prosemirror 0.1.0 → 0.3.0

data/lib/prosemirror/html/selector.rb

@@ -0,0 +1,95 @@
+# frozen_string_literal: true
+
+module Prosemirror
+  module HTML
+    # The CSS selectors parse rules need, matched against one element: lists of compound selectors,
+    # with a type (`p`, `*`), `#id`, `.class`, attribute selectors (`[href]`, `[a=b]`, `[a^=b]`,
+    # `[a$=b]`, `[a*=b]`, `[a~=b]`, `[a|=b]`) and `:not(…)`. Combinators aren't supported.
+    class Selector
+      IDENTIFIER = /-?[_a-zA-Z][-\w]*/
+      ATTRIBUTE = /\[\s*(#{IDENTIFIER})\s*(?:([~|^$*]?=)\s*(?:"([^"]*)"|'([^']*)'|(#{IDENTIFIER}))\s*)?\]/
+
+      def self.parse(source)
+        (@cache ||= {})[source] ||= new(source)
+      end
+
+      def initialize(source)
+        @source = source
+        @compounds = source.split(",").map { |part| parse_compound(part.strip) }
+      end
+
+      def match?(element)
+        @compounds.any? { |tests| tests.all? { |test| test.call(element) } }
+      end
+
+      private
+
+      def parse_compound(part)
+        invalid! if part.empty?
+
+        tests = []
+        rest = part
+        if (type = rest[/\A(?:\*|#{IDENTIFIER})/o])
+          tests << ->(element) { element.name == type.downcase } unless type == "*"
+          rest = rest[type.length..]
+        end
+        until rest.empty?
+          test, length = parse_simple(rest)
+          tests << test
+          rest = rest[length..]
+        end
+        tests
+      end
+
+      # A simple selector at the start of `rest`, as a test and the length it takes.
+      def parse_simple(rest)
+        if (match = /\A#(#{IDENTIFIER})/o.match(rest))
+          [id_test(match[1]), match[0].length]
+        elsif (match = /\A\.(#{IDENTIFIER})/o.match(rest))
+          [class_test(match[1]), match[0].length]
+        elsif (match = /\A#{ATTRIBUTE}/o.match(rest))
+          [attribute_test(*match.captures), match[0].length]
+        elsif (match = /\A:not\(([^()]*)\)/.match(rest))
+          [not_test(parse_compound(match[1].strip)), match[0].length]
+        else
+          invalid!
+        end
+      end
+
+      def id_test(id)
+        ->(element) { element["id"] == id }
+      end
+
+      def class_test(name)
+        ->(element) { element.classes.include?(name) }
+      end
+
+      def attribute_test(name, operator, double_quoted, single_quoted, bare)
+        name = name.downcase
+        expected = double_quoted || single_quoted || bare
+        lambda do |element|
+          value = element[name]
+          next false if value.nil?
+
+          case operator
+          when nil then true
+          when "=" then value == expected
+          when "^=" then !expected.empty? && value.start_with?(expected)
+          when "$=" then !expected.empty? && value.end_with?(expected)
+          when "*=" then !expected.empty? && value.include?(expected)
+          when "~=" then value.split.include?(expected)
+          when "|=" then value == expected || value.start_with?("#{expected}-")
+          end
+        end
+      end
+
+      def not_test(tests)
+        ->(element) { !tests.all? { |test| test.call(element) } }
+      end
+
+      def invalid!
+        raise ArgumentError, "unsupported selector #{@source.inspect}: only compound selectors, without combinators"
+      end
+    end
+  end
+end

data/lib/prosemirror/html/splitter.rb

@@ -0,0 +1,68 @@
+# frozen_string_literal: true
+
+require "strscan"
+
+module Prosemirror
+  module HTML
+    # Splits HTML into pieces of about `size` bytes, between top-level elements, so Herb can parse
+    # them one at a time: its memory grows to hundreds of times the size of the input it parses
+    # (https://github.com/marcoroth/herb/issues/2839). Once Herb fixes it, this class can probably be removed.
