` heading followed by the content subtree's markdown, * per accordion item, blank-line separated. * * `Element_Base::render_markdown()` (the inherited default - fact 6) already recurses * `get_children()` and joins every non-empty result with a blank line, with no regard for * structure. Left unchanged, item -> header -> title -> paragraph and item -> content -> paragraph * would all flatten into one undifferentiated string per item ("Title\n\nContent"), and there is * no delimiter left in that output to tell where the title ends and the content begins - the * heading marker has to be added *before* the two are joined, not recovered afterwards. That * ruled out simply wrapping `parent::render_markdown()` and reformatting its return value. * * Instead this walks the `e-accordion-item` children directly - the same * `Collection::make($this->get_children())->filter()` idiom `get_item_index()` and * `build_faq_schema_json()` already use in this file - keeping each item's title and content * markdown separate until the `### ` heading is applied. The actual text/markdown extraction * still comes from calling `render_markdown()` on the title and content sub-elements (per the * plan's reuse-over-reimplementation preference, the same way `get_faq_item_text()` does above); * only the top-level assembly differs from the inherited default. * * @return string */ public function render_markdown(): string { $items = Collection::make( $this->get_children() ) ->filter( fn( $child ) => $child->get_type() === self::ELEMENT_TYPE_ITEM ); $blocks = []; foreach ( $items as $item ) { $title_markdown = trim( $this->get_item_title_markdown( $item ) ); $content_markdown = trim( $this->get_item_content_markdown( $item ) ); if ( '' === $title_markdown && '' === $content_markdown ) { continue; } $blocks[] = '### ' . $title_markdown . "\n\n" . $content_markdown; } return implode( "\n\n", $blocks ); } /** * Markdown of an item's title, via `e-accordion-item` -> `e-accordion-item-header` -> * `e-accordion-item-title` -> `render_markdown()`. * * @param Atomic_Element_Base $item * @return string */ private function get_item_title_markdown( $item ): string { $header = Collection::make( $item->get_children() ) ->filter( fn( $child ) => $child->get_type() === self::ELEMENT_TYPE_HEADER ) ->first(); if ( null === $header ) { return ''; } $title = Collection::make( $header->get_children() ) ->filter( fn( $child ) => $child->get_type() === self::ELEMENT_TYPE_TITLE ) ->first(); return null === $title ? '' : $title->render_markdown(); } /** * Markdown of an item's content, via `e-accordion-item` -> `e-accordion-item-content` -> * `render_markdown()`. * * @param Atomic_Element_Base $item * @return string */ private function get_item_content_markdown( $item ): string { $content = Collection::make( $item->get_children() ) ->filter( fn( $child ) => $child->get_type() === self::ELEMENT_TYPE_CONTENT ) ->first(); return null === $content ? '' : $content->render_markdown(); } protected function get_templates(): array { return [ 'elementor/elements/atomic-accordion' => __DIR__ . '/atomic-accordion.html.twig', ]; } /** * Editor-only: auto-opens the item containing the descendant the editor just selected * (`handlers/editor-accordion-state.js`). Nothing is added on the frontend - the toggle itself * stays native, no JS involved. * * @return array */ public function get_script_depends() { $global_depends = parent::get_script_depends(); if ( Plugin::$instance->preview->is_preview_mode() ) { return array_merge( $global_depends, [ 'elementor-accordion-preview-handler' ] ); } return $global_depends; } public function register_frontend_handlers() { $assets_url = ELEMENTOR_ASSETS_URL; $min_suffix = ( Utils::is_script_debug() || Utils::is_elementor_tests() ) ? '' : '.min'; wp_register_script( 'elementor-accordion-preview-handler', "{$assets_url}js/accordion-preview-handler{$min_suffix}.js", [ Frontend_Assets_Loader::FRONTEND_HANDLERS_HANDLE ], ELEMENTOR_VERSION, true ); } /** * Adds