` in the head (AgentReady
* AR-READ-09) and the matching Link header. The href is the page's own URL —
* the same URL serves markdown on `Accept: text/markdown`.
*
* @return void
*/
function output_markdown_alternate_link(): void {
if ( ! is_feature_enabled( 'markdown' ) || is_admin() || is_feed() || is_404() ) {
return;
}
$url = markdown_alternate_url();
if ( '' === $url ) {
return;
}
echo '' . "\n";
}
/**
* URL of the current view for the alternate link ('' when unknown).
*
* @return string
*/
function markdown_alternate_url(): string {
if ( is_front_page() ) {
return home_url( '/' );
}
if ( is_singular() ) {
$permalink = get_permalink();
return is_string( $permalink ) ? $permalink : '';
}
global $wp;
if ( $wp instanceof \WP && is_string( $wp->request ) ) {
return home_url( user_trailingslashit( $wp->request ) );
}
return '';
}
/**
* YAML frontmatter block for the markdown representation: title, URL, site,
* modification date, and the HTTP status for error pages. Lets an agent
* identify what it fetched without parsing the body.
*
* @return string
*/
function markdown_frontmatter(): string {
$yaml = static function ( string $value ): string {
return '"' . str_replace( array( '\\', '"' ), array( '\\\\', '\\"' ), markdown_safe_text( $value ) ) . '"';
};
$lines = array( '---' );
$lines[] = 'title: ' . $yaml( is_404() ? __( 'Not found', 'aj-agent-crawl-optimizer' ) : wp_get_document_title() );
$url = markdown_alternate_url();
if ( '' !== $url ) {
$lines[] = 'url: ' . esc_url_raw( $url );
}
$lines[] = 'site: ' . $yaml( get_bloginfo( 'name' ) );
if ( is_singular() ) {
$modified = get_post_modified_time( 'c', true );
if ( is_string( $modified ) ) {
$lines[] = 'modified: ' . $modified;
}
}
if ( is_404() ) {
$lines[] = 'status: 404';
}
$lines[] = '---';
return implode( "\n", $lines ) . "\n\n";
}
/**
* Check if the current request accepts Markdown.
*
* @return bool
*/
function accepts_markdown(): bool {
if ( ! isset( $_SERVER['HTTP_ACCEPT'] ) ) {
return false;
}
$accept = sanitize_text_field( wp_unslash( $_SERVER['HTTP_ACCEPT'] ) );
// strpos, not str_contains — the WP polyfill only exists since WP 5.9 and
// this plugin supports 5.5+.
return strpos( $accept, 'text/markdown' ) !== false;
}
/**
* Emit `Vary: Accept` on every frontend response while markdown negotiation
* is enabled — the representation varies on the Accept header, so caches and
* CDNs must key on it. Without this, a page cache either serves cached HTML
* to agents (breaking negotiation) or caches the markdown variant for humans.
*
* @return void
*/
function send_markdown_vary_header(): void {
if ( ! is_feature_enabled( 'markdown' ) ) {
return;
}
if ( is_admin() ) {
return;
}
header( 'Vary: Accept', false );
if ( ! is_feed() && ! is_404() ) {
$url = markdown_alternate_url();
if ( '' !== $url ) {
header( 'Link: <' . esc_url_raw( $url ) . '>; rel="alternate"; type="text/markdown"', false );
}
}
}
/**
* Convert HTML to Markdown.
*
* @param string $html HTML string.
* @return string
*/
function html_to_markdown( string $html ): string {
$markdown = $html;
// Remove script, style, noscript, and comments.
$markdown = preg_replace( '/]*>.*?<\/script>/is', '', $markdown );
$markdown = preg_replace( '/]*>.*?<\/style>/is', '', $markdown );
$markdown = preg_replace( '/]*>.*?<\/noscript>/is', '', $markdown );
$markdown = preg_replace( '//s', '', $markdown );
$markdown = preg_replace( '/<[^>]*hidden[^>]*>/i', '', $markdown );
// Headings.
$markdown = preg_replace( '/]*>(.*?)<\/h1>/is', "# $1\n\n", $markdown );
$markdown = preg_replace( '/]*>(.*?)<\/h2>/is', "## $1\n\n", $markdown );
$markdown = preg_replace( '/]*>(.*?)<\/h3>/is', "### $1\n\n", $markdown );
$markdown = preg_replace( '/]*>(.*?)<\/h4>/is', "#### $1\n\n", $markdown );
$markdown = preg_replace( '/]*>(.*?)<\/h5>/is', "##### $1\n\n", $markdown );
$markdown = preg_replace( '/]*>(.*?)<\/h6>/is', "###### $1\n\n", $markdown );
// Paragraphs and line breaks.
$markdown = preg_replace( '//i', "\n", $markdown );
$markdown = preg_replace( '/<\/p>/i', "\n\n", $markdown );
$markdown = preg_replace( '/]*>/i', '', $markdown );
// Bold and italic.
