Improve MFM parser (#3337)

* wip

* wip

* Refactor

* Refactor

* wip

* wip

* wip

* wip

* Refactor

* Refactor

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* wip

* Clean up

* Update misskey-flavored-markdown.ts

* wip

* wip

* wip

* wip

* Update parser.ts

* wip

* Add new test

* wip

* Add new test

* Add new test

* wip

* Refactor

* Update parse.ts

* Refactor

* Update parser.ts

* wip

1 files changed, 81 insertions, 0 deletions

diff --git a/src/mfm/parse.ts b/src/mfm/parse.ts
new file mode 100644
index 0000000000..e7ae3d012e
--- /dev/null
+++ b/src/mfm/parse.ts
@@ -0,0 +1,81 @@
+import parser, { Node } from './parser';
+import * as A from '../prelude/array';
+import * as S from '../prelude/string';
+
+export default (source: string): Node[] => {
+	if (source == null || source == '') {
+		return null;
+	}
+
+	let nodes: Node[] = parser.root.tryParse(source);
+
+	const combineText = (es: Node[]): Node =>
+		({ name: 'text', props: { text: S.concat(es.map(e => e.props.text)) } });
+
+	const concatText = (nodes: Node[]): Node[] =>
+		A.concat(A.groupOn(x => x.name, nodes).map(es =>
+			es[0].name === 'text' ? [combineText(es)] : es
+		));
+
+	const concatTextRecursive = (es: Node[]): void =>
+		es.filter(x => x.children).forEach(x => {
+			x.children = concatText(x.children);
+			concatTextRecursive(x.children);
+		});
+
+	nodes = concatText(nodes);
+	concatTextRecursive(nodes);
+
+	function getBeforeTextNode(node: Node): Node {
+		if (node == null) return null;
+		if (node.name == 'text') return node;
+		if (node.children) return getBeforeTextNode(node.children[node.children.length - 1]);
+		return null;
+	}
+
+	function getAfterTextNode(node: Node): Node {
+		if (node == null) return null;
+		if (node.name == 'text') return node;
+		if (node.children) return getBeforeTextNode(node.children[0]);
+		return null;
+	}
+
+	function isBlockNode(node: Node): boolean {
+		return ['blockCode', 'quote', 'title'].includes(node.name);
+	}
+
+	/**
+	 * ブロック要素の前後にある改行を削除します(ブロック要素自体が改行の役割も果たすため、余計に改行されてしまうため)
+	 * @param nodes
+	 */
+	const removeNeedlessLineBreaks = (nodes: Node[]) => {
+		nodes.forEach((node, i) => {
+			if (node.children) removeNeedlessLineBreaks(node.children);
+			if (isBlockNode(node)) {
+				const before = getBeforeTextNode(nodes[i - 1]);
+				const after = getAfterTextNode(nodes[i + 1]);
+				if (before && before.props.text.endsWith('\n')) {
+					before.props.text = before.props.text.substring(0, before.props.text.length - 1);
+				}
+				if (after && after.props.text.startsWith('\n')) {
+					after.props.text = after.props.text.substring(1);
+				}
+			}
+		});
+	};
+
+	const removeEmptyTextNodes = (nodes: Node[]) => {
+		nodes.forEach(n => {
+			if (n.children) {
+				n.children = removeEmptyTextNodes(n.children);
+			}
+		});
+		return nodes.filter(n => !(n.name == 'text' && n.props.text == ''));
+	};
+
+	removeNeedlessLineBreaks(nodes);
+
+	nodes = removeEmptyTextNodes(nodes);
+
+	return nodes;
+};