diff options
Diffstat (limited to 'lib/banzai')
| -rw-r--r-- | lib/banzai/filter/blockquote_fence_filter.rb | 6 | ||||
| -rw-r--r-- | lib/banzai/filter/kroki_filter.rb | 11 | ||||
| -rw-r--r-- | lib/banzai/filter/math_filter.rb | 100 | ||||
| -rw-r--r-- | lib/banzai/filter/plantuml_filter.rb | 3 | ||||
| -rw-r--r-- | lib/banzai/filter/references/reference_filter.rb | 12 | ||||
| -rw-r--r-- | lib/banzai/pipeline/markup_pipeline.rb | 4 |
6 files changed, 116 insertions, 20 deletions
diff --git a/lib/banzai/filter/blockquote_fence_filter.rb b/lib/banzai/filter/blockquote_fence_filter.rb index e07cbfe8d85..e5cf20d00df 100644 --- a/lib/banzai/filter/blockquote_fence_filter.rb +++ b/lib/banzai/filter/blockquote_fence_filter.rb @@ -6,13 +6,13 @@ module Banzai REGEX = %r{ #{::Gitlab::Regex.markdown_code_or_html_blocks} | - (?=^>>>\ *\n.*\n>>>\ *$)(?: + (?=(?<=^\n|\A)>>>\ *\n.*\n>>>\ *(?=\n$|\z))(?: # Blockquote: # >>> # Anything, including code and HTML blocks # >>> - ^>>>\ *\n + (?<=^\n|\A)>>>\ *\n (?<quote> (?: # Any character that doesn't introduce a code or HTML block @@ -30,7 +30,7 @@ module Banzai \g<html> )+? ) - \n>>>\ *$ + \n>>>\ *(?=\n$|\z) ) }mx.freeze diff --git a/lib/banzai/filter/kroki_filter.rb b/lib/banzai/filter/kroki_filter.rb index 845c7f2bc0a..713ff2439fc 100644 --- a/lib/banzai/filter/kroki_filter.rb +++ b/lib/banzai/filter/kroki_filter.rb @@ -14,7 +14,10 @@ module Banzai return doc unless settings.kroki_enabled diagram_selectors = ::Gitlab::Kroki.formats(settings) - .map { |diagram_type| %(pre[lang="#{diagram_type}"] > code) } + .map do |diagram_type| + %(pre[lang="#{diagram_type}"] > code, + pre > code[lang="#{diagram_type}"]) + end .join(', ') xpath = Gitlab::Utils::Nokogiri.css_to_xpath(diagram_selectors) @@ -22,7 +25,7 @@ module Banzai diagram_format = "svg" doc.xpath(xpath).each do |node| - diagram_type = node.parent['lang'] + diagram_type = node.parent['lang'] || node['lang'] diagram_src = node.content image_src = create_image_src(diagram_type, diagram_format, diagram_src) img_tag = Nokogiri::HTML::DocumentFragment.parse(%(<img src="#{image_src}" />)) @@ -33,8 +36,8 @@ module Banzai img_tag.set_attribute('hidden', '') if lazy_load img_tag.set_attribute('class', 'js-render-kroki') - img_tag.set_attribute('data-diagram', node.parent['lang']) - img_tag.set_attribute('data-diagram-src', "data:text/plain;base64,#{Base64.strict_encode64(node.content)}") + img_tag.set_attribute('data-diagram', diagram_type) + img_tag.set_attribute('data-diagram-src', "data:text/plain;base64,#{Base64.strict_encode64(diagram_src)}") node.parent.replace(img_tag) end diff --git a/lib/banzai/filter/math_filter.rb b/lib/banzai/filter/math_filter.rb index 0ac506776be..1ca4b2c89db 100644 --- a/lib/banzai/filter/math_filter.rb +++ b/lib/banzai/filter/math_filter.rb @@ -7,7 +7,7 @@ require 'uri' # - app/assets/javascripts/behaviors/markdown/nodes/code_block.js module Banzai module Filter - # HTML filter that adds class="code math" and removes the dollar sign in $`2+2`$. + # HTML filter that implements