+    #
+    # It only cuts where it can follow the structure: when a closing tag doesn't match the last open
+    # element, as with unclosed or implicitly closed elements, the rest stays in one piece. HTML with
+    # ERB tags isn't split, since Herb groups ERB blocks with the HTML between them.
+    class Splitter
+      # As Herb's lists.
+      VOID_TAGS = %w[
+        area base basefont bgsound br col embed frame hr img input keygen link meta param source track wbr
+      ].to_set.freeze
+      RAW_TEXT_TAGS = %w[script style iframe xmp noembed noframes textarea title].to_set.freeze
+
+      NAME = /[a-zA-Z][^\s\/>]*/
+      # The rest of an opening tag: attributes, with quoted values that can hold `>`.
+      TAG_REST = /(?:[^>"']|"[^"]*"|'[^']*')*>/
+
+      def initialize(size)
+        @size = size
+      end
+
+      def call(html)
+        return [html] if html.bytesize <= @size || html.include?("<%")
+
+        pieces = []
+        start = 0
+        open = []
+        scanner = StringScanner.new(html)
+        until scanner.eos?
+          break unless scanner.skip_until(/</)
+
+          if scanner.skip(/!--/)
+            break unless scanner.skip_until(/-->/)
+          elsif scanner.skip(/[!?]/)
+            break unless scanner.skip_until(/>/)
+          elsif scanner.skip(%r{/(#{NAME})\s*>}o)
+            break unless open.pop == scanner[1].downcase
+          elsif scanner.skip(/(#{NAME})/o)
+            name = scanner[1].downcase
+            break unless (rest = scanner.scan(TAG_REST))
+
+            if RAW_TEXT_TAGS.include?(name)
+              break unless scanner.skip_until(%r{</#{Regexp.escape(name)}\s*>}i)
+            elsif !VOID_TAGS.include?(name) && !rest.end_with?("/>")
+              open << name
+            end
+          else
+            next
+          end
+
+          if open.empty? && scanner.pos - start >= @size
+            pieces << html.byteslice(start...scanner.pos)
+            start = scanner.pos
+          end
+        end
+        pieces << html.byteslice(start..) if start < html.bytesize
+        pieces
+      end
+    end
+  end
+end

data/lib/prosemirror/html/text.rb

@@ -0,0 +1,23 @@
+# frozen_string_literal: true
+
+module Prosemirror
+  module HTML
+    # A text node, with its character references decoded.
+    class Text
+      attr_reader :value
+      attr_accessor :parent
+
+      def initialize(value)
+        @value = value
+      end
+
+      def text_content
+        @value
+      end
+
+      def element?
+        false
+      end
+    end
+  end
+end

data/lib/prosemirror/html_parser.rb

@@ -0,0 +1,503 @@
+# frozen_string_literal: true
+
+module Prosemirror
+  # Builds a document from HTML with the parse rules of the schema's node and mark types, as ProseMirror's
+  # DOMParser does, from which this is ported: https://prosemirror.net/docs/ref/#model.DOMParser
+  #
+  # - whitespace is collapsed, except in nodes whose whitespace is preserved, like code blocks;
+  # - content that doesn't fit where it is gets wrapped (text in a list becomes a list item with a
+  #   paragraph), or moves up to a parent that accepts it;
+  # - required content is filled in (an empty list item gets a paragraph);
+  # - elements without a rule are unwrapped, and `head`, `noscript`, `object`, `script`, `style` and `title`
+  #   are ignored.
+  class HTMLParser
+    BLOCK_TAGS = %w[
+      address article aside blockquote canvas dd div dl fieldset figcaption figure footer form h1 h2 h3 h4 h5 h6
+      header hgroup hr li noscript ol output p pre section table tfoot ul
+    ].to_set.freeze
+    IGNORE_TAGS = %w[head noscript object script style title].to_set.freeze
+    LIST_TAGS = %w[ol ul].to_set.freeze
+
+    PRESERVE_WS = 1
+    PRESERVE_WS_FULL = 2
+
+    WHITESPACE = /[ \t\r\n\f]/
+    WHITESPACE_RUN = /[ \t\r\n\f]+/
+
+    # A rule that matched an element: the rule, its index, the attributes it gives, and the schema's type.
