asciidoctor-epub3 1.5.0.alpha.6 → 1.5.0.alpha.7

data/asciidoctor-epub3.gemspec

@@ -0,0 +1,42 @@
+# -*- encoding: utf-8 -*-
+require File.expand_path('lib/asciidoctor-epub3/version', File.dirname(__FILE__))
+require 'open3' unless defined? Open3
+
+Gem::Specification.new do |s|
+  s.name = 'asciidoctor-epub3'
+  s.version = Asciidoctor::Epub3::VERSION
+
+  s.summary = 'Converts AsciiDoc documents to EPUB3 and KF8/MOBI (Kindle) e-book formats'
+  s.description = <<-EOS
+An extension for Asciidoctor that converts AsciiDoc documents to EPUB3 and KF8/MOBI (Kindle) e-book archives.
+  EOS
+
+  s.authors = ['Dan Allen', 'Sarah White']
+  s.email = 'dan@opendevise.com'
+  s.homepage = 'https://github.com/asciidoctor/asciidoctor-epub3'
+  s.license = 'MIT'
+
+  s.required_ruby_version = '>= 1.9.3'
+
+  files = begin
+    (result = Open3.popen3('git ls-files -z') {|_, out| out.read }.split %(\0)).empty? ? Dir['**/*'] : result
+  rescue
+    Dir['**/*']
+  end
+  s.files = files.grep %r/^(?:(?:data\/(?:fonts|images|styles)|lib)\/.+|Gemfile|Rakefile|(?:CHANGELOG|LICENSE|NOTICE|README)\.adoc|#{s.name}\.gemspec)$/
+  s.executables = %w(asciidoctor-epub3 adb-push-ebook)
+  s.test_files = s.files.grep(/^(?:test|spec|feature)\/.*$/)
+
+  s.require_paths = ['lib']
+
+  s.has_rdoc = true
+  s.rdoc_options = ['--charset=UTF-8', '--title="Asciidoctor EPUB3"', '--main=README.adoc', '-ri']
+  s.extra_rdoc_files = ['CHANGELOG.adoc', 'LICENSE.adoc', 'NOTICE.adoc', 'README.adoc']
+
+  s.add_development_dependency 'rake'
+  #s.add_development_dependency 'rdoc', '~> 4.1.0'
+
+  s.add_runtime_dependency 'asciidoctor', '~> 1.5.0'
+  s.add_runtime_dependency 'gepub', '~> 0.6.9.2'
+  s.add_runtime_dependency 'thread_safe', '~> 0.3.6'
+end

data/bin/adb-push-ebook

@@ -1,6 +1,10 @@
 #!/usr/bin/env ruby
 
 ADB = ENV['ADB'] || 'adb'
+TARGETS = {
+  '.epub' => '/sdcard/',
+  '.mobi' => '/sdcard/Android/data/com.amazon.kindle/files/'
+}
 
 unless ::File.executable? ADB
   warn %(adb-push-ebook: `adb` not found.\nPlease set the ADB environment variable or add `adb` to your PATH.)
@@ -10,16 +14,21 @@ end
 require 'open3'
 require 'shellwords'
 
-docname = ARGV[0] || '_output/sample-book'
+payload_file = ARGV[0] || '_output/sample-book'
 
-targets = {
-  'epub' => '/sdcard/',
-  'mobi' => '/sdcard/Android/data/com.amazon.kindle/files/'
-}
+if (payload_file_ext = File.extname payload_file).empty?
+  transfers = TARGETS.map do |(ext, target_dir)|
+    {
+      src: %(#{payload_file}#{ext}),
+      dest: target_dir
+    }
+  end
+else
+  transfers = [{ src: payload_file, dest: TARGETS[payload_file_ext] }]
+end
 
-targets.each {|(ext, target)|
-  file = %(#{docname}.#{ext})
-  Open3.popen2e(Shellwords.join [ADB, 'push', file, target]) {|input, output, wait_thr|
+transfers.each do |transfer|
+  Open3.popen2e(Shellwords.join [ADB, 'push', transfer[:src], transfer[:dest]]) do |input, output, wait_thr|
     output.each {|line| puts line }
-  } if File.file? file
-}
+  end if File.file? transfer[:src]
+end

data/data/styles/epub3-css3-only.css

@@ -1,17 +1,30 @@
-/* Gitden & Namo default to 16px font-size; bump it to 20px (125%) */
+/* @page is for EPUB2 only */
+@page {
+  margin: 0;
+}
+
+body.calibre-desktop {
+  padding: 20pt 0 !important;
+}
+
+body.calibre-desktop > section {
+  margin: 0 25pt;
+}
+
+/* Gitden & Namo default to 16px font-size; bump it to 19px (118.75%) */
 body.gitden-reader,
 body.namo-epub-library {
-  font-size: 125%;
+  font-size: 118.75%;
 }
 
-/* Gitden doesn't give us much margin, so let's match Kindle */
+/* Gitden doesn't give us much margin, so let's roughly match Aldiko and Kindle (narrow setting) */
 body.gitden-reader {
-  margin: 0 25pt;
+  margin: 0 5pt !important;
 }
 
 /* Namo has the same margin problem, except setting side margins doesn't work */
 /*body.namo-epub-library > section.chapter {
-  margin: 0 25pt;
+  margin: 0 5pt;
 }*/
 
 /* Use tighter margins and smaller font (18px) on phones (Nexus 4 and smaller) */
@@ -22,15 +35,19 @@ only screen and (max-device-width: 1280px) and (max-device-height: 768px) {
     font-size: 112.5%;
   }
 
-  body.gitden-reader {
-    margin: 0 5pt;
-  }
+  /*body.gitden-reader {
+    margin: 0 5pt !important;
+  }*/
 