the FAQPage JSON-LD payload to the template context, when enabled. * * Kept out of the editor preview: structured data is a frontend/SEO concern, and building it * from live descendant markdown on every preview re-render would be wasted work with no * user-visible effect (the `<script type="application/ld+json">` tag renders invisibly). * * @return array */ protected function build_template_context(): array { $faq_schema_json = null; if ( $this->get_atomic_setting( 'faq_schema' ) && ! Plugin::$instance->preview->is_preview_mode() ) { $faq_schema_json = $this->build_faq_schema_json(); } return array_merge( $this->build_base_template_context(), [ 'faq_schema_json' => $faq_schema_json, ] ); } /** * Builds the `wp_json_encode()`-ready FAQPage JSON-LD string from the accordion's items, or * `null` when there is nothing valid to emit. * * @return string|null */ private function build_faq_schema_json() { $entities = []; $items = Collection::make( $this->get_children() ) ->filter( fn( $child ) => $child->get_type() === self::ELEMENT_TYPE_ITEM ); foreach ( $items as $item ) { $question = $this->get_faq_item_text( $item, self::ELEMENT_TYPE_HEADER, self::ELEMENT_TYPE_TITLE ); $answer = $this->get_faq_item_text( $item, null, self::ELEMENT_TYPE_CONTENT ); if ( '' === $question || '' === $answer ) { continue; } $entities[] = [ '@type' => 'Question', 'name' => $question, 'acceptedAnswer' => [ '@type' => 'Answer', 'text' => $answer, ], ]; } if ( empty( $entities ) ) { return null; } return wp_json_encode( [ '@context' => 'https://schema.org', '@type' => 'FAQPage', 'mainEntity' => $entities, ] ); } /** * Plain text of a sub-element within an `e-accordion-item`, found by element type. * * When `$via_type` is given, the target is looked up one level deeper (e.g. the title lives * inside the header, not directly on the item). `render_markdown()` is `Element_Base`'s existing * recursive default (fact 6) — reused as-is, not re-implemented. Its output is markdown syntax * (`**bold**`, `[text](url)`, …), not HTML, so `strip_markdown_syntax()` — not * `wp_strip_all_tags()` — does the real cleanup; `wp_strip_all_tags()` stays only as a * defensive second pass in case some element's `render_markdown()` ever emits raw HTML. * * @param Atomic_Element_Base $item * @param string|null $via_type * @param string $target_type * @return string */ private function get_faq_item_text( $item, $via_type, string $target_type ): string { $scope = $item; if ( null !== $via_type ) { $scope = Collection::make( $item->get_children() ) ->filter( fn( $child ) => $child->get_type() === $via_type ) ->first(); if ( null === $scope ) { return ''; } } $target = Collection::make( $scope->get_children() ) ->filter( fn( $child ) => $child->get_type() === $target_type ) ->first(); if ( null === $target ) { return ''; } $plain = $this->strip_markdown_syntax( $target->render_markdown() ); return trim( wp_strip_all_tags( $plain ) ); } /** * Strips the markdown syntax `render_markdown()` produces via `Html_To_Markdown::convert()` * (see `modules/markdown-render/html-to-markdown.php`) so it never leaks into JSON-LD text * fields verbatim — `**bold**`, `[text](url)`, `# heading`, etc. are not valid prose, and a * FAQPage consumer (Google, any schema.org parser) should see clean text. `wp_strip_all_tags()` * cannot do this job: it strips HTML tags, and markdown syntax is punctuation, not HTML. * * Deliberately scoped to exactly what `Html_To_Markdown::convert()` can emit in this codebase's * configuration, not a general-purpose markdown parser: * - fenced ``` code blocks and inline `code` spans (backtick fence length varies) * - `![alt](src)` images and `[text](url)` links — the visible text is kept, the target dropped * - `**bold**` / `__bold__`, `*italic*` / `_italic_`, `~~strikethrough~~` * - `#` … `######` heading markers and `-`/`*`/`+`/`1.