Add an HTML parser and URL extractor

This will be used for adding discovered URLs to
the crawl queue.

Author: john-shaffer

composer.json

@@ -29,7 +29,8 @@
         "ext-mbstring": "*",
         "ext-simplexml": "*",
         "lib-libxml": "*",
-        "yahnis-elsts/plugin-update-checker": "^5.6"
+        "yahnis-elsts/plugin-update-checker": "^5.6",
+        "masterminds/html5": "^2.9"
     },
     "require-dev": {
         "thecodingmachine/phpstan-strict-rules": "*",

composer.lock

@@ -22,7 +22,7 @@
               let rel = baseNameOf path;
               in rel == "composer.json" || rel == "composer.lock";
           };
-          vendorHash = "sha256-lN/24P9LbuRuU/bc3DNPR0kZGtIhZfWcNqeL5puXj2w=";
+          vendorHash = "sha256-4DTiW/rykJGZAnWCRSqlAaBtDthXKkZshPNe3yYCfA4=";
         });
         wp2static = runCommand "wp2static" {} ''
           export PLUGIN_DIR="$TMPDIR/${name}"

flake.nix

@@ -0,0 +1,45 @@
+<?php
+
+namespace WP2Static;
+
+class ParseHTML {
+    /**
+     * Return URLs found in a DOMNode, recursively.
+     *
+     * @param \DOMNode $node
+     * @return \Iterator<string>
+     */
+    public static function parseURLsDOMNode( \DOMNode $node ) : \Iterator {
+        foreach ( $node->childNodes as $child ) {
+            if ( $child instanceof \DOMElement ) {
+                $tag_name = strtolower( $child->tagName );
+                switch ( $tag_name ) {
+                    case 'a':
+                    case 'link':
+                        yield $child->getAttribute( 'href' );
+                        break;
+                    case 'img':
+                    case 'script':
+                    case 'source':
+                        yield $child->getAttribute( 'src' );
+                        break;
+                }
+                foreach ( self::parseURLsDOMNode( $child ) as $url ) {
+                    yield $url;
+                }
+            }
+        }
+    }
+
+    /**
+     * Return an iterator of URLs parsed from the provided HTML
+     *
+     * @param string $html
+     * @return \Iterator<string>
+     */
+    public static function parseURLsString( string $html ) : \Iterator {
+        $html5 = new \Masterminds\HTML5();
+        $dom = $html5->loadHTML( $html );
+        return self::parseURLsDOMNode( $dom );
+    }
+}