   /*body.namo-epub-library > section.chapter {
     margin: 0 5pt;
   }*/
 }
 
+body.gitden-reader pre {
+  white-space: pre-wrap !important; /* Gitden must be applying white-space: pre !important */
+}
+
 body h1, body h2, body h3:not(.list-heading), body h4, body h5, body h6,
 h1 :not(code), h2 :not(code), h3:not(.list-heading) :not(code), h4 :not(code), h5 :not(code), h6 :not(code) {
   /* !important required to override custom font setting in Kindle / Gitden / Namo */
@@ -79,7 +96,7 @@ body code, body kbd, body pre, pre :not(code) {
   h1.chapter-title .subtitle > b:last-child {
     padding-right: 0;
   }
-  
+
   h1.chapter-title .subtitle::after {
     display: table;
     content: ' ';
@@ -104,13 +121,13 @@ body code, body kbd, body pre, pre :not(code) {
 .icon-1_5x {
   padding: 0 0.25em;
   -webkit-transform: scale(1.5, 1.5);
-  transform: scale(1.5, 1.5); 
+  transform: scale(1.5, 1.5);
 }
 
 .icon-2x {
   padding: 0 0.5em;
   -webkit-transform: scale(2, 2);
-  transform: scale(2, 2); 
+  transform: scale(2, 2);
 }
 
 .icon-small {

data/data/styles/epub3.css

@@ -25,7 +25,7 @@ html {
 body {
   padding: 0 !important;
   /* add margin to ~ match Kindle's narrow setting */
-  /* don't use !important on margin as it breaks calibre */
+  /* don't use !important on margin as it interferes with reader overrides (Calibre and Kindle) */
   margin: 0;
   font-size: 100%;
   /* NOTE putting optimizeLegibility on the body slows down rendering considerably */
@@ -38,10 +38,10 @@ html body {
   background-color: #FFFFFF;
 }
 
-/* sets minimum margin permitted */
-/* @page not supported by Kindle or GitDen */
+/* @page only applies to EPUB2 readers; not supported by EPUB3 readers such as Kindle and Gitden */
 @page {
-  /* push the top & bottom margins down in Aldiko to emulate Kindle (Kindle uses ~ 10% of screen by default )*/
+  /* sets minimum margin permitted */
+  /* pushes the top & bottom margins down in Aldiko to emulate Kindle (Kindle uses ~ 10% of screen by default )*/
   margin: 1cm;
 }
 
@@ -895,7 +895,7 @@ blockquote footer {
 
 blockquote footer .context {
   font-size: 0.9em;
-  letter-spacing: -0.1em;
+  letter-spacing: -0.05em;
   color: #666665;
 }
 
@@ -916,8 +916,7 @@ pre {
   margin-top: 1em; /* 0.85rem */
   /*margin-top: 1.176em;*/ /* 1rem */
   white-space: pre-wrap;
-  /*word-break: break-all;*/ /* break at the end of the line, no matter what */
-  word-wrap: break-word; /* break in middle of long word if no other break opportunities are available */
+  overflow-wrap: break-word; /* break in middle of long word if no other break opportunities are available */
   font-size: 0.85em;
   line-height: 1.4; /* matches what Kindle uses and can't go less */
   background-color: #F2F2F2;
@@ -985,28 +984,28 @@ aside[class~="admonition"] {
 }
 
 aside.note {
-  border-left-color: #B3B3B1;
-  background-color: #E1E1E1; /* 25% opacity of border */
+  border-left-color: #FFC14F;
+  background-color: #FFF0D4; /* 25% opacity of border */
 }
 
 aside.tip {
-  border-left-color: #57AD68;
-  background-color: #D4EAD9; /* 25% opacity of border */
+  border-left-color: #40403E;
+  background-color: #D0D0CF; /* 25% opacity of border */
 }
 
 aside.caution {
-  border-left-color: #666665;
-  background-color: #D8D8D8; /* 25% opacity of border */
+  border-left-color: #7F7F7D;
+  background-color: #DFDFDF; /* 25% opacity of border */
 }
 
 aside.warning {
   border-left-color: #C83737;
-  background-color: #F1CCCC; /* 25% opacity of border */
+  background-color: #F1CECE; /* 25% opacity of border */
 }
 
 aside.important {
-  border-left-color: #FFC14F;
-  background-color: #FFEFD2; /* 25% opacity of border */
+  border-left-color: #F2642A;
+  background-color: #FCD9CA; /* 25% opacity of border */
 }
 
 aside.admonition::before {
@@ -1018,7 +1017,7 @@ aside.admonition::before {
   text-align: center;
   margin-bottom: -0.25em;
   margin-left: -0.5em;
-  text-shadow: 0px 1px 2px rgba(102, 102, 101, 0.3);
+  text-shadow: 0px 1px 1px rgba(102, 102, 101, 0.15);
 }
 
 aside.admonition > div.content {
@@ -1039,41 +1038,38 @@ aside[class~="admonition"] > div[class~="content"] {
 }
 
 aside.note::before {
-  /*content: "\f0f4";*/ /* fa-coffee */
   content: "\f040"; /* fa-pencil */
-  color: #B3B3B1; /* 179,179,177 */
+  color: #FFC14F;
 }
 
 aside[class~="note"] > div[class~="content"] {
-  background-image: -webkit-linear-gradient(left, rgba(255,255,255,0) 42.5%, #B3B3B1 45%, #B3B3B1 55%, rgba(255,255,255,0) 57.5%);
-  background-image: linear-gradient(to right, rgba(255,255,255,0) 42.5%, #B3B3B1 45%, #B3B3B1 55%, rgba(255,255,255,0) 57.5%);
+  background-image: -webkit-linear-gradient(left, rgba(255,255,255,0) 42.5%, #FFC14F 45%, #FFC14F 55%, rgba(255,255,255,0) 57.5%);
+  background-image: linear-gradient(to right, rgba(255,255,255,0) 42.5%, #FFC14F 45%, #FFC14F 55%, rgba(255,255,255,0) 57.5%);
 }
 