` list markers at line start * - `> ` blockquote markers (repeated per nesting depth) at line start — reachable from a plain * `Atomic_Paragraph`, which whitelists `<blockquote>` in its own output filter and feeds it * straight into `Html_To_Markdown::convert()`, so this is not a hypothetical case * - the backslash-escaping `Html_To_Markdown::escape_text()` applies to literal * `\ \` * _ [ ]` characters in ordinary text, so those round-trip back to plain characters * instead of surfacing a stray backslash * * The formatting-delimiter regexes below (code/bold/italic/link/image) all require an * *unescaped* delimiter via `(?<!\\)`: `escape_text()` backslash-escapes a literal `* _ \` [ ]` * character wherever it appears in plain text (not just when paired), so a string like * `1* free*` (literal asterisks, no formatting intended) round-trips through `render_markdown()` * as `1\* free\*`. Without the guard, the two escaped asterisks look like a real `*...*` pair to * a naive regex, which eats the enclosed text and leaves orphaned backslashes behind — this was * exactly the Round 1→2 regression. Unescaping runs *after* the delimiter regexes so the escaped * characters are still backslash-prefixed (and therefore excluded by the lookbehind) while those * regexes run. * * Tables aren't handled: no reachable path in this codebase currently produces table markdown * for an accordion title/content subtree, and adding unused coverage would be guessing at * requirements rather than matching them. * * @param string $markdown * @return string */ private function strip_markdown_syntax( string $markdown ): string { $text = $markdown; // Fenced code blocks: keep only the code content. $text = preg_replace( '/```[^\n`]*\n(.*?)\n```/s', '$1', $text ); // Inline code spans (only when both fences are unescaped — see docblock). $text = preg_replace( '/(?<!\\\\)`+(.*?)(?<!\\\\)`+/s', '$1', $text ); // Images and links: keep the visible text, drop the target (unescaped opening bracket only). $text = preg_replace( '/(?<!\\\\)!\[([^\]]*)\]$[^)]*$/', '$1', $text ); $text = preg_replace( '/(?<!\\\\)\[([^\]]*)\]$[^)]*$/', '$1', $text ); // Headings: leading #'s at the start of a line. $text = preg_replace( '/^#{1,6}[ \t]+/m', '', $text ); // List markers: "- ", "* ", "+ " or "1. " at a (possibly indented) line start. $text = preg_replace( '/^[ \t]*(?:[-*+]|\d+\.)[ \t]+/m', '', $text ); // Blockquote markers: "> " (repeated for nested quotes) at the start of a line. $text = preg_replace( '/^(?:>[ \t]?)+/m', '', $text ); // Bold, italic (unescaped delimiters only — see docblock), strikethrough (never escaped). $text = preg_replace( '/(?<!\\\\)(\*\*|__)(.+?)(?<!\\\\)\1/s', '$2', $text ); $text = preg_replace( '/(?<!\\\\)\*(.+?)(?<!\\\\)\*/s', '$1', $text ); $text = preg_replace( '/(?<!\\\\)_(.+?)(?<!\\\\)_/s', '$1', $text ); $text = preg_replace( '/~~(.+?)~~/s', '$1', $text ); // Undo the converter's backslash-escaping of literal markdown-special characters. $text = preg_replace( '/\\\\([\\\\`*_\[\]])/', '$1', $text ); // Collapse whitespace left behind by the removals above. $text = preg_replace( '/[ \t]+/', ' ', $text ); $text = preg_replace( '/\s*\n\s*/', ' ', $text ); return trim( $text ); } /** * Exposes the accordion's identity and per-item lookup to descendants that render inside its * pass (item, header, title, icon, content) via `Render_Context::get( self::class )`. * * @return array */ protected function define_render_context(): array { return [ [ 'context' => [ 'accordion-id' => $this->get_id(), 'get-item-index' => fn( $item_id ) => $this->get_item_index( $item_id ), 'default-state' => $this->get_atomic_setting( 'default_state' ), 'max-expanded' => $this->get_atomic_setting( 'max