+    Match = Data.define(:rule, :index, :attrs, :type)
+
+    attr_reader :schema, :normalize_lists, :matched_styles
+
+    def initialize(schema)
+      @schema = schema
+      rules = (schema.marks + schema.nodes).flat_map(&:parse_rules)
+      rules = rules.each_with_index.sort_by { |rule, index| [-rule.priority, index] }.map(&:first)
+      @tags = rules.select(&:tag)
+      @styles = rules.select(&:style)
+      @matched_styles = @styles.map(&:style_property).uniq
+      @normalize_lists = @tags.none? { |rule| lists_contain_themselves?(rule) }
+    end
+
+    # The document parsed from the HTML, as its top node.
+    def parse(html)
+      parse_element(HTML::Builder.new.call(html))
+    end
+
+    # The document for an Element's content, like the one HTML::Builder gives.
+    def parse_element(root)
+      context = Context.new(self)
+      context.add_all(root, [])
+      context.finish
+    end
+
+    # The first tag rule after `after` that matches the element in this context.
+    def match_tag(element, context, after = nil)
+      ((after ? after + 1 : 0)...@tags.size).each do |index|
+        rule = @tags[index]
+        next unless HTML::Selector.parse(rule.tag).match?(element)
+        next if rule.context && !context.matches_context?(rule.context)
+
+        match = build_match(rule, index, element) { |get_attrs| call_get_attrs(get_attrs, element) }
+        return match if match
+      end
+      nil
+    end
+
+    # The first style rule after `after` that matches the property and value in this context.
+    def match_style(property, value, context, after = nil)
+      ((after ? after + 1 : 0)...@styles.size).each do |index|
+        rule = @styles[index]
+        next unless rule.style_matches?(property, value)
+        next if rule.context && !context.matches_context?(rule.context)
+
+        match = build_match(rule, index, nil) { |get_attrs| get_attrs.call(value) }
+        return match if match
+      end
+      nil
+    end
+
+    # The schema's type for a rule: a subclass registered in the schema inherits its parent's rules.
+    def type_for(rule)
+      rule.node? ? @schema.node_type(rule.type.type_name) : @schema.mark_type(rule.type.type_name)
+    end
+
+    private
+
+    def build_match(rule, index, element)
+      type = type_for(rule)
+      return nil unless type
+
+      given = rule.attrs
+      if rule.get_attrs
+        result = yield(rule.get_attrs)
+        return nil if result == false
+
+        given = result.is_a?(Hash) ? result : nil
+      end
+      return Match.new(rule:, index:, attrs: nil, type:) if rule.ignore || rule.skip || rule.close_parent
+
+      attrs = build_attrs(type, given, element)
+      attrs && Match.new(rule:, index:, attrs:, type:)
+    end
+
+    # The rule's attributes, with the ones read from the element, built with the type's properties.
+    def build_attrs(type, given, element)
+      given = (given || {}).transform_keys(&:to_s)
+      given = given.merge(type.attrs_from_html(element)) if element
+      type.build_attrs(given, report: ->(message) { @schema.report(message, path: "html") })
+    end
+
+    def call_get_attrs(get_attrs, element)
+      (get_attrs.arity == 2) ? get_attrs.call(element, @schema) : get_attrs.call(element)
+    end
+
+    def lists_contain_themselves?(rule)
+      return false unless rule.tag.match?(/\A(ul|ol)\b/) && rule.node?
+
+      type = type_for(rule)
+      type && @schema.content_match(type).match_type(type)
+    end
+
+    # A node being built: its type, attributes and marks, its content so far, and its content match state.
+    class NodeContext
+      attr_reader :type, :attrs, :marks, :solid, :content
+      attr_accessor :match, :options
+
+      def initialize(type, attrs, marks, solid, options, schema)
+        @type = type
+        @attrs = attrs
+        @marks = marks
+        @solid = solid
+        @options = options
+        @schema = schema
+        @match = schema.content_match(type)
+        @content = []
+      end
+
+      def find_wrapping(type)
+        @match&.find_wrapping(type, @schema)
+      end
+
+      def finish
+        strip_trailing_whitespace unless @options.anybits?(PRESERVE_WS)
+        fill = @match&.fill_before([], @schema, to_end: true) || []
+        content = Node.join_text(@content + fill.filter_map { |type| @schema.create_and_fill(type) })
+        Node.new(type: @type.type_name, attrs: @attrs, content:, marks: @marks)
+      end
+
+      def inline_context?