 aside.tip::before {
-  /*content: "\f069";*/ /* fa-asterisk */
-  /*content: "\f0d6";*/ /* fa-money */
-  content: "\f15a"; /* fa-bitcoin */
-  color: #57AD68; /* 87,173,104 */
+  content: "\f0eb"; /* fa-lightbulb-o */
+  color: #40403E;
 }
 
 aside[class~="tip"] > div[class~="content"] {
-  background-image: -webkit-linear-gradient(left, rgba(255,255,255,0) 42.5%, #57AD68 45%, #57AD68 55%, rgba(255,255,255,0) 57.5%);
-  background-image: linear-gradient(to right, rgba(255,255,255,0) 42.5%, #57AD68 45%, #57AD68 55%, rgba(255,255,255,0) 57.5%);
+  background-image: -webkit-linear-gradient(left, rgba(255,255,255,0) 42.5%, #40403E 45%, #40403E 55%, rgba(255,255,255,0) 57.5%);
+  background-image: linear-gradient(to right, rgba(255,255,255,0) 42.5%, #40403E 45%, #40403E 55%, rgba(255,255,255,0) 57.5%);
 }
 
 aside.caution::before {
   content: "\f0c2"; /* fa-cloud */
-  color: #666665; /* 102,102,101 */ 
+  color: #7F7F7D;
 }
 
 aside[class~="caution"] > div[class~="content"] {
-  background-image: -webkit-linear-gradient(left, rgba(255,255,255,0) 42.5%, #666665 45%, #666665 55%, rgba(255,255,255,0) 57.5%);
-  background-image: linear-gradient(to right, rgba(255,255,255,0) 42.5%, #666665 45%, #666665 55%, rgba(255,255,255,0) 57.5%);
+  background-image: -webkit-linear-gradient(left, rgba(255,255,255,0) 42.5%, #7F7F7D 45%, #7F7F7D 55%, rgba(255,255,255,0) 57.5%);
+  background-image: linear-gradient(to right, rgba(255,255,255,0) 42.5%, #7F7F7D 45%, #7F7F7D 55%, rgba(255,255,255,0) 57.5%);
 }
 
 aside.warning::before {
   content: "\f0e7"; /* fa-bolt */
-  color: #C83737; /* 200,55,55 */
+  color: #C83737;
 }
 
 aside[class~="warning"] > div[class~="content"] {
@@ -1083,12 +1079,12 @@ aside[class~="warning"] > div[class~="content"] {
 
 aside.important::before {
   content: "\f12a"; /* fa-exclamation */
-  color: #FFC14F; /* 255,193,79 */
+  color: #F2642A;
 }
 
 aside[class~="important"] > div[class~="content"] {
-  background-image: -webkit-linear-gradient(left, rgba(255,255,255,0) 42.5%, #FFC14F 45%, #FFC14F 55%, rgba(255,255,255,0) 57.5%);
-  background-image: linear-gradient(to right, rgba(255,255,255,0) 42.5%, #FFC14F 45%, #FFC14F 55%, rgba(255,255,255,0) 57.5%);
+  background-image: -webkit-linear-gradient(left, rgba(255,255,255,0) 42.5%, #F2642A 45%, #F2642A 55%, rgba(255,255,255,0) 57.5%);
+  background-image: linear-gradient(to right, rgba(255,255,255,0) 42.5%, #F2642A 45%, #F2642A 55%, rgba(255,255,255,0) 57.5%);
 }
 
 aside.admonition > h2 {
@@ -1137,13 +1133,18 @@ table.table thead th {
   border-bottom: 1px solid #80807F;
 }
 
-table.table td > p {
+table.table td > p,
+table.table div.embed > * {
   margin-top: 0;
+}
+
+table.table td > p {
   text-align: left;
 }
 
 /* REVIEW */
-table.table td > p + p {
+table.table td > p + p,
+table.table div.embed > * + * {
   margin-top: 1em;
 }
 
@@ -1167,7 +1168,7 @@ table.table-framed-sides {
 
 table.table-grid th,
 table.table-grid td {
-  border-width: 0 1px 1px 0; 
+  border-width: 0 1px 1px 0;
   border-style: solid;
   border-color: #80807F;
 }
@@ -1183,14 +1184,14 @@ table.table-grid tbody tr:last-child > td {
 
 table.table-grid-rows tbody th,
 table.table-grid-rows tbody td {
-  border-width: 1px 0 0 0; 
+  border-width: 1px 0 0 0;
   border-style: solid;
   border-color: #80807F;
 }
 
 table.table-grid-cols th,
 table.table-grid-cols td {
-  border-width: 0 1px 0 0; 
+  border-width: 0 1px 0 0;
   border-style: solid;
   border-color: #80807F;
 }

data/lib/asciidoctor-epub3/converter.rb

@@ -4,9 +4,6 @@ require_relative 'font_icon_map'
 
 module Asciidoctor
 module Epub3
-# tried 8288, but it didn't work in older readers
-WordJoiner = [65279].pack 'U*'
-WordJoinerRx = RUBY_ENGINE_JRUBY ? /\uFEFF/ : WordJoiner
 
 # Public: The main converter for the epub3 backend that handles packaging the
 # EPUB3 or KF8 publication file.
@@ -25,15 +22,23 @@ class Converter
     @extract = false
   end
 