semantics, with no JavaScript needed for the toggle. Structure: e-accordion contains e-accordion-item elements; each item contains an e-accordion-item-header (holding e-accordion-item-title and an optional e-accordion-item-icon) and an e-accordion-item-content that accepts any element.'; public function __construct( $data = [], $args = null ) { parent::__construct( $data, $args ); $this->meta( 'is_container', true ); $this->meta( 'is_compound', true ); } public static function get_type() { return 'e-accordion'; } public static function get_element_type(): string { return 'e-accordion'; } public function get_title() { return esc_html__( 'Accordion', 'elementor' ); } public function get_keywords() { return [ 'ato', 'atom', 'atoms', 'atomic', 'accordion', 'faq', 'collapse' ]; } public function get_icon() { return 'eicon-accordion'; } public static function get_computed_html_tag( array $settings ): string { return Html_Tag_Computer::compute( $settings, 'div' ); } protected static function define_props_schema(): array { return [ 'classes' => Classes_Prop_Type::make() ->default( [] ), 'attributes' => Attributes_Prop_Type::make()->meta( Overridable_Prop_Type::ignore() ), 'default_state' => String_Prop_Type::make() ->enum( [ 'first_expanded', 'all_collapsed' ] ) ->default( 'first_expanded' ) ->description( 'Which items are open when the accordion first renders. Valid values: first_expanded, all_collapsed' ), 'max_expanded' => String_Prop_Type::make() ->enum( [ 'one', 'multiple' ] ) ->default( 'one' ) ->description( 'How many items can be open at the same time. Valid values: one, multiple' ), // Single, global, user-facing toggle - there is no per-item Show Icon (confirmed with the // PM: users want icons on every item or on none, never mixed). It is exposable as a // component property, so no `Overridable_Prop_Type::ignore()`. Every `e-accordion-item-header` // carries its own mirrored `show_icon` prop (see that class) purely because the children- // dependencies reconciler evaluates a rule against the *declaring* element's own settings // and can only attach/detach that element's *direct* children - a root-level prop here can // never drive a grandchild's presence. Do not "clean up" the mirror; the duplication is // structural, not an oversight. 'show_icon' => Boolean_Prop_Type::make()->default( true ) ->description( 'Whether every item header shows an open/closed indicator icon. Applies to all items; there is no per-item override.' ), 'faq_schema' => Boolean_Prop_Type::make()->default( false ) ->description( 'Whether to output an FAQPage JSON-LD structured data script on the frontend, built from each item\'s title (question) and content (answer).' ), ]; } protected function define_atomic_controls(): array { return [ Section::make() ->set_label( __( 'Content', 'elementor' ) ) ->set_id( 'content' ) ->set_items( [ Accordion_Items_Control::make() ->set_label( __( 'Accordion Items', 'elementor' ) ) ->set_meta( [ 'layout' => 'custom', ] ), Switch_Control::bind_to( 'show_icon' ) ->set_label( esc_html__( 'Show Icon', 'elementor' ) ), ] ), Section::make() ->set_label( __( 'Settings', 'elementor' ) ) ->set_id( 'settings' ) ->set_items( [ Toggle_Control::bind_to( 'default_state' ) ->set_label( esc_html__( 'Default State', 'elementor' ) ) ->add_options( [ 'first_expanded' => [ 'title' => esc_html__( 'First expanded', 'elementor' ) ], 'all_collapsed' => [ 'title' => esc_html__( 'All collapsed', 'elementor' ) ], ] ) ->set_exclusive( true ) ->set_convert_options( true ) ->set_size( 'tiny' ) ->set_full_width( true ), Toggle_Control::bind_to( 'max_expanded' ) ->set_label( esc_html__( 'Max Items Expanded', 'elementor' ) ) ->add_options( [ 'one' => [ 'title' => esc_html__( 'One', 'elementor' ) ], 'multiple' => [ 'title' => esc_html__( 'Multiple', 'elementor' ) ], ] ) ->set_exclusive( true ) ->set_convert_options( true ) ->set_size( 'tiny' ) ->set_full_width( true ), Switch_Control::bind_to( 'faq_schema' ) ->set_label( esc_html__( 'FAQ Schema', 'elementor' ) ), Text_Control::bind_to( '_cssid' ) ->set_label( __( 'ID', 'elementor' ) ) ->set_meta( $this->get_css_id_control_meta() ), ] ), ]; } protected function define_base_styles(): array { return [ static::BASE_STYLE_KEY => Style_Definition::make() ->add_variant( Style_Variant::make() ->add_props( [ 'display' => String_Prop_Type::generate( 'flex' ), 'flex-direction' => String_Prop_Type::generate( 'column' ), // Neutralises the 10px inline padding that `.e-con` (added to every atomic // element by `render_base_classes`) resolves from the container defaults. // The item's own 10px lives on the header and the content slot, so any // padding here or on the item would just inset the whole accordion. 