Support templated content

The template element in HTML 5 is special in the sense that it does not
add its contents into the DOM tree, but instead keeps them in a separate
shadow DOM document fragment. Interacting with the DOM tree cannot touch
the elements in the document fragment.

Author: nielsdos

ext/dom/config.m4

@@ -27,7 +27,7 @@ if test "$PHP_DOM" != "no"; then
       $LEXBOR_DIR/ns/ns.c \
       $LEXBOR_DIR/tag/tag.c"
     PHP_NEW_EXTENSION(dom, [php_dom.c attr.c document.c infra.c \
-                            xml_document.c html_document.c xml_serializer.c html5_serializer.c html5_parser.c namespace_compat.c \
+                            xml_document.c html_document.c xml_serializer.c html5_serializer.c html5_parser.c namespace_compat.c private_data.c \
                             domexception.c \
                             parentnode/tree.c parentnode/css_selectors.c \
                             processinginstruction.c cdatasection.c \

ext/dom/config.w32

@@ -8,7 +8,7 @@ if (PHP_DOM == "yes") {
 		CHECK_HEADER_ADD_INCLUDE("libxml/parser.h", "CFLAGS_DOM", PHP_PHP_BUILD + "\\include\\libxml2")
 	) {
 		EXTENSION("dom", "php_dom.c attr.c document.c infra.c \
-			xml_document.c html_document.c xml_serializer.c html5_serializer.c html5_parser.c namespace_compat.c \
+			xml_document.c html_document.c xml_serializer.c html5_serializer.c html5_parser.c namespace_compat.c private_data.c \
 			domexception.c processinginstruction.c \
 			cdatasection.c documentfragment.c domimplementation.c element.c  inner_html_mixin.c \
 			node.c characterdata.c documenttype.c \

ext/dom/domimplementation.c

@@ -23,6 +23,7 @@
 #if defined(HAVE_LIBXML) && defined(HAVE_DOM)
 #include "php_dom.h"
 #include "namespace_compat.h"
+#include "private_data.h"
 
 /*
 * class DOMImplementation

ext/dom/element.c

@@ -24,6 +24,7 @@
 #include "zend_enum.h"
 #include "php_dom.h"
 #include "namespace_compat.h"
+#include "private_data.h"
 #include "internal_helpers.h"
 #include "dom_properties.h"
 #include "token_list.h"
@@ -2030,6 +2031,11 @@ PHP_METHOD(Dom_Element, rename)
 			}
 			goto cleanup;
 		}
+
+		/* If we currently have a template but the new element type won't be a template, then throw away the templated content. */
+		if (is_currently_html_ns && xmlStrEqual(nodep->name, BAD_CAST "template") && !xmlStrEqual(localname, BAD_CAST "template")) {
+			php_dom_remove_templated_content(php_dom_get_private_data(intern), nodep);
+		}
 	}
 
 	php_libxml_invalidate_node_list_cache(intern->document);

ext/dom/html5_parser.c

@@ -22,12 +22,13 @@
 #if defined(HAVE_LIBXML) && defined(HAVE_DOM)
 #include "php_dom.h"
 #include "html5_parser.h"
+#include "private_data.h"
 #include <lexbor/html/parser.h>
 #include <lexbor/html/interfaces/element.h>
+#include <lexbor/html/interfaces/template_element.h>
 #include <lexbor/dom/dom.h>
 #include <libxml/parserInternals.h>
 #include <libxml/HTMLtree.h>
-#include <Zend/zend.h>
 
 #define WORK_LIST_INIT_SIZE 128
 /* libxml2 reserves 2 pointer-sized words for interned strings */
@@ -102,11 +103,12 @@ static lexbor_libxml2_bridge_status lexbor_libxml2_bridge_convert(
     xmlNodePtr root,
     bool compact_text_nodes,
     bool create_default_ns,
-    php_dom_libxml_ns_mapper *ns_mapper
+    php_dom_private_data *private_data
 )
 {
     lexbor_libxml2_bridge_status retval = LEXBOR_LIBXML2_BRIDGE_STATUS_OK;
 
+	php_dom_libxml_ns_mapper *ns_mapper = php_dom_ns_mapper_from_private(private_data);
     xmlNsPtr html_ns = php_dom_libxml_ns_mapper_ensure_html_ns(ns_mapper);
     xmlNsPtr xlink_ns = NULL;
     xmlNsPtr prefixed_xmlns_ns = NULL;
@@ -158,12 +160,32 @@ static lexbor_libxml2_bridge_status lexbor_libxml2_bridge_convert(
             /* Instead of xmlSetNs() because we know the arguments are valid. Prevents overhead. */
             lxml_element->ns = current_lxml_ns;
 