+        @schema.content_match(@type).inline_content?
+      end
+
+      private
+
+      def strip_trailing_whitespace
+        last = @content.last
+        return unless last&.text? && (trailing = last.text[/#{WHITESPACE_RUN}\z/o])
+
+        if trailing.length == last.text.length
+          @content.pop
+        else
+          @content[-1] = last.with(text: last.text[0...-trailing.length])
+        end
+      end
+    end
+
+    # The state of one parse, as ProseMirror's ParseContext.
+    class Context
+      def initialize(parser)
+        @parser = parser
+        @schema = parser.schema
+        @nodes = [NodeContext.new(@schema.node_type(@schema.top_node), @schema.node_type(@schema.top_node).default_attrs,
+          [], true, 0, @schema)]
+        @open = 0
+        @needs_block = false
+        @local_preserve_ws = false
+        @linebreak = @schema.nodes.find(&:linebreak_replacement?)
+      end
+
+      def top
+        @nodes[@open]
+      end
+
+      def add_all(parent, marks)
+        parent.children.each { |child| add_dom(child, marks) }
+      end
+
+      def finish
+        @open = 0
+        close_extra
+        @nodes.first.finish
+      end
+
+      # Whether the parents match a rule's context, as `"list_item/"`, `"blockquote//"` or `"a/|b/"`.
+      def matches_context?(context)
+        return context.split(/\s*\|\s*/).any? { |option| matches_context?(option) } if context.include?("|")
+
+        parts = context.split("/", -1)
+        context_matches?(parts, parts.size - 1, @open)
+      end
+
+      private
+
+      def context_matches?(parts, index, depth)
+        while index >= 0
+          part = parts[index]
+          if part.empty?
+            if index == parts.size - 1 || index.zero?
+              index -= 1
+              next
+            end
+            return depth.downto(0).any? { |deeper| context_matches?(parts, index - 1, deeper) }
+          end
+
+          type = (depth >= 0) ? @nodes[depth].type : nil
+          return false if type.nil? || (type.type_name.to_s != part && !type.group.include?(part.to_sym))
+
+          depth -= 1
+          index -= 1
+        end
+        true
+      end
+
+      def add_dom(dom, marks)
+        dom.element? ? add_element(dom, marks) : add_text_node(dom, marks)
+      end
+
+      def add_text_node(dom, marks)
+        value = dom.value
+        top = self.top
+        preserve = if top.options.anybits?(PRESERVE_WS_FULL)
+          :full
+        else
+          @local_preserve_ws || top.options.anybits?(PRESERVE_WS)
+        end
+        return unless preserve == :full || top.inline_context? || value.match?(/[^ \t\r\n\f]/)
+
+        if !preserve
+          value = value.gsub(WHITESPACE_RUN, " ")
+          value = value[1..] if value.start_with?(" ") && @open == @nodes.size - 1 && strip_leading_space?(dom, top)
+        elsif preserve == :full
+          value = value.gsub(/\r\n?/, "\n")
+        elsif @linebreak && value.match?(/[\r\n]/) && top.find_wrapping(@linebreak)
+          value.split(/\r?\n|\r/, -1).each_with_index do |line, index|
+            insert_node(Node.new(type: @linebreak.type_name, attrs: @linebreak.default_attrs), marks, true) if index.positive?
+            insert_node(text_node(line), marks, !line.match?(/\S/)) unless line.empty?
+          end
+          value = ""
+        else
+          value = value.gsub(/\r?\n|\r/, " ")
+        end
+        insert_node(text_node(value), marks, !value.match?(/\S/)) unless value.empty?
+      end
+
+      # A leading space is dropped at the start of a node, after a line break, and after a space.