our math syntax, adding class="code math" # class MathFilter < HTML::Pipeline::Filter CSS_MATH = 'pre.code.language-math' @@ -15,14 +15,42 @@ module Banzai CSS_CODE = 'code' XPATH_CODE = Gitlab::Utils::Nokogiri.css_to_xpath(CSS_CODE).freeze + # These are based on the Pandoc heuristics, + # https://pandoc.org/MANUAL.html#extension-tex_math_dollars + # Note: at this time, using a dollar sign literal, `\$` inside + # a math statement does not work correctly. + # Corresponds to the "$...$" syntax + DOLLAR_INLINE_PATTERN = %r{ + (?<matched>\$(?<math>(?:\S[^$\n]*?\S|[^$\s]))\$)(?:[^\d]|$) + }x.freeze + + # Corresponds to the "$$...$$" syntax + DOLLAR_DISPLAY_INLINE_PATTERN = %r{ + (?<matched>\$\$\ *(?<math>[^$\n]+?)\ *\$\$) + }x.freeze + + # Corresponds to the $$\n...\n$$ syntax + DOLLAR_DISPLAY_BLOCK_PATTERN = %r{ + ^(?<matched>\$\$\ *\n(?<math>.*)\n\$\$\ *)$ + }x.freeze + + # Order dependent. Handle the `$$` syntax before the `$` syntax + DOLLAR_MATH_PIPELINE = [ + { pattern: DOLLAR_DISPLAY_INLINE_PATTERN, tag: :code, style: :display }, + { pattern: DOLLAR_DISPLAY_BLOCK_PATTERN, tag: :pre, style: :display }, + { pattern: DOLLAR_INLINE_PATTERN, tag: :code, style: :inline } + ].freeze + + # Do not recognize math inside these tags + IGNORED_ANCESTOR_TAGS = %w[pre code tt].to_set + # Attribute indicating inline or display math. STYLE_ATTRIBUTE = 'data-math-style' # Class used for tagging elements that should be rendered TAG_CLASS = 'js-render-math' - INLINE_CLASSES = "code math #{TAG_CLASS}" - + MATH_CLASSES = "code math #{TAG_CLASS}" DOLLAR_SIGN = '$' # Limit to how many nodes can be marked as math elements. @@ -31,8 +59,48 @@ module Banzai RENDER_NODES_LIMIT = 50 def call - nodes_count = 0 + @nodes_count = 0 + + process_dollar_pipeline if Feature.enabled?(:markdown_dollar_math, group) + + process_dollar_backtick_inline + process_math_codeblock + + doc + end + + def process_dollar_pipeline + doc.xpath('descendant-or-self::text()').each do |node| + next if has_ancestor?(node, IGNORED_ANCESTOR_TAGS) + + node_html = node.to_html + next unless node_html.match?(DOLLAR_INLINE_PATTERN) || + node_html.match?(DOLLAR_DISPLAY_INLINE_PATTERN) || + node_html.match?(DOLLAR_DISPLAY_BLOCK_PATTERN) + + temp_doc = Nokogiri::HTML.fragment(node_html) + DOLLAR_MATH_PIPELINE.each do |pipeline| + temp_doc.xpath('child::text()').each do |temp_node| + html = temp_node.to_html + temp_node.content.scan(pipeline[:pattern]).each do |matched, math| + html.sub!(matched, math_html(tag: pipeline[:tag], style: pipeline[:style], math: math)) + @nodes_count += 1 + break if @nodes_count >= RENDER_NODES_LIMIT + end + + temp_node.replace(html) + + break if @nodes_count >= RENDER_NODES_LIMIT + end + end + + node.replace(temp_doc) + end + end + + # Corresponds to the "$`...`$" syntax + def process_dollar_backtick_inline doc.xpath(XPATH_CODE).each do |code| closing = code.next opening = code.previous @@ -44,22 +112,38 @@ module Banzai closing.content.first == DOLLAR_SIGN && opening.content.last == DOLLAR_SIGN - code[:class] = INLINE_CLASSES + code[:class] = MATH_CLASSES code[STYLE_ATTRIBUTE] = 'inline' closing.content = closing.content[1..] opening.content = opening.content[0..