feat: Support multiline quote without ">>>" <https://github.com/pocka/slack-message-parser/issues/38> close #38 The parser previously parse these messages as multiple single-line quotes instead of one quote with children. ``` foo &gt; bar &gt; baz ``` While ">>>" is still valid mrkdwn syntax, it's not documented at Slack's official mrkdwn syntax page. Therefore, we should consider this style of multiline quote the most appropriate way and add best-in-class support. My implementation introduced a new `oneOrMore` parser combinator. In order to safely implement this, I added optional Node type constraint (type parameter) to `Parser`.

Changes

4 changed files (+112/-38)

• combinator.ts
  • @@ -17,15 +17,47 @@ return null;
    }; }; - export const regexp = ( + export const oneOrMore = <A extends Node, B extends Node>( + parser: Parser<A>, + andThen: (nodes: readonly A[]) => B, + ): Parser<B> => { + const rec = ( + text: string, + position: number, + rootParser: ParseText, + ): readonly NonNullable<ReturnType<typeof parser>>[] => { + const match = parser(text, position, rootParser); + if (!match) { + return []; + } + + const [, nextPosition] = match; + + return [match, ...rec(text, nextPosition, rootParser)]; + }; + + return (text, position, rootParser) => { + const ret = rec(text, position, rootParser); + + if (ret.length === 0) { + return null; + } + + const [, lastPosition] = ret[ret.length - 1]; + + return [andThen(ret.map(([a]) => a)), lastPosition]; + }; + }; + + export const regexp = <T extends Node = Node>( pattern: RegExp, callback: ( match: string[], text: string, position: number, parseText: ParseText, - ) => [Node, number] | null, - ): Parser => + ) => [T, number] | null, + ): Parser<T> => (text, position, parseText) => { const match = text.substring(position).match(pattern);
  • @@ -51,15 +83,16 @@
    return parser(text, position, parseText); }; - export const topOfLine = (parser: Parser): Parser => - ( - text, - position, - parseText, - ) => { - if (position > 0 && text.charAt(position - 1) !== "\n") { - return null; - } + export const topOfLine = + <T extends Node = Node>(parser: Parser<T>): Parser<T> => + ( + text, + position, + parseText, + ) => { + if (position > 0 && text.charAt(position - 1) !== "\n") { + return null; + } - return parser(text, position, parseText); - }; + return parser(text, position, parseText); + };
• parser.ts
  • @@ -1,6 +1,6 @@
    - import { NodeType } from "./types/Node.ts"; + import { type Node, NodeType, type Quote } from "./types/Node.ts"; - import { explicit, or, regexp, topOfLine } from "./combinator.ts"; + import { explicit, oneOrMore, or, regexp, topOfLine } from "./combinator.ts"; const parseBold = explicit( regexp(
  • @@ -89,12 +89,20 @@ },
    ), ); - const parseSingleLineQuote = topOfLine( - regexp(/^&gt;(.*)(\n|$)/, (match, _text, position, parseText) => { + const parseQuoteLine = topOfLine( + regexp(/^&gt;(.*)(\n&gt;|\n|$)/, (match, _text, position, parseText) => { const [matchedText, content] = match; const repeatedGt = content.match(/^((&gt;)+)(.*)$/); + // If the next line is also starts with ">", do not include the character + // (simulating RegExp's non-capturing group) + const source = matchedText.replace(/\n&gt;$/, "\n"); + + // `source` and `matchedText` are same unless the next line starts with ">" + // due to the above line removes the ">". + const isNextLineStartsWithGt = source !== matchedText; + return [ { type: NodeType.Quote,
  • @@ -107,28 +115,39 @@ source: repeatedGt[1],
    }, ...parseText(repeatedGt[3]), ] - : parseText(content), - source: matchedText, + // Only the last LF could be a terminator character of quote. + // Non-last LFs should be parsed as well as `content`. + : parseText(isNextLineStartsWithGt ? content + "\n" : content), + source, }, - position + matchedText.length, + position + source.length, ]; }), ); - const parseMultilineQuote = topOfLine( - regexp(/^&gt;&gt;&gt;([\s\S]+)$/, (match, _text, position, parseText) => { - const [matchedText, content] = match; + const parseQuote = or([ + topOfLine( + regexp(/^&gt;&gt;&gt;([\s\S]+)$/, (match, _text, position, parseText) => { + const [matchedText, content] = match; - return [ - { - type: NodeType.Quote, - children: parseText(content), - source: matchedText, - }, - position + matchedText.length, - ]; + return [ + { + type: NodeType.Quote, + children: parseText(content), + source: matchedText, + }, + position + matchedText.length, + ]; + }), + ), + oneOrMore<Quote, Quote>(parseQuoteLine, (quotes) => { + return { + type: NodeType.Quote, + children: quotes.map((quote) => quote.children).flat(), + source: quotes.map((quote) => quote.source).join(""), + }; }), - ); + ]); const parseEmoji = regexp( /^:([^:<`*#@!\s()$%]+):(:(skin-tone-.+?):)?/,
  • @@ -147,7 +166,7 @@ ];
    }, ); - const parseLink = regexp( + const parseLink = regexp<Node>( /^<([^\s<>][^\n<>]*?)(\|([^<>]+?))?>/, (match, _text, position, parseText) => { const [matchedText, link, _, label] = match;
  • @@ -209,8 +228,7 @@ parsePreText,
    parseCode, parseEmoji, parseItalic, - parseMultilineQuote, - parseSingleLineQuote, + parseQuote, parseLink, parseStrike, ]);
• tests/issues.test.ts
  • @@ -7,6 +7,7 @@ bold,
    code, emoji, italic, + quote, root, strike, text,
  • @@ -129,3 +130,25 @@ bold([text("Y")]),
    ]), ); }); + + // https://github.com/pocka/slack-message-parser/issues/38 + Deno.test(`#38 / parse multiline quote without >>>`, () => { + const actual = parse( + "This is unquoted text\n&gt; This is quoted text\n&gt; This is still quoted text\nThis is unquoted text again", + ); + + assertEquals( + actual, + root([ + text("This is unquoted text\n"), + { + ...quote([ + text(" This is quoted text\n"), + text(" This is still quoted text"), + ], true), + source: "&gt; This is quoted text\n&gt; This is still quoted text\n", + }, + text("This is unquoted text again"), + ]), + ); + });
• types/Parser.ts
  • @@ -1,9 +1,9 @@
    import { Node } from "./Node.ts"; - export type Parser = ( + export type Parser<T extends Node = Node> = ( text: string, position: number, parseText: ParseText, - ) => [Node, number] | null; + ) => [T, number] | null; export type ParseText = (text: string) => Node[];