'padding' => Size_Prop_Type::generate( [ 'size' => 0, 'unit' => 'px', ] ), ] ) ), ]; } protected function define_allowed_child_types() { return [ self::ELEMENT_TYPE_ITEM ]; } protected function define_default_children() { $items = []; foreach ( range( 1, self::DEFAULT_ITEM_COUNT ) as $i ) { $items[] = $this->build_default_item( $i ); } return $items; } /** * Builds one `e-accordion-item` with its numbered title seeded explicitly. * * Each level of `default_children` is hydrated independently client-side * (`Atomic_Element_Base_Model::onElementCreate()`, driven by `hydrateDefaultChildren: true`), * so the title slot's own `define_default_children()` has no way to know which item it * belongs to. Building the item → header → title → paragraph chain explicitly here lets the * numbered text ("Accordion Item 1", "Accordion Item 2") reach the rendered paragraph. The * icon and content branches don't need per-index content, so they keep using their own * `define_default_children()` via `hydrate_default_children( true )`. * * @param int $index * @return array */ private function build_default_item( int $index ): array { /* translators: %d: Accordion item position. */ $numbered_title = sprintf( esc_html__( 'Accordion Item %d', 'elementor' ), $index ); $title = Element_Builder::make( self::ELEMENT_TYPE_TITLE ) ->editor_settings( [ 'title' => esc_html__( 'Title', 'elementor' ), ] ) ->children( [ Atomic_Paragraph::generate() ->settings( [ 'paragraph' => Escaped_Html_Prop_Type::generate( $numbered_title ), 'tag' => String_Prop_Type::generate( 'span' ), ] ) ->build(), ] ) ->build(); $icon = Element_Builder::make( self::ELEMENT_TYPE_ICON ) ->hydrate_default_children( true ) ->editor_settings( [ 'title' => esc_html__( 'Icon', 'elementor' ), ] ) ->build(); $header = Element_Builder::make( self::ELEMENT_TYPE_HEADER ) ->editor_settings( [ 'title' => esc_html__( 'Header', 'elementor' ), ] ) // Seeded explicitly, mirroring the TS repeater's `buildItemModel()` // (`accordion-items-control/use-actions.ts`) - both sides of the item-building split need // the header's mirrored `show_icon` prop set on creation, not left to its own schema default, // or the two default items this method builds and any item added later through the panel // would start from different `show_icon` states (set vs. unset). See the comment on the // mirrored prop in `Atomic_Accordion_Item_Header` for why the duplication exists at all. // // Reads the *schema's* default here, not `$this->get_atomic_setting( 'show_icon' )`: this // runs while `define_default_children()` is itself building the initial config, before // `ensure_settings()` has finished resolving this element's own settings, so calling // `get_atomic_setting()` here recurses back into the same in-progress settings resolution // and fatals. There is also no "current" root value to seed from yet - the two default // items always start from the schema default, the same as the root's own `show_icon` // control does the first time it renders. ->settings( [ 'show_icon' => self::define_props_schema()['show_icon']->get_default(), ] ) ->children( [ $title, $icon ] ) ->build(); $content = Element_Builder::make( self::ELEMENT_TYPE_CONTENT ) ->hydrate_default_children( true ) ->editor_settings( [ 'title' => esc_html__( 'Content', 'elementor' ), ] ) ->build(); return Element_Builder::make( self::ELEMENT_TYPE_ITEM ) ->editor_settings( [ 'title' => $numbered_title, 'initial_position' => $index, ] ) ->children( [ $header, $content ] ) ->build(); } /** * Index of a direct `e-accordion-item` child, by element id. * * Items read this through the render context to decide which one carries the `open` * attribute. Mirrors `Atomic_Tabs::get_tab_index()`. * * @param string $item_id * @return int|null */ private function get_item_index( string $item_id ) { $item_ids = Collection::make( $this->get_children() ) ->filter( fn( $child ) => $child->get_type() === self::ELEMENT_TYPE_ITEM ) ->map( fn( $child ) => $child->get_id() ) ->flip() ->all(); return $item_ids[ $item_id ] ?? null; } /** * Root markdown output: one `### ` heading followed by the content subtree's markdown, * per accordion item, blank-line separated. * * `Element_Base::render_markdown()` (the inherited default - fact 6) already recurses * `get_children()` and joins every non-empty result with a blank line, with no regard for * structure. Left unchanged, item -> header -> title -> paragraph and item -> content -> paragraph * would all flatten into one undifferentiated string per item ("Title\n\nContent"), and there is * no delimiter left in that output to tell where the title ends and the content begins - the * heading marker has to be added *before* the two are joined, not recovered afterwards. That * ruled out simply wrapping `parent::render_markdown()` and reformatting its return value. * * Instead this walks the `e-accordion-item` children directly - the same * `Collection::make($this->get_children())->filter()` idiom `get_item_index()` and * `build_faq_schema_json()` already use in this file - keeping each item's title and content * markdown separate until the `### ` heading is applied. The actual text/markdown extraction * still comes from calling `render_markdown()` on the title and content sub-elements (per the * plan's reuse-over-reimplementation preference, the same way `get_faq_item_text()` does above); * only the top-level assembly differs from the inherited default. * * @return string */ public function render_markdown(): string { $items = Collection::make( $this->get_children() ) ->filter( fn( $child ) => $child->get_type() === self::ELEMENT_TYPE_ITEM ); $blocks = []; foreach ( $items as $item ) { $title_markdown = trim( $this->get_item_title_markdown( $item ) ); $content_markdown = trim( $this->get_item_content_markdown( $item ) ); if ( '' === $title_markdown && '' === $content_markdown ) { continue; } $blocks[] = '### ' . $title_markdown . "\n\n" . $content_markdown; } return implode( "\n\n", $blocks ); } /** * Markdown of an item's title, via `e-accordion-item` -> `e-accordion-item-header` -> * `e-accordion-item-title` -> `render_markdown()`. * * @param Atomic_Element_Base $item * @return string */ private function get_item_title_markdown( $item ): string { $header = Collection::make( $item->get_children() ) ->filter( fn( $child ) => $child->get_type() === self::ELEMENT_TYPE_HEADER ) ->first(); if ( null === $header ) { return ''; } $title = Collection::make( $header->get_children() ) ->filter( fn( $child ) => $child->get_type() === self::ELEMENT_TYPE_TITLE ) ->first(); return null === $title ? '' : $title->render_markdown(); } /** * Markdown of an item's content, via `e-accordion-item` -> `e-accordion-item-content` -> * `render_markdown()`. * * @param Atomic_Element_Base $item * @return string */ private function get_item_content_markdown( $item ): string { $content = Collection::make( $item->get_children() ) ->filter( fn( $child ) => $child->get_type() === self::ELEMENT_TYPE_CONTENT ) ->first(); return null === $content ? '' : $content->render_markdown(); } protected function get_templates(): array { return [ 'elementor/elements/atomic-accordion' => __DIR__ . '/atomic-accordion.html.twig', ]; } /** * Editor-only: auto-opens the item containing the descendant the editor just selected * (`handlers/editor-accordion-state.js`). Nothing is added on the frontend - the toggle itself * stays native, no JS involved. * * @return array */ public function get_script_depends() { $global_depends = parent::get_script_depends(); if ( Plugin::$instance->preview->is_preview_mode() ) { return array_merge( $global_depends, [ 'elementor-accordion-preview-handler' ] ); } return $global_depends; } public