-2] - nodes_count += 1 - break if nodes_count >= RENDER_NODES_LIMIT + @nodes_count += 1 + break if @nodes_count >= RENDER_NODES_LIMIT end end + end + # corresponds to the "```math...```" syntax + def process_math_codeblock doc.xpath(XPATH_MATH).each do |el| el[STYLE_ATTRIBUTE] = 'display' el[:class] += " #{TAG_CLASS}" end + end - doc + private + + def math_html(tag:, math:, style:) + case tag + when :code + "<code class=\"#{MATH_CLASSES}\" data-math-style=\"#{style}\">#{math}</code>" + when :pre + "<pre class=\"#{MATH_CLASSES}\" data-math-style=\"#{style}\"><code>#{math}</code></pre>" + end + end + + def group + context[:group] || context[:project]&.group end end end diff --git a/lib/banzai/filter/plantuml_filter.rb b/lib/banzai/filter/plantuml_filter.rb index cbcd547120d..82f6247cf03 100644 --- a/lib/banzai/filter/plantuml_filter.rb +++ b/lib/banzai/filter/plantuml_filter.rb @@ -31,7 +31,8 @@ module Banzai private def lang_tag - @lang_tag ||= Gitlab::Utils::Nokogiri.css_to_xpath('pre[lang="plantuml"] > code').freeze + @lang_tag ||= Gitlab::Utils::Nokogiri + .css_to_xpath('pre[lang="plantuml"] > code, pre > code[lang="plantuml"]').freeze end def settings diff --git a/lib/banzai/filter/references/reference_filter.rb b/lib/banzai/filter/references/reference_filter.rb index 97ef71036a2..37734f6a45a 100644 --- a/lib/banzai/filter/references/reference_filter.rb +++ b/lib/banzai/filter/references/reference_filter.rb @@ -15,6 +15,8 @@ module Banzai include RequestStoreReferenceCache include OutputSafety + REFERENCE_TYPE_DATA_ATTRIBUTE = 'data-reference-type=' + class << self # Implement in child class # Example: self.reference_type = :merge_request @@ -132,13 +134,19 @@ module Banzai def data_attribute(attributes = {}) attributes = attributes.reject { |_, v| v.nil? } - attributes[:reference_type] ||= self.class.reference_type + # "data-reference-type=" attribute got moved into a constant because we need + # to use it on ReferenceRewriter class to detect if the markdown contains any reference + reference_type_attribute = "#{REFERENCE_TYPE_DATA_ATTRIBUTE}#{escape_once(self.class.reference_type)} " + attributes[:container] ||= 'body' attributes[:placement] ||= 'top' attributes.delete(:original) if context[:no_original_data] + attributes.map do |key, value| %Q(data-#{key.to_s.dasherize}="#{escape_once(value)}") - end.join(' ') + end + .join(' ') + .prepend(reference_type_attribute) end def ignore_ancestor_query diff --git a/lib/banzai/pipeline/markup_pipeline.rb b/lib/banzai/pipeline/markup_pipeline.rb index 17a73f29afb..330914f7238 100644 --- a/lib/banzai/pipeline/markup_pipeline.rb +++ b/lib/banzai/pipeline/markup_pipeline.rb @@ -9,8 +9,8 @@ module Banzai Filter::AssetProxyFilter, Filter::ExternalLinkFilter, Filter::PlantumlFilter, - Filter::SyntaxHighlightFilter, - Filter::KrokiFilter + Filter::KrokiFilter, + Filter::SyntaxHighlightFilter ] end |