS.include?(glyph.codepoint)
+      end
+      private :break_after?
     end
   end
 end

data/lib/idml/text_engine/pdfrb_font_metrics.rb

@@ -37,6 +37,12 @@ module Idml
         0
       end
 
+      # Cap height in font units; nil when the font tables don't
+      # report one (some pdfrb-loaded fonts).
+      def cap_height
+        metrics_data[:cap_height]
+      end
+
       # Returns raw advance width in font units (not scaled by size).
       def glyph_width(codepoint)
         cp = codepoint.is_a?(String) ? codepoint.each_codepoint.first : codepoint

data/lib/idml/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module Idml
-  VERSION = "0.10.6"
+  VERSION = "0.10.7"
 end

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: idml
 version: !ruby/object:Gem::Version
-  version: 0.10.6
+  version: 0.10.7
 platform: ruby
 authors:
 - Ribose
 autorequire:
 bindir: exe
 cert_chain: []
-date: 2026-08-26 00:00:00.000000000 Z
+date: 2026-08-27 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: bigdecimal
@@ -201,7 +201,10 @@ files:
 - TODO.pdf/136-diagonal-cell-strokes.md
 - TODO.pdf/137-simple-render-footnote-text.md
 - TODO.pdf/138-text-wrap-sides.md
+- TODO.pdf/139-hyphen-break-opportunities.md
 - TODO.pdf/14-page-item-element-models.md
+- TODO.pdf/140-cap-xheight-first-baselines.md
+- TODO.pdf/141-perf-hundred-pages.md
 - TODO.pdf/15-wire-spread-children.md
 - TODO.pdf/16-typed-model-pipeline.md
 - TODO.pdf/17-shape-rendering.md