-            for (lxb_dom_node_t *child_node = element->node.last_child; child_node != NULL; child_node = child_node->prev) {
+			/* Handle template element by creating a fragment node to contain its children.
+			 * Other types of nodes contain their children directly. */
+			xmlNodePtr lxml_child_parent = lxml_element;
+			lxb_dom_node_t *child_node = element->node.last_child;
+			if (lxb_html_tree_node_is(&element->node, LXB_TAG_TEMPLATE)) {
+				lxml_child_parent = xmlNewDocFragment(lxml_doc);
+				if (UNEXPECTED(lxml_child_parent == NULL)) {
+					retval = LEXBOR_LIBXML2_BRIDGE_STATUS_OOM;
+					break;
+				}
+
+				lxml_child_parent->parent = lxml_element;
+				php_dom_add_templated_content(private_data, lxml_element, lxml_child_parent);
+
+				lxb_html_template_element_t *template = lxb_html_interface_template(&element->node);
+				if (template->content != NULL) {
+					child_node = template->content->node.last_child;
+				}
+			}
+
+            for (; child_node != NULL; child_node = child_node->prev) {
                 lexbor_libxml2_bridge_work_list_item_push(
                     &work_list,
                     child_node,
                     entering_namespace,
-                    lxml_element,
+                    lxml_child_parent,
                     current_lxml_ns
                 );
             }
@@ -307,7 +329,7 @@ lexbor_libxml2_bridge_status lexbor_libxml2_bridge_convert_document(
     xmlDocPtr *doc_out,
     bool compact_text_nodes,
     bool create_default_ns,
-    php_dom_libxml_ns_mapper *ns_mapper
+	php_dom_private_data *private_data
 )
 {
     xmlDocPtr lxml_doc = php_dom_create_html_doc();
@@ -320,7 +342,7 @@ lexbor_libxml2_bridge_status lexbor_libxml2_bridge_convert_document(
         (xmlNodePtr) lxml_doc,
         compact_text_nodes,
         create_default_ns,
-        ns_mapper
+        private_data
     );
     if (status != LEXBOR_LIBXML2_BRIDGE_STATUS_OK) {
         xmlFreeDoc(lxml_doc);
@@ -336,7 +358,7 @@ lexbor_libxml2_bridge_status lexbor_libxml2_bridge_convert_fragment(
     xmlNodePtr *fragment_out,
     bool compact_text_nodes,
     bool create_default_ns,
-    php_dom_libxml_ns_mapper *ns_mapper
+	php_dom_private_data *private_data
 )
 {
     xmlNodePtr fragment = xmlNewDocFragment(lxml_doc);
@@ -349,7 +371,7 @@ lexbor_libxml2_bridge_status lexbor_libxml2_bridge_convert_fragment(
         fragment,
         compact_text_nodes,
         create_default_ns,
-        ns_mapper
+        private_data
     );
     if (status != LEXBOR_LIBXML2_BRIDGE_STATUS_OK) {
         xmlFreeNode(fragment);

ext/dom/html5_parser.h

@@ -71,15 +71,15 @@ lexbor_libxml2_bridge_status lexbor_libxml2_bridge_convert_document(
     xmlDocPtr *doc_out,
     bool compact_text_nodes,
     bool create_default_ns,
-    php_dom_libxml_ns_mapper *ns_mapper
+    php_dom_private_data *private_data
 );
 lexbor_libxml2_bridge_status lexbor_libxml2_bridge_convert_fragment(
     lxb_dom_node_t *start_node,
     xmlDocPtr lxml_doc,
     xmlNodePtr *fragment_out,
     bool compact_text_nodes,
     bool create_default_ns,
-    php_dom_libxml_ns_mapper *ns_mapper
+    php_dom_private_data *private_data
 );
 void lexbor_libxml2_bridge_report_errors(
     const lexbor_libxml2_bridge_parse_context *ctx,

ext/dom/html5_serializer.c

@@ -289,9 +289,13 @@ static zend_result dom_html5_serialize_node(dom_html5_serialize_context *ctx, co
 
 			case XML_ELEMENT_NODE: {
 				TRY(dom_html5_serialize_element_start(ctx, node));
-				if (node->children) {
+				const xmlNode *children = node->children;
+				if (php_dom_ns_is_fast(node, php_dom_ns_is_html_magic_token) && xmlStrEqual(node->name, BAD_CAST "template")) {
+					children = php_dom_retrieve_templated_content(ctx->private_data, node);
+				}
+				if (children) {
 					if (!dom_html5_serializes_as_void(node)) {
-						node = node->children;
+						node = children;
 						continue;
 					}
 				} else {
@@ -301,6 +305,14 @@ static zend_result dom_html5_serialize_node(dom_html5_serialize_context *ctx, co
 				break;
 			}
 
