Merge pull request #13 from blex41/parse-comments

feat: add support for comment parsing

README.md

@@ -79,7 +79,8 @@ Parse given data, and return root of the generated DOM.
     lowerCaseTagName: false,  // convert tag name to lower case (hurt performance heavily)
     script: false,            // retrieve content in <script> (hurt performance slightly)
     style: false,             // retrieve content in <style> (hurt performance slightly)
-    pre: false                // retrieve content in <pre> (hurt performance slightly)
+    pre: false,               // retrieve content in <pre> (hurt performance slightly)
+    comment: false            // retrieve comments (hurt performance slightly)
   }
   ```
 

src/index.ts

@@ -2,7 +2,8 @@ import { decode } from 'he';
 
 export enum NodeType {
 	ELEMENT_NODE = 1,
-	TEXT_NODE = 3
+	TEXT_NODE = 3,
+	COMMENT_NODE = 8
 }
 
 /**
@@ -52,6 +53,31 @@ export class TextNode extends Node {
 	}
 }
 
+export class CommentNode extends Node {
+	constructor(value: string) {
+		super();
+		this.rawText = value;
+	}
+
+	/**
+	 * Node Type declaration.
+	 * @type {Number}
+	 */
+	nodeType = NodeType.COMMENT_NODE;
+
+	/**
+	 * Get unescaped text value of current node and its children.
+	 * @return {string} text content
+	 */
+	get text() {
+		return decode(this.rawText);
+	}
+
+	toString() {
+		return `<!--${this.rawText}-->`;
+	}
+}
+
 const kBlockElements = {
 	div: true,
 	p: true,
@@ -193,7 +219,7 @@ export class HTMLElement extends Node {
 					currentBlock.push(text);
 				}
 			}
-		}
+		} 
 		dfs(this);
 		return blocks
 			.map(function (block) {
@@ -754,6 +780,7 @@ export function parse(data: string, options?: {
 	script?: boolean;
 	style?: boolean;
 	pre?: boolean;
+	comment?: boolean;
 }) {
 	const root = new HTMLElement(null, {});
 	let currentParent = root;
@@ -772,6 +799,11 @@ export function parse(data: string, options?: {
 		lastTextPos = kMarkupPattern.lastIndex;
 		if (match[0][1] == '!') {
 			// this is a comment
+			if (options.comment) {
+				// Only keep what is in between <!-- and -->
+				const text = data.substring(lastTextPos - 3 , lastTextPos - match[0].length + 4);
+				currentParent.appendChild(new CommentNode(text));
+			}
 			continue;
 		}
 		if (options.lowerCaseTagName)

test/html.js

@@ -9,6 +9,7 @@ describe('HTML Parser', function () {
 	var Matcher = HTMLParser.Matcher;
 	var HTMLElement = HTMLParser.HTMLElement;
 	var TextNode = HTMLParser.TextNode;
+	var CommentNode = HTMLParser.CommentNode;
 
 	describe('Matcher', function () {
 		it('should match corrent elements', function () {
@@ -97,6 +98,34 @@ describe('HTML Parser', function () {
 
 		});
 
+		it('should parse "<div><a><!-- my comment --></a></div>" and return root element without comments', function () {
+			var root = parseHTML('<div><a><!-- my comment --></a></div>');
+
+			var div = new HTMLElement('div', {}, '');
+			var a = div.appendChild(new HTMLElement('a', {}, ''));
+
+			root.firstChild.should.eql(div);
+		});
+
+		it('should parse "<div><a><!-- my comment --></a></div>" and return root element with comments', function () {
+			var root = parseHTML('<div><a><!-- my comment --></a></div>', { comment: true });
+
+			var div = new HTMLElement('div', {}, '');
+			var a = div.appendChild(new HTMLElement('a', {}, ''));
+			var comment = a.appendChild(new CommentNode(' my comment '));
+
+			root.firstChild.should.eql(div);
+		});
+
+		it('should not parse HTML inside comments', function () {
+			var root = parseHTML('<div><!--<a></a>--></div>', { comment: true });
+
+			var div = new HTMLElement('div', {}, '');
+			var comment = div.appendChild(new CommentNode('<a></a>'));
+
+			root.firstChild.should.eql(div);
+		});
+
 		it('should parse picture element', function () {
 
 			var root = parseHTML('<picture><source srcset="/images/example-1.jpg 1200w, /images/example-2.jpg 1600w" sizes="100vw"><img src="/images/example.jpg" alt="Example"/></picture>');
@@ -319,6 +348,11 @@ describe('HTML Parser', function () {
 				var root = parseHTML('<span>o<p>a</p><p>b</p>c</span>');
 				root.structuredText.should.eql('o\na\nb\nc');
 			});
+
+			it('should not return comments in structured text', function () {
+				var root = parseHTML('<span>o<p>a</p><!-- my comment --></span>', { comment: true });
+				root.structuredText.should.eql('o\na');
+			});
 		});
 		describe('#set_content', function () {
 			it('set content string', function () {
@@ -350,6 +384,26 @@ describe('HTML Parser', function () {
 			const root = parseHTML(html);
 			root.toString().should.eql(html)
 		});
+
+		it('#toString() should not return comments by default', function () {
+			const html = '<p><!-- my comment --></p>';
+			const result = '<p></p>';
+			const root = parseHTML(html);
+			root.toString().should.eql(result);
+		});
+
+		it('#toString() should return comments when specified', function () {
+			const html = '<!----><p><!-- my comment --></p>';
+			const root = parseHTML(html, { comment: true });
+			root.toString().should.eql(html);
+		});
+	});
+
+	describe('Comment Element', function () {
+		it('comment nodeType should be 8', function () {
+			var root = parseHTML('<!-- my comment -->', { comment: true });
+			root.firstChild.nodeType.should.eql(8);
+		});
 	});
 
 	describe('Custom Element', function () {