-  def convert spine_doc, name = nil
-    @validate = true if spine_doc.attr? 'ebook-validate'
-    @extract = true if spine_doc.attr? 'ebook-extract'
-    Packager.new spine_doc, (spine_doc.references[:spine_items] || [spine_doc]), spine_doc.attributes['ebook-format'].to_sym
+  def convert node, name = nil
+    if (name ||= node.node_name) == 'document'
+      @validate = node.attr? 'ebook-validate'
+      @extract = node.attr? 'ebook-extract'
+      @compress = node.attr 'ebook-compress'
+      Packager.new node, (node.references[:spine_items] || [node]), node.attributes['ebook-format'].to_sym
+    # converting an element from the spine document, such as an inline node in the doctitle
+    elsif name.start_with? 'inline_'
+      (@content_converter ||= ::Asciidoctor::Converter::Factory.default.create('epub3-xhtml5')).convert node, name
+    else
+      raise ::ArgumentError, %(Encountered unexpected node in epub3 package converter: #{name})
+    end
   end
 
   # FIXME we have to package in write because we don't have access to target before this point
   def write packager, target
-    packager.package validate: @validate, extract: @extract, target: target
+    packager.package validate: @validate, extract: @extract, compress: @compress, target: target
     nil
   end
 end
@@ -45,14 +50,14 @@ class ContentConverter
 
   register_for 'epub3-xhtml5'
 
-  WordJoiner = Epub3::WordJoiner
-  EOL = "\n"
+  EOL = %(\n)
   NoBreakSpace = ' '
   ThinNoBreakSpace = ' '
   RightAngleQuote = '›'
+  CalloutStartNum = %(\u2460)
 
   XmlElementRx = /<\/?.+?>/
-  CharEntityRx = /&#(\d{2,5});/
+  CharEntityRx = /&#(\d{2,6});/
   NamedEntityRx = /&([A-Z]+);/
   UppercaseTagRx = /<(\/)?([A-Z]+)>/
 
@@ -80,7 +85,7 @@ class ContentConverter
     basebackend 'html'
     outfilesuffix '.xhtml'
     htmlsyntax 'xml'
-    @xrefs_used = ::Set.new
+    @xrefs_seen = ::Set.new
     @icon_names = []
   end
 
@@ -88,36 +93,10 @@ class ContentConverter
     if respond_to?(name ||= node.node_name)
       send name, node
     else
-      warn %(conversion missing in epub3 backend for #{name})
+      warn %(asciidoctor: WARNING: conversion missing in epub3 backend for #{name})
     end
   end
 
-  # TODO aggregate authors of spine document into authors attribute(s) on main document
-  def navigation_document node, spine
-    doctitle_sanitized = (node.doctitle sanitize: true, use_fallback: true).gsub WordJoinerRx, ''
-    lines = [%(<!DOCTYPE html>
-<html xmlns="http://www.w3.org/1999/xhtml" xmlns:epub="http://www.idpf.org/2007/ops" xml:lang="#{lang = (node.attr 'lang', 'en')}" lang="#{lang}">
-<head>
-<meta charset="UTF-8"/>
-<title>#{doctitle_sanitized}</title>
-<link rel="stylesheet" type="text/css" href="styles/epub3.css"/>
-<link rel="stylesheet" type="text/css" href="styles/epub3-css3-only.css" media="(min-device-width: 0px)"/>
-</head>
-<body>
-<h1>#{doctitle_sanitized}</h1>
-<nav epub:type="toc" id="toc">
-<h2>#{node.attr 'toc-title'}</h2>
-<ol>)]
-    spine.each do |item|
-      lines << %(<li><a href="#{item.id || (item.attr 'docname')}.xhtml">#{(item.doctitle sanitize: true, use_fallback: true).gsub WordJoinerRx, ''}</a></li>)
-    end
-    lines << %(</ol>
-</nav>
-</body>
-</html>)
-    lines * EOL
-  end
-
   def document node
     docid = node.id
 
@@ -131,19 +110,22 @@ class ContentConverter
       subtitle = doctitle.combined
     end
 
-    doctitle_sanitized = (node.doctitle sanitize: true, use_fallback: true).gsub WordJoinerRx, ''
-    subtitle_formatted = subtitle.gsub(WordJoinerRx, '').split(' ').map {|w| %(<b>#{w}</b>) } * ' '
-    # FIXME make this uppercase routine more intelligent, less fragile
+    doctitle_sanitized = doctitle.combined
+    subtitle_formatted = subtitle.split.map {|w| %(<b>#{w}</b>) } * ' '
+    # FIXME use uppercase pcdata helper to make less fragile (see logic in Asciidoctor PDF)
     subtitle_formatted_upper = subtitle_formatted.upcase
         .gsub(UppercaseTagRx) { %(<#{$1}#{$2.downcase}>) }
         .gsub(NamedEntityRx) { %(&#{$1.downcase};) }
 
-    author = node.attr 'author'
-    username = node.attr 'username', 'default'
-    # FIXME needs to resolve to the imagesdir of the spine document, not this document
-    #imagesdir = (node.attr 'imagesdir', '.').chomp '/'
-    #imagesdir = (imagesdir == '.' ? nil : %(#{imagesdir}/))
-    imagesdir = 'images/'
+    if (node.attr 'publication-type', 'book') == 'book'
+      byline = nil
+    else
+      author = node.attr 'author'
+      username = node.attr 'username', 'default'
+      imagesdir = (node.references[:spine].attr 'imagesdir', '.').chomp '/'
+      imagesdir = (imagesdir == '.' ? nil : %(#{imagesdir}/))
+      byline = %(<p class="byline"><img src="#{imagesdir}avatars/#{username}.jpg"/> <b class="author">#{author}</b></p>#{EOL})
+    end
 