$markdown = preg_replace( '/]*>(.*?)<\/strong>/is', '**$1**', $markdown );
$markdown = preg_replace( '/]*>(.*?)<\/b>/is', '**$1**', $markdown );
$markdown = preg_replace( '/]*>(.*?)<\/em>/is', '*$1*', $markdown );
$markdown = preg_replace( '/]*>(.*?)<\/i>/is', '*$1*', $markdown );
// Links.
$markdown = preg_replace( '/]*href=["\']([^"\']+)["\'][^>]*>(.*?)<\/a>/is', '[$2]($1)', $markdown );
// Images.
$markdown = preg_replace( '/]*src=["\']([^"\']+)["\'][^>]*(?:\/?)>/is', '', $markdown );
// Lists.
$markdown = preg_replace( '/]*>/i', "\n", $markdown );
$markdown = preg_replace( '/<\/ul>/i', "\n", $markdown );
$markdown = preg_replace( '/]*>/i', "\n", $markdown );
$markdown = preg_replace( '/<\/ol>/i', "\n", $markdown );
$markdown = preg_replace( '/]*>(.*?)<\/li>/is', "- $1\n", $markdown );
// Blockquotes.
$markdown = preg_replace( '/]*>(.*?)<\/blockquote>/is', "> $1\n\n", $markdown );
// Code.
$markdown = preg_replace( '/]*>(.*?)<\/code>/is', '`$1`', $markdown );
$markdown = preg_replace( '/]*>(.*?)<\/pre>/is', "```\n$1\n```\n\n", $markdown );
// Horizontal rules.
$markdown = preg_replace( '//i', "\n---\n\n", $markdown );
// Remove remaining HTML tags and decode entities.
$markdown = wp_strip_all_tags( $markdown );
$markdown = html_entity_decode( $markdown, ENT_QUOTES | ENT_HTML5, 'UTF-8' );
// Clean up whitespace.
$markdown = preg_replace( "/\n{3,}/", "\n\n", $markdown );
$markdown = trim( $markdown );
return $markdown;
}
/**
* Extract main content region from a full HTML document.
*
* @param string $html HTML string.
* @return string
*/
function extract_main_content( string $html ): string {
$doc = new \DOMDocument();
libxml_use_internal_errors( true );
$wrapped = '' . $html . '';
$doc->loadHTML( $wrapped, LIBXML_HTML_NOIMPLIED | LIBXML_HTML_NODEFDTD );
$xpath = new \DOMXPath( $doc );
$selectors = array(
'//article[contains(@class, "post")]',
'//article[contains(@class, "entry")]',
'//main[contains(@class, "content")]',
'//main//article',
'//div[contains(@class, "entry-content")]',
'//div[contains(@class, "post-content")]',
'//div[contains(@class, "content")]',
'//article',
'//main',
);
foreach ( $selectors as $selector ) {
$nodes = $xpath->query( $selector );
if ( $nodes && $nodes->length > 0 ) {
$node = $nodes->item( 0 );
$inner = '';
foreach ( $node->childNodes as $child ) {
$inner .= $node->ownerDocument->saveHTML( $child );
}
return $inner;
}
}
// Fallback to body.
$body = $xpath->query( '//body' );
if ( $body && $body->length > 0 ) {
$inner = '';
foreach ( $body->item( 0 )->childNodes as $child ) {
$inner .= $body->item( 0 )->ownerDocument->saveHTML( $child );
}
return $inner;
}
return $html;
}
/**
* Handle a request that accepts Markdown.
*
* @return void
*/
function handle_markdown_request(): void {
if ( ! accepts_markdown() ) {
return;
}
if ( is_admin() || is_feed() || is_robots() || is_favicon() ) {
return;
}
if ( ! is_feature_enabled( 'markdown' ) ) {
return;
}
nocache_headers();
header( 'Content-Type: text/markdown; charset=UTF-8' );
header( 'Vary: Accept', false );
// Capture the rendered page in an output buffer whose callback rewrites
// the body as Markdown when the buffer is flushed. The conversion has to
// live in the buffer callback, not in a late `shutdown` action: core
// flushes every open buffer in wp_ob_end_flush_all() at shutdown
// priority 1, so by the time a later action ran the HTML had already
// gone to the client under a text/markdown Content-Type.
ob_start( __NAMESPACE__ . '\\markdown_output_callback' );
}
/**
* Output-buffer callback: turn the buffered HTML page into Markdown.
*
* @param string $content Buffered HTML.
* @return string Markdown body.
*/
function markdown_output_callback( string $content ): string {
if ( '' === trim( $content ) ) {
return $content;
}
$markdown = markdown_frontmatter() . html_to_markdown( extract_main_content( $content ) );
if ( ! headers_sent() ) {
header( 'X-Markdown-Tokens: ' . (int) ceil( strlen( $markdown ) / 4 ) );
}
return $markdown;
}