+      def strip_leading_space?(dom, top)
+        before = top.content.last
+        previous = dom.parent&.previous_sibling(dom)
+        before.nil? || (previous&.element? && previous.name == "br") || (before.text? && before.text.match?(/#{WHITESPACE}\z/o))
+      end
+
+      def add_element(dom, marks, match_after = nil)
+        outer_ws = @local_preserve_ws
+        top = self.top
+        @local_preserve_ws = true if dom.name == "pre" || dom.style["white-space"].to_s.include?("pre")
+        normalize_list(dom) if LIST_TAGS.include?(dom.name) && @parser.normalize_lists
+        match = @parser.match_tag(dom, self, match_after)
+        rule = match&.rule
+
+        if rule ? rule.ignore : IGNORE_TAGS.include?(dom.name)
+          ignore_fallback(dom, marks)
+        elsif !rule || rule.skip || rule.close_parent
+          add_element_without_rule(dom, marks, rule, top)
+        else
+          inner_marks = read_styles(dom, marks)
+          add_element_by_rule(dom, match, inner_marks, rule.consuming ? nil : match.index) if inner_marks
+        end
+        @local_preserve_ws = outer_ws
+      end
+
+      def add_element_without_rule(dom, marks, rule, top)
+        @open = [0, @open - 1].max if rule&.close_parent
+        sync = false
+        old_needs_block = @needs_block
+        if BLOCK_TAGS.include?(dom.name)
+          if top.content.any? && inline?(top.content.first) && @open.positive?
+            @open -= 1
+            top = self.top
+          end
+          sync = true
+          @needs_block = true unless top.type
+        elsif dom.children.empty?
+          return leaf_fallback(dom, marks)
+        end
+        inner_marks = rule&.skip ? marks : read_styles(dom, marks)
+        add_all(dom, inner_marks) if inner_marks
+        self.sync(top) if sync
+        @needs_block = old_needs_block
+      end
+
+      # A `br` without a rule becomes a newline, in nodes with inline content.
+      def leaf_fallback(dom, marks)
+        add_text_node(HTML::Text.new("\n"), marks) if dom.name == "br" && top.inline_context?
+      end
+
+      # An ignored `br` still starts inline content.
+      def ignore_fallback(dom, marks)
+        find_place(@schema.node_type(:text), marks, true) if dom.name == "br" && !top.inline_context?
+      end
+
+      # The marks with the ones from the element's styles, or nil when a style rule ignores the element.
+      def read_styles(dom, marks)
+        return marks if dom.style.empty?
+
+        @parser.matched_styles.each do |property|
+          value = dom.style[property]
+          next if value.nil? || value.empty?
+
+          after = nil
+          while (match = @parser.match_style(property, value, self, after))
+            return nil if match.rule.ignore
+
+            marks = if match.rule.clear_mark
+              marks.reject { |mark| clears?(match.rule.clear_mark, match.type, mark) }
+            else
+              marks + [Mark.new(type: match.type.type_name, attrs: match.attrs)]
+            end
+            break if match.rule.consuming
+
+            after = match.index
+          end
+        end
+        marks
+      end
+
+      def clears?(clear_mark, type, mark)
+        (clear_mark == true) ? mark.type == type.type_name : clear_mark.call(mark)
+      end
+
+      def add_element_by_rule(dom, match, marks, continue_after)
+        rule = match.rule
+        type = match.type
+        sync = false
+        if rule.node?
+          if !type.leaf?
+            inner = enter(type, match.attrs, marks, rule.preserve_whitespace)
+            if inner
+              sync = true
+              marks = inner
+            end
+          elsif !insert_node(Node.new(type: type.type_name, attrs: match.attrs), marks, dom.name == "br")
+            leaf_fallback(dom, marks)
+          end
+        else
+          marks += [Mark.new(type: type.type_name, attrs: match.attrs)]
+        end
+        start_in = top
+
+        if rule.node? && type.leaf?
+          # Leaves have no content to parse.
+        elsif continue_after
+          add_element(dom, marks, continue_after)
+        elsif rule.get_content
+          Array(rule.get_content.call(dom, @schema)).each { |node| insert_node(node, marks, false) }
+        else
+          add_all(content_element(dom, rule) || dom, marks)
+        end
+        @open -= 1 if sync && self.sync(start_in)
+      end
+
+      def content_element(dom, rule)
+        case rule.content_element
+        when String then dom.query(rule.content_element)
+        when Proc then rule.content_element.call(dom)
+        end
+      end
+
+      # Finds where a node of this type fits: in the open nodes, from the innermost, possibly with wrappers.