     mark_last_paragraph node
     content = node.content
@@ -176,14 +158,12 @@ class ContentConverter
 <link rel="stylesheet" type="text/css" href="styles/epub3.css"/>
 <link rel="stylesheet" type="text/css" href="styles/epub3-css3-only.css" media="(min-device-width: 0px)"/>
 #{icon_css_head}<script type="text/javascript">
-document.addEventListener('DOMContentLoaded', function(event) {
-  var epubReader = navigator.epubReadingSystem;
-  if (!epubReader) {
-    if (window.parent == window || !(epubReader = window.parent.navigator.epubReadingSystem)) {
-      return;
-    }
+document.addEventListener('DOMContentLoaded', function(event, reader) {
+  if (!(reader = navigator.epubReadingSystem)) {
+    if (navigator.userAgent.indexOf(' calibre/') >= 0) reader = { name: 'calibre-desktop' };
+    else if (window.parent == window || !(reader = window.parent.navigator.epubReadingSystem)) return;
   }
-  document.body.setAttribute('class', epubReader.name.toLowerCase().replace(/ /g, '-'));
+  document.body.setAttribute('class', reader.name.toLowerCase().replace(/ /g, '-'));
 });
 </script>
 </head>
@@ -191,8 +171,7 @@ document.addEventListener('DOMContentLoaded', function(event) {
 <section class="chapter" title="#{doctitle_sanitized.gsub '"', '&quot;'}" epub:type="chapter" id="#{docid}">
 #{icon_css_scoped}<header>
 <div class="chapter-header">
-<p class="byline"><img src="#{imagesdir}avatars/#{username}.jpg"/> <b class="author">#{author}</b></p>
-<h1 class="chapter-title">#{title_upper}#{subtitle ? %[ <small class="subtitle">#{subtitle_formatted_upper}</small>] : nil}</h1>
+#{byline}<h1 class="chapter-title">#{title_upper}#{subtitle ? %[ <small class="subtitle">#{subtitle_formatted_upper}</small>] : nil}</h1>
 </div>
 </header>
 #{content})]
@@ -219,6 +198,11 @@ document.addEventListener('DOMContentLoaded', function(event) {
     lines * EOL
   end
 
+  # NOTE embedded is used for AsciiDoc table cell content
+  def embedded node
+    node.content
+  end
+
   def section node
     hlevel = node.level + 1
     epub_type_attr = node.special ? %( epub:type="#{node.sectname}") : nil
@@ -248,9 +232,16 @@ document.addEventListener('DOMContentLoaded', function(event) {
     end
   end
 
-  # QUESTION use convert_content?
   def open node
-    node.content
+    id_attr = node.id ? %( id="#{node.id}") : nil
+    class_attr = node.role ? %( class="#{node.role}") : nil
+    if id_attr || class_attr
+      %(<div#{id_attr}#{class_attr}>
+#{convert_content node}
+</div>)
+    else
+      convert_content node
+    end
   end
 
   def abstract node
@@ -263,11 +254,10 @@ document.addEventListener('DOMContentLoaded', function(event) {
     role = node.role
     # stack-head is the alternative to the default, inline-head (where inline means "run-in")
     head_stop = node.attr 'head-stop', (role && (node.has_role? 'stack-head') ? nil : '.')
+    # FIXME promote regexp to constant
     head = node.title? ? %(<strong class="head">#{title = node.title}#{head_stop && title !~ /[[:punct:]]$/ ? head_stop : nil}</strong> ) : nil
     if role
-      if node.has_role? 'signature'
-        node.set_option 'hardbreaks'
-      end
+      node.set_option 'hardbreaks' if node.has_role? 'signature'
       %(<p class="#{role}">#{head}#{node.content}</p>)
     else
       %(<p>#{head}#{node.content}</p>)
@@ -330,15 +320,11 @@ document.addEventListener('DOMContentLoaded', function(event) {
   def listing node
     figure_classes = ['listing']
     figure_classes << 'coalesce' if node.option? 'unbreakable'
-    pre_classes = if node.style == 'source'
-      ['source', %(language-#{node.attr 'language'})]
-    else
-      ['screen']
-    end
+    pre_classes = node.style == 'source' ? ['source', %(language-#{node.attr 'language'})] : ['screen']
     title_div = node.title? ? %(<figcaption>#{node.captioned_title}</figcaption>
 ) : nil
     # patches conums to fix extra or missing leading space
-    # TODO apply this patch upstream to Asciidoctor
+    # TODO remove patch once upgrading to Asciidoctor 1.5.6
     %(<figure class="#{figure_classes * ' '}">
 #{title_div}<pre class="#{pre_classes * ' '}"><code>#{(node.content || '').gsub(/(?<! )<i class="conum"| +<i class="conum"/, ' <i class="conum"')}</code></pre>
 </figure>)
@@ -359,11 +345,11 @@ document.addEventListener('DOMContentLoaded', function(event) {
 
   def quote node
     footer_content = []
-    if attribution = (node.attr 'attribution')
-      footer_content << attribution  
+    if (attribution = node.attr 'attribution')
+      footer_content << attribution
     end
 
-    if citetitle = (node.attr 'citetitle')
+    if (citetitle = node.attr 'citetitle')
       citetitle_sanitized = xml_sanitize citetitle
       footer_content << %(<cite title="#{citetitle_sanitized}">#{citetitle}</cite>)
     end
@@ -386,11 +372,11 @@ document.addEventListener('DOMContentLoaded', function(event) {
 
   def verse node
     footer_content = []
-    if attribution = (node.attr 'attribution')
-      footer_content << attribution  
+    if (attribution = node.attr 'attribution')
+      footer_content << attribution
     end
 