function register_frontend_handlers() { $assets_url = ELEMENTOR_ASSETS_URL; $min_suffix = ( Utils::is_script_debug() || Utils::is_elementor_tests() ) ? '' : '.min'; wp_register_script( 'elementor-accordion-preview-handler', "{$assets_url}js/accordion-preview-handler{$min_suffix}.js", [ Frontend_Assets_Loader::FRONTEND_HANDLERS_HANDLE ], ELEMENTOR_VERSION, true ); } /** * Adds the FAQPage JSON-LD payload to the template context, when enabled. * * Kept out of the editor preview: structured data is a frontend/SEO concern, and building it * from live descendant markdown on every preview re-render would be wasted work with no * user-visible effect (the `<script type="application/ld+json">` tag renders invisibly). * * @return array */ protected function build_template_context(): array { $faq_schema_json = null; if ( $this->get_atomic_setting( 'faq_schema' ) && ! Plugin::$instance->preview->is_preview_mode() ) { $faq_schema_json = $this->build_faq_schema_json(); } return array_merge( $this->build_base_template_context(), [ 'faq_schema_json' => $faq_schema_json, ] ); } /** * Builds the `wp_json_encode()`-ready FAQPage JSON-LD string from the accordion's items, or * `null` when there is nothing valid to emit. * * @return string|null */ private function build_faq_schema_json() { $entities = []; $items = Collection::make( $this->get_children() ) ->filter( fn( $child ) => $child->get_type() === self::ELEMENT_TYPE_ITEM ); foreach ( $items as $item ) { $question = $this->get_faq_item_text( $item, self::ELEMENT_TYPE_HEADER, self::ELEMENT_TYPE_TITLE ); $answer = $this->get_faq_item_text( $item, null, self::ELEMENT_TYPE_CONTENT ); if ( '' === $question || '' === $answer ) { continue; } $entities[] = [ '@type' => 'Question', 'name' => $question, 'acceptedAnswer' => [ '@type' => 'Answer', 'text' => $answer, ], ]; } if ( empty( $entities ) ) { return null; } return wp_json_encode( [ '@context' => 'https://schema.org', '@type' => 'FAQPage', 'mainEntity' => $entities, ] ); } /** * Plain text of a sub-element within an `e-accordion-item`, found by element type. * * When `$via_type` is given, the target is looked up one level deeper (e.g. the title lives * inside the header, not directly on the item). `render_markdown()` is `Element_Base`'s existing * recursive default (fact 6) — reused as-is, not re-implemented. Its output is markdown syntax * (`**bold**`, `[text](url)`, …), not HTML, so `strip_markdown_syntax()` — not * `wp_strip_all_tags()` — does the real cleanup; `wp_strip_all_tags()` stays only as a * defensive second pass in case some element's `render_markdown()` ever emits raw HTML. * * @param Atomic_Element_Base $item * @param string|null $via_type * @param string $target_type * @return string */ private function get_faq_item_text( $item, $via_type, string $target_type ): string { $scope = $item; if ( null !== $via_type ) { $scope = Collection::make( $item->get_children() ) ->filter( fn( $child ) => $child->get_type() === $via_type ) ->first(); if ( null === $scope ) { return ''; } } $target = Collection::make( $scope->get_children() ) ->filter( fn( $child ) => $child->get_type() === $target_type ) ->first(); if ( null === $target ) { return ''; } $plain = $this->strip_markdown_syntax( $target->render_markdown() ); return trim( wp_strip_all_tags( $plain ) ); } /** * Strips the markdown syntax `render_markdown()` produces via `Html_To_Markdown::convert()` * (see `modules/markdown-render/html-to-markdown.php`) so it never leaks into JSON-LD text * fields verbatim — `**bold**`, `[text](url)`, `# heading`, etc. are not valid prose, and a * FAQPage consumer (Google, any schema.org parser) should see clean text. `wp_strip_all_tags()` * cannot do this job: it strips HTML tags, and markdown syntax is punctuation, not HTML. * * Deliberately scoped to exactly what `Html_To_Markdown::convert()` can emit in this codebase's * configuration, not a general-purpose markdown parser: * - fenced ``` code blocks and inline `code` spans (backtick fence length varies) * - `![alt](src)` images and `[text](url)` links — the visible text is kept, the target dropped * - `**bold**` / `__bold__`, `*italic*` / `_italic_`, `~~strikethrough~~` * - `#` … `######` heading markers and `-`/`*`/`+`/`1.