+      # Returns the marks left to apply, or nil when it fits nowhere.
+      def find_place(type, marks, cautious)
+        route = nil
+        sync_to = nil
+        penalty = 0
+        @open.downto(0) do |depth|
+          context = @nodes[depth]
+          found = context.find_wrapping(type)
+          if found && (route.nil? || route.size > found.size + penalty)
+            route = found
+            sync_to = context
+            break if found.empty?
+          end
+          next unless context.solid
+          break if cautious
+
+          penalty += 2
+        end
+        return nil unless route
+
+        sync(sync_to)
+        route.each { |wrapper| marks = enter_inner(wrapper, wrapper.default_attrs, marks, false) }
+        marks
+      end
+
+      def insert_node(node, marks, cautious)
+        type = node_type(node)
+        inner_marks = find_place(type, marks, cautious)
+        return false unless inner_marks
+
+        close_extra
+        top = self.top
+        top.match &&= top.match.match_type(type)
+        node_marks = (inner_marks + node.marks).each_with_object([]) do |mark, set|
+          mark_type = @schema.mark_type(mark.type)
+          set.replace(@schema.add_mark(mark, set)) if mark_type && top.type.allows_mark?(mark_type, @schema)
+        end
+        top.content << node.with(marks: node_marks)
+        true
+      end
+
+      def enter(type, attrs, marks, preserve_ws)
+        inner_marks = find_place(type, marks, false)
+        inner_marks && enter_inner(type, attrs, inner_marks, true, preserve_ws)
+      end
+
+      # Opens a node of this type. Returns the marks that don't apply to it, which apply to its content.
+      def enter_inner(type, attrs, marks, solid, preserve_ws = nil)
+        close_extra
+        top = self.top
+        top.match &&= top.match.match_type(type)
+        options = ws_options(type, preserve_ws, top.options)
+        applied = []
+        remaining = marks.reject do |mark|
+          mark_type = @schema.mark_type(mark.type)
+          next false unless mark_type && top.type.allows_mark?(mark_type, @schema)
+
+          applied = @schema.add_mark(mark, applied)
+          true
+        end
+        @nodes << NodeContext.new(type, attrs, applied, solid, options, @schema)
+        @open += 1
+        remaining
+      end
+
+      def ws_options(type, preserve_ws, base)
+        unless preserve_ws.nil?
+          return (preserve_ws ? PRESERVE_WS : 0) | ((preserve_ws == :full) ? PRESERVE_WS_FULL : 0)
+        end
+
+        (type.whitespace == :pre) ? (PRESERVE_WS | PRESERVE_WS_FULL) : base
+      end
+
+      # Finishes the nodes above the open one, and adds them to their parents.
+      def close_extra
+        (@nodes.size - 1).downto(@open + 1) { |index| @nodes[index - 1].content << @nodes[index].finish }
+        @nodes.slice!((@open + 1)..)
+      end
+
+      def sync(to)
+        @open.downto(0) do |index|
+          if @nodes[index].equal?(to)
+            @open = index
+            return true
+          elsif @local_preserve_ws
+            @nodes[index].options |= PRESERVE_WS
+          end
+        end
+        false
+      end
+
+      # Moves lists that directly follow a list item into it, as some tools produce them.
+      def normalize_list(dom)
+        previous_item = nil
+        kept = []
+        dom.children.each do |child|
+          name = child.element? ? child.name : nil
+          if name && LIST_TAGS.include?(name) && previous_item
+            previous_item.append(child)
+          else
+            kept << child
+            if name == "li"
+              previous_item = child
+            elsif name
+              previous_item = nil
+            end
+          end
+        end
+        dom.children.replace(kept)
+      end
+
+      def text_node(value)
+        Node.new(type: :text, text: value)
+      end
+
+      def node_type(node)
+        @schema.node_type(node.type)
+      end
+
+      def inline?(node)
+        node.text? || node_type(node)&.inline?
+      end
+    end
+  end
+end