-    if citetitle = (node.attr 'citetitle')
+    if (citetitle = node.attr 'citetitle')
       citetitle_sanitized = xml_sanitize citetitle
       footer_content << %(<cite title="#{citetitle_sanitized}">#{citetitle}</cite>)
     end
@@ -409,6 +395,7 @@ document.addEventListener('DOMContentLoaded', function(event) {
       title = node.title
       title_sanitized = xml_sanitize title
       title_attr = %( title="#{title_sanitized}")
+      # FIXME use uppercase pcdata helper to make less fragile (see logic in Asciidoctor PDF)
       title_upper = title.upcase.gsub(NamedEntityRx) { %(&#{$1.downcase};) }
       title_el = %(<h2>#{title_upper}</h2>
 )
@@ -444,10 +431,10 @@ document.addEventListener('DOMContentLoaded', function(event) {
     end
     table_class_attr = %( class="#{table_classes * ' '}")
     table_styles = []
-    unless node.option? 'autowidth'
-      table_styles << %(width: #{node.attr 'tablepcwidth'}%;)
+    unless (node.option? 'autowidth') && !(node.attr? 'width', nil, false)
+      table_styles << %(width: #{node.attr 'tablepcwidth'}%)
     end
-    table_style_attr = table_styles.size > 0 ? %( style="#{table_styles * ' '}") : nil
+    table_style_attr = table_styles.size > 0 ? %( style="#{table_styles * '; '}") : nil
 
     lines << %(<table#{table_id_attr}#{table_class_attr}#{table_style_attr}>)
     lines << %(<caption>#{node.captioned_title}</caption>) if node.title?
@@ -460,7 +447,7 @@ document.addEventListener('DOMContentLoaded', function(event) {
         end
       #else
       #  node.columns.each do |col|
-      #    lines << %(<col style="width: #{col.attr 'colpcwidth'}%;"/>)
+      #    lines << %(<col style="width: #{col.attr 'colpcwidth'}%"/>)
       #  end
       #end
       lines << '</colgroup>'
@@ -474,7 +461,7 @@ document.addEventListener('DOMContentLoaded', function(event) {
             else
               case cell.style
               when :asciidoc
-                cell_content = %(<div>#{cell.content}</div>)
+                cell_content = %(<div class="embed">#{cell.content}</div>)
               when :verse
                 cell_content = %(<div class="verse">#{cell.text}</div>)
               when :literal
@@ -498,7 +485,7 @@ document.addEventListener('DOMContentLoaded', function(event) {
             cell_class_attr = cell_classes.size > 0 ? %( class="#{cell_classes * ' '}") : nil
             cell_colspan_attr = cell.colspan ? %( colspan="#{cell.colspan}") : nil
             cell_rowspan_attr = cell.rowspan ? %( rowspan="#{cell.rowspan}") : nil
-            cell_style_attr = (node.document.attr? 'cellbgcolor') ? %( style="background-color: #{node.document.attr 'cellbgcolor'};") : nil
+            cell_style_attr = (node.document.attr? 'cellbgcolor') ? %( style="background-color: #{node.document.attr 'cellbgcolor'}") : nil
             lines << %(<#{cell_tag_name}#{cell_class_attr}#{cell_colspan_attr}#{cell_rowspan_attr}#{cell_style_attr}>#{cell_content}</#{cell_tag_name}>)
           end
           lines << '</tr>'
@@ -515,7 +502,7 @@ document.addEventListener('DOMContentLoaded', function(event) {
   def colist node
     lines = ['<div class="callout-list">
 <ol>']
-    num = "\u2460"
+    num = CalloutStartNum
     node.items.each_with_index do |item, i|
       lines << %(<li><i class="conum" data-value="#{i + 1}">#{num}</i> #{item.text}</li>)
       num = num.next
@@ -545,7 +532,7 @@ document.addEventListener('DOMContentLoaded', function(event) {
         lines << '<li>'
         if dd
           # NOTE: must wrap remaining text in a span to help webkit justify the text properly
-          lines << %(<span class="principal">#{subject_element}#{dd.text? ? %[ <span class="supporting">#{dd.text}</span>] : nil}</span>) 
+          lines << %(<span class="principal">#{subject_element}#{dd.text? ? %[ <span class="supporting">#{dd.text}</span>] : nil}</span>)
           lines << dd.content if dd.blocks?
         else
           lines << %(<span class="principal">#{subject_element}</span>)
@@ -611,7 +598,6 @@ document.addEventListener('DOMContentLoaded', function(event) {
   def ulist node
     complex = false
     div_classes = ['itemized-list', node.style, node.role].compact
-    # TODO could strip WordJoiner if brief since not using justify
     ul_classes = [node.style, ((node.option? 'brief') ? 'brief' : nil)].compact
     ul_class_attr = ul_classes.empty? ? nil : %( class="#{ul_classes * ' '}")
     id_attribute = node.id ? %( id="#{node.id}") : nil
@@ -642,7 +628,7 @@ document.addEventListener('DOMContentLoaded', function(event) {
     img_attrs = [%(alt="#{node.attr 'alt'}")]
     case type
     when 'svg'
-      img_attrs << %(style="width: #{node.attr 'scaledwidth', '100%'};")
+      img_attrs << %(style="width: #{node.attr 'scaledwidth', '100%'}")
       # TODO make this a convenience method on document
       epub_properties = (node.document.attr 'epub-properties') || []
       unless epub_properties.include? 'svg'
@@ -651,7 +637,7 @@ document.addEventListener('DOMContentLoaded', function(event) {
       end
     else
       if node.attr? 'scaledwidth'
-        img_attrs << %(style="width: #{node.attr 'scaledwidth'};")
+        img_attrs << %(style="width: #{node.attr 'scaledwidth'}")
       end
     end
 =begin
@@ -664,7 +650,7 @@ document.addEventListener('DOMContentLoaded', function(event) {
       #  img_attrs << %(width="#{node.attr 'scaledheight'}" height="#{node.attr 'scaledheight'}")
       # ePub3
       elsif node.attr? 'scaledheight'
-        img_attrs << %(height="#{node.attr 'scaledheight'}" style="max-height: #{node.attr 'scaledheight'} !important;")
+        img_attrs << %(height="#{node.attr 'scaledheight'}" style="max-height: #{node.attr 'scaledheight'} !important")
       else
         # Aldiko doesn't not scale width to 100% by default
         img_attrs << %(width="100%")
@@ -682,22 +668,58 @@ document.addEventListener('DOMContentLoaded', function(event) {
   def inline_anchor node
     target = node.target
     case node.type
-    when :xref
-      refid = (node.attr 'refid') || target
-      id_attr = unless @xrefs_used.include? refid
-        @xrefs_used << refid
-        %( id="xref-#{refid}")
+    when :xref # TODO would be helpful to know what type the target is (e.g., bibref)
+      doc, refid, text, path = node.document, ((node.attr 'refid') || target), node.text, (node.attr 'path')
+      # NOTE if path is non-nil, we have an inter-document xref
+      # QUESTION should we drop the id attribute for an inter-document xref?
+      if path
+        # ex. chapter-id#section-id
+        if node.attr 'fragment'
+          refdoc_id, refdoc_refid = refid.split '#', 2
+          if refdoc_id == refdoc_refid
+            target = target[0...(target.index '#')]
+            id_attr = %( id="xref--#{refdoc_id}")
+          else
+            id_attr = %( id="xref--#{refdoc_id}--#{refdoc_refid}")
+          end
+        # ex. chapter-id#
+        else
+          refdoc_id = refdoc_refid = refid
+          # inflate key to spine item root (e.g., transform chapter-id to chapter-id#chapter-id)
+          refid = %(#{refid}##{refid})
+          id_attr = %( id="xref--#{refdoc_id}")
+        end
+        id_attr = nil unless @xrefs_seen.add? refid
+        refdoc = doc.references[:spine_items].find {|it| refdoc_id == (it.id || (it.attr 'docname')) }
+        if refdoc
+          if (reftext = refdoc.references[:ids][refdoc_refid])
+            text ||= reftext
+          else
+            warn %(asciidoctor: WARNING: #{::File.basename(doc.attr 'docfile')}: invalid reference to unknown anchor in #{refdoc_id} chapter: #{refdoc_refid})
+          end
+        else
+          warn %(asciidoctor: WARNING: #{::File.basename(doc.attr 'docfile')}: invalid reference to anchor in unknown chapter: #{refdoc_id})
+        end
+      else
+        id_attr = (@xrefs_seen.add? refid) ? %( id="xref-#{refid}") : nil
+        if (reftext = doc.references[:ids][refid])
+          text ||= reftext
+        else
+          # FIXME we get false negatives for reference to bibref
+          warn %(asciidoctor: WARNING: #{::File.basename(doc.attr 'docfile')}: invalid reference to unknown local anchor (or valid bibref): #{refid})
+        end
       end
-      # FIXME seems like text should be prepared already
-      # FIXME would be nice to know what type the target is (e.g., bibref)
-      text = node.text || (node.document.references[:ids][refid] || %([#{refid}]))
-      %(<a#{id_attr} href="#{target}" class="xref">#{text}</a>#{WordJoiner})
+      %(<a#{id_attr} href="#{target}" class="xref">#{text || "[#{refid}]"}</a>)
     when :ref
       %(<a id="#{target}"></a>)
     when :link
-      %(<a href="#{target}" class="link">#{node.text}</a>#{WordJoiner})
+      %(<a href="#{target}" class="link">#{node.text}</a>)
     when :bibref
-      %(<a id="#{target}" href="#xref-#{target}">[#{target}]</a>#{WordJoiner})
+      if @xrefs_seen.include? target
+        %(<a id="#{target}" href="#xref-#{target}">[#{target}]</a>)
+      else
+        %(<a id="#{target}"></a>[#{target}])
+      end
     end
   end
 