` list markers at line start * - `> ` blockquote markers (repeated per nesting depth) at line start — reachable from a plain * `Atomic_Paragraph`, which whitelists `<blockquote>` in its own output filter and feeds it * straight into `Html_To_Markdown::convert()`, so this is not a hypothetical case * - the backslash-escaping `Html_To_Markdown::escape_text()` applies to literal * `\ \` * _ [ ]` characters in ordinary text, so those round-trip back to plain characters * instead of surfacing a stray backslash * * The formatting-delimiter regexes below (code/bold/italic/link/image) all require an * *unescaped* delimiter via `(?<!\\)`: `escape_text()` backslash-escapes a literal `* _ \` [ ]` * character wherever it appears in plain text (not just when paired), so a string like * `1* free*` (literal asterisks, no formatting intended) round-trips through `render_markdown()` * as `1\* free\*`. Without the guard, the two escaped asterisks look like a real `*...*` pair to * a naive regex, which eats the enclosed text and leaves orphaned backslashes behind — this was * exactly the Round 1→2 regression. Unescaping runs *after* the delimiter regexes so the escaped * characters are still backslash-prefixed (and therefore excluded by the lookbehind) while those * regexes run. * * Tables aren't handled: no reachable path in this codebase currently produces table markdown * for an accordion title/content subtree, and adding unused coverage would be guessing at * requirements rather than matching them. * * @param string $markdown * @return string */ private function strip_markdown_syntax( string $markdown ): string { $text = $markdown; // Fenced code blocks: keep only the code content. $text = preg_replace( '/```[^\n`]*\n(.*?)\n```/s', '$1', $text ); // Inline code spans (only when both fences are unescaped — see docblock). $text = preg_replace( '/(?<!\\\\)`+(.*?)(?<!\\\\)`+/s', '$1', $text ); // Images and links: keep the visible text, drop the target (unescaped opening bracket only). $text = preg_replace( '/(?<!\\\\)!\[([^\]]*)\]$[^)]*$/', '$1', $text ); $text = preg_replace( '/(?<!\\\\)\[([^\]]*)\]$[^)]*$/', '$1', $text ); // Headings: leading #'s at the start of a line. $text = preg_replace( '/^#{1,6}[ \t]+/m', '', $text ); // List markers: "- ", "* ", "+ " or "1. " at a (possibly indented) line start. $text = preg_replace( '/^[ \t]*(?:[-*+]|\d+\.)[ \t]+/m', '', $text ); // Blockquote markers: "> " (repeated for nested quotes) at the start of a line. $text = preg_replace( '/^(?:>[ \t]?)+/m', '', $text ); // Bold, italic (unescaped delimiters only — see docblock), strikethrough (never escaped). $text = preg_replace( '/(?<!\\\\)(\*\*|__)(.+?)(?<!\\\\)\1/s', '$2', $text ); $text = preg_replace( '/(?<!\\\\)\*(.+?)(?<!\\\\)\*/s', '$1', $text ); $text = preg_replace( '/(?<!\\\\)_(.+?)(?<!\\\\)_/s', '$1', $text ); $text = preg_replace( '/~~(.+?)~~/s', '$1', $text ); // Undo the converter's backslash-escaping of literal markdown-special characters. $text = preg_replace( '/\\\\([\\\\`*_\[\]])/', '$1', $text ); // Collapse whitespace left behind by the removals above. $text = preg_replace( '/[ \t]+/', ' ', $text ); $text = preg_replace( '/\s*\n\s*/', ' ', $text ); return trim( $text ); } /** * Exposes the accordion's identity and per-item lookup to descendants that render inside its * pass (item, header, title, icon, content) via `Render_Context::get( self::class )`. * * @return array */ protected function define_render_context(): array { return [ [ 'context' => [ 'accordion-id' => $this->get_id(), 'get-item-index' => fn( $item_id ) => $this->get_item_index( $item_id ), 'default-state' => $this->get_atomic_setting( 'default_state' ), 'max-expanded' => $this->get_atomic_setting( 'max_expanded' ), ], ], ]; } }