+			case XML_DOCUMENT_FRAG_NODE: {
+				if (node->children) {
+					node = node->children;
+					continue;
+				}
+				break;
+			}
+
 			/* Only exists for compatibility with XML and old DOM. */
 			case XML_ENTITY_REF_NODE: {
 				TRY(dom_html5_serialize_entity_ref(ctx, node));
@@ -346,10 +358,15 @@ zend_result dom_html5_serialize(dom_html5_serialize_context *ctx, const xmlNode
 	}
 
 	/* Step 2 not needed because we're not using a string to store the serialized data */
-	/* Step 3 not needed because we don't support template contents yet */
+
+	/* Step 3. If the node is a template element, then let the node instead be the template element's template contents (a DocumentFragment node). */
+	xmlNodePtr children = php_dom_retrieve_templated_content(ctx->private_data, node);
+	if (!children) {
+		children = node->children;
+	}
 
 	/* Step 4 */
-	return dom_html5_serialize_node(ctx, node->children, node);
+	return dom_html5_serialize_node(ctx, children, node);
 }
 
 /* Variant on the above that is equivalent to the "outer HTML". */

ext/dom/html5_serializer.h

@@ -19,11 +19,13 @@
 
 #include <Zend/zend_types.h>
 #include <libxml/tree.h>
+#include "private_data.h"
 
 typedef struct {
 	zend_result (*write_string)(void *application_data, const char *buf);
 	zend_result (*write_string_len)(void *application_data, const char *buf, size_t len);
 	void *application_data;
+	php_dom_private_data *private_data;
 } dom_html5_serialize_context;
 
 zend_result dom_html5_serialize(dom_html5_serialize_context *ctx, const xmlNode *node);

ext/dom/html_document.c

@@ -25,6 +25,7 @@
 #include "html5_parser.h"
 #include "html5_serializer.h"
 #include "namespace_compat.h"
+#include "private_data.h"
 #include "dom_properties.h"
 #include <Zend/zend_smart_string.h>
 #include <lexbor/html/encoding.h>
@@ -879,15 +880,14 @@ PHP_METHOD(Dom_HTMLDocument, createFromString)
 	}
 
 	php_dom_private_data *private_data = php_dom_private_data_create();
-	php_dom_libxml_ns_mapper *ns_mapper = php_dom_ns_mapper_from_private(private_data);
 
 	xmlDocPtr lxml_doc;
 	lexbor_libxml2_bridge_status bridge_status = lexbor_libxml2_bridge_convert_document(
 		document,
 		&lxml_doc,
 		options & XML_PARSE_COMPACT,
 		!(options & DOM_HTML_NO_DEFAULT_NS),
-		ns_mapper
+		private_data
 	);
 	lexbor_libxml2_bridge_copy_observations(parser->tree, &ctx.observations);
 	if (UNEXPECTED(bridge_status != LEXBOR_LIBXML2_BRIDGE_STATUS_OK)) {
@@ -1071,15 +1071,14 @@ PHP_METHOD(Dom_HTMLDocument, createFromFile)
 	}
 
 	private_data = php_dom_private_data_create();
-	php_dom_libxml_ns_mapper *ns_mapper = php_dom_ns_mapper_from_private(private_data);
 
 	xmlDocPtr lxml_doc;
 	lexbor_libxml2_bridge_status bridge_status = lexbor_libxml2_bridge_convert_document(
 		document,
 		&lxml_doc,
 		options & XML_PARSE_COMPACT,
 		!(options & DOM_HTML_NO_DEFAULT_NS),
-		ns_mapper
+		private_data
 	);
 	lexbor_libxml2_bridge_copy_observations(parser->tree, &ctx.observations);
 	if (UNEXPECTED(bridge_status != LEXBOR_LIBXML2_BRIDGE_STATUS_OK)) {
@@ -1206,7 +1205,7 @@ static zend_result dom_saveHTML_write_string(void *application_data, const char
 	return dom_saveHTML_write_string_len(application_data, buf, strlen(buf));
 }
 
-static zend_result dom_common_save(dom_output_ctx *output_ctx, const xmlDoc *docp, const xmlNode *node)
+static zend_result dom_common_save(dom_output_ctx *output_ctx, dom_object *intern, const xmlDoc *docp, const xmlNode *node)
 {
 	/* Initialize everything related to encoding & decoding */
 	const lxb_encoding_data_t *decoding_data = lxb_encoding_data(LXB_ENCODING_UTF_8);
@@ -1239,6 +1238,7 @@ static zend_result dom_common_save(dom_output_ctx *output_ctx, const xmlDoc *doc
 	ctx.write_string_len = dom_saveHTML_write_string_len;
 	ctx.write_string = dom_saveHTML_write_string;
 	ctx.application_data = output_ctx;
+	ctx.private_data = php_dom_get_private_data(intern);
 	if (UNEXPECTED(dom_html5_serialize_outer(&ctx, node) != SUCCESS)) {
 		return FAILURE;
 	}
@@ -1297,7 +1297,7 @@ PHP_METHOD(Dom_HTMLDocument, saveHtmlFile)
 	dom_output_ctx output_ctx;
 	output_ctx.output_data = stream;
 	output_ctx.write_output = dom_write_output_stream;
-	if (UNEXPECTED(dom_common_save(&output_ctx, docp, (const xmlNode *) docp) != SUCCESS)) {
+	if (UNEXPECTED(dom_common_save(&output_ctx, intern, docp, (const xmlNode *) docp) != SUCCESS)) {
 		php_stream_close(stream);
 		RETURN_FALSE;
 	}
@@ -1336,7 +1336,7 @@ PHP_METHOD(Dom_HTMLDocument, saveHtml)
 	output_ctx.output_data = &buf;
 	output_ctx.write_output = dom_write_output_smart_str;
 	/* Can't fail because dom_write_output_smart_str() can't fail. */
-	zend_result result = dom_common_save(&output_ctx, docp, node);
+	zend_result result = dom_common_save(&output_ctx, intern, docp, node);
 	ZEND_ASSERT(result == SUCCESS);
 
 	RETURN_STR(smart_str_extract(&buf));

ext/dom/inner_html_mixin.c

@@ -68,6 +68,7 @@ zend_result dom_element_inner_html_read(dom_object *obj, zval *retval)
 	if (context_document->type == XML_HTML_DOCUMENT_NODE) {
 		smart_str output = {0};
 		dom_html5_serialize_context ctx;
+		ctx.private_data = php_dom_get_private_data(obj);
 		ctx.application_data = &output;
 		ctx.write_string = dom_inner_html_write_string;
 		ctx.write_string_len = dom_inner_html_write_string_len;
@@ -86,11 +87,12 @@ zend_result dom_element_inner_html_read(dom_object *obj, zval *retval)
 			xmlCharEncodingHandlerPtr handler = xmlFindCharEncodingHandler("UTF-8");
 			xmlOutputBufferPtr out = xmlOutputBufferCreateIO(dom_write_smart_str, NULL, &str, handler);
 			if (EXPECTED(out != NULL)) {
+				php_dom_private_data *private_data = php_dom_get_private_data(obj);
 				/* Note: the innerHTML mixin sets the well-formed flag to true. */
 				xmlNodePtr child = node->children;
 				status = 0;
 				while (child != NULL && status == 0) {
-					status = dom_xml_serialize(ctxt, out, child, false, true);
+					status = dom_xml_serialize(ctxt, out, child, false, true, private_data);
 					child = child->next;
 				}
 				status |= xmlOutputBufferFlush(out);
@@ -205,7 +207,7 @@ static xmlNodePtr dom_html_fragment_parsing_algorithm(dom_object *obj, xmlNodePt
 	xmlNodePtr fragment = NULL;
 	if (node != NULL) {
 		/* node->last_child could be NULL, but that is allowed. */
-		lexbor_libxml2_bridge_status status = lexbor_libxml2_bridge_convert_fragment(node->last_child, context_node->doc, &fragment, true, true, php_dom_get_ns_mapper(obj));
+		lexbor_libxml2_bridge_status status = lexbor_libxml2_bridge_convert_fragment(node->last_child, context_node->doc, &fragment, true, true, php_dom_get_private_data(obj));
 		if (UNEXPECTED(status != LEXBOR_LIBXML2_BRIDGE_STATUS_OK)) {
 			php_dom_throw_error(INVALID_STATE_ERR, true);
 		}
@@ -349,6 +351,11 @@ zend_result dom_element_inner_html_write(dom_object *obj, zval *newval)
 		return FAILURE;
 	}
 
+	xmlNodePtr template_content = php_dom_retrieve_templated_content(php_dom_get_private_data(obj), context_node);
+	if (template_content != NULL) {
+		context_node = template_content;
+	}
+
 	/* We skip the steps involving the template element as context node since we don't do special handling for that. */
 	dom_remove_all_children(context_node);
 	return php_dom_pre_insert(obj->document, fragment, context_node, NULL) ? SUCCESS : FAILURE;

ext/dom/internal_helpers.h

@@ -89,4 +89,12 @@ static zend_always_inline bool dom_is_document_cache_modified_since_parsing(php_
 	return !doc_ptr || doc_ptr->cache_tag.modification_nr > dom_minimum_modification_nr_since_parsing(doc_ptr);
 }
 
+static zend_always_inline zend_long dom_mangle_pointer_for_key(const void *ptr)
+{
+	zend_ulong value = (zend_ulong) (uintptr_t) ptr;
+	/* Rotate 3/4 bits for better hash distribution because the low 3