@@ -706,11 +728,11 @@ document.addEventListener('DOMContentLoaded', function(event) {
   end
 
   def inline_button node
-    %(<b class="button">[<span class="label">#{node.text}</span>]</b>#{WordJoiner})
+    %(<b class="button">[<span class="label">#{node.text}</span>]</b>)
   end
 
   def inline_callout node
-    num = "\u2460"
+    num = CalloutStartNum
     int_num = node.text.to_i
     (int_num - 1).times { num = num.next }
     %(<i class="conum" data-value="#{int_num}">#{num}</i>)
@@ -725,7 +747,7 @@ document.addEventListener('DOMContentLoaded', function(event) {
   end
 
   def inline_image node
-    if (type = node.type) == 'icon'
+    if node.type == 'icon'
       @icon_names << (icon_name = node.target)
       i_classes = ['icon', %(i-#{icon_name})]
       i_classes << %(icon-#{node.attr 'size'}) if node.attr? 'size'
@@ -748,7 +770,7 @@ document.addEventListener('DOMContentLoaded', function(event) {
     if (keys = node.attr 'keys').size == 1
       %(<kbd>#{keys[0]}</kbd>)
     else
-      key_combo = keys.map {|key| %(<kbd>#{key}</kbd>+) }.join.chop
+      key_combo = keys.map {|key| %(<kbd>#{key}</kbd>) }.join '+'
       %(<span class="keyseq">#{key_combo}</span>)
     end
   end
@@ -770,11 +792,11 @@ document.addEventListener('DOMContentLoaded', function(event) {
   def inline_quoted node
     case node.type
     when :strong
-      %(<strong>#{node.text}</strong>#{WordJoiner})
+      %(<strong>#{node.text}</strong>)
     when :emphasis
-      %(<em>#{node.text}</em>#{WordJoiner})
+      %(<em>#{node.text}</em>)
     when :monospaced
-      %(<code class="literal">#{node.text}</code>#{WordJoiner})
+      %(<code class="literal">#{node.text}</code>)
     when :double
       #%(&#x201c;#{node.text}&#x201d;)
       %(“#{node.text}”)
@@ -782,28 +804,26 @@ document.addEventListener('DOMContentLoaded', function(event) {
       #%(&#x2018;#{node.text}&#x2019;)
       %(‘#{node.text}’)
     when :superscript
-      %(<sup>#{node.text}</sup>#{WordJoiner})
+      %(<sup>#{node.text}</sup>)
     when :subscript
-      %(<sub>#{node.text}</sub>#{WordJoiner})
+      %(<sub>#{node.text}</sub>)
     else
       node.text
     end
   end
 
   def convert_content node
-    if node.content_model == :simple
-      %(<p>#{node.content}</p>)
-    else
-      node.content
-    end
+    node.content_model == :simple ? %(<p>#{node.content}</p>) : node.content
   end
 
+  # FIXME merge into with xml_sanitize helper
   def xml_sanitize value, target = :attribute
-    sanitized = (value.include? '<') ? value.gsub(XmlElementRx, '').tr_s(' ', ' ').strip : value
+    sanitized = (value.include? '<') ? value.gsub(XmlElementRx, '').strip.tr_s(' ', ' ') : value
     if target == :plain && (sanitized.include? ';')
-      sanitized = sanitized.gsub(CharEntityRx) { [$1.to_i].pack('U*') }.gsub(FromHtmlSpecialCharsRx, FromHtmlSpecialCharsMap)
+      sanitized = sanitized.gsub(CharEntityRx) { [$1.to_i].pack 'U*' } if sanitized.include? '&#'
+      sanitized = sanitized.gsub(FromHtmlSpecialCharsRx, FromHtmlSpecialCharsMap)
     elsif target == :attribute
-      sanitized = sanitized.gsub(WordJoinerRx, '').gsub('"', '&quot;')
+      sanitized = sanitized.gsub '"', '&quot;' if sanitized.include? '"'
     end
     sanitized
   end
@@ -822,17 +842,58 @@ document.addEventListener('DOMContentLoaded', function(event) {
 end
 
 class DocumentIdGenerator
+  ReservedIds = %w(cover nav ncx)
+  CharRefRx = /&(?:([a-zA-Z]{2,})|#(\d{2,6})|#x([a-fA-F0-9]{2,5}));/
+  if defined? __dir__
+    InvalidIdCharsRx = /[^\p{Word}]+/
+    LeadingDigitRx = /^\p{Nd}/
+  else
+    InvalidIdCharsRx = /[^[:word:]]+/
+    LeadingDigitRx = /^[[:digit:]]/
+  end
   class << self
-    def generate_id doc
+    def generate_id doc, pre = nil, sep = nil
+      synthetic = false
       unless (id = doc.id)
-        id = if doc.header?
-          doc.doctitle(sanitize: true).gsub(WordJoinerRx, '').downcase.delete(':').tr_s(' ', '-').tr_s('-', '-')
+        # NOTE we assume pre is a valid ID prefix and that pre and sep only contain valid ID chars
+        pre ||= '_'
+        sep = sep ? sep.chr : '_'
+        if doc.header?
+          id = doc.doctitle sanitize: true
+          id = id.gsub CharRefRx do
+            $1 ? ($1 == 'amp' ? 'and' : sep) : ((d = $2 ? $2.to_i : $3.hex) == 8217 ? '' : ([d].pack 'U*'))
+          end if id.include? '&'
+          id = id.downcase.gsub InvalidIdCharsRx, sep
+          if id.empty?
+            id, synthetic = nil, true
+          else
+            unless sep.empty?
+              if (id = id.tr_s sep, sep).end_with? sep
+                if id == sep
+                  id, synthetic = nil, true
+                else
+                  id = (id.start_with? sep) ? id[1..-2] : id.chop
+                end
+              elsif id.start_with? sep
+                id = id[1..-1]
+              end
+            end
+            unless synthetic
+              if pre.empty?
+                id = %(_#{id}) if LeadingDigitRx =~ id
+              elsif !(id.start_with? pre)
+                id = %(#{pre}#{id})
+              end
+            end
+          end
         elsif (first_section = doc.first_section)
-          first_section.id
+          id = first_section.id
         else
-          %(document-#{doc.object_id})
+          synthetic = true
         end
+        id = %(#{pre}document#{sep}#{doc.object_id}) if synthetic
       end
+      warn %(asciidoctor: ERROR: chapter uses a reserved ID: #{id}) if !synthetic && (ReservedIds.include? id)
       id
     end
   end
@@ -854,14 +915,20 @@ Extensions.register do
     when 'epub3', 'kf8'
       # all good
     when 'mobi'
-      document.attributes['ebook-format'] = 'kf8'
+      ebook_format = document.attributes['ebook-format'] = 'kf8'
     else
+      # QUESTION should we display a warning?
       ebook_format = document.attributes['ebook-format'] = 'epub3'
     end
     document.attributes[%(ebook-format-#{ebook_format})] = ''
     # Only fire SpineItemProcessor for top-level include directives
     include_processor SpineItemProcessor.new(document)
-    treeprocessor { process {|doc| doc.id = DocumentIdGenerator.generate_id doc } }
+    treeprocessor do
+      process do |doc|
+        doc.id = DocumentIdGenerator.generate_id doc, (doc.attr 'idprefix'), (doc.attr 'idseparator')
+        nil
+      end
+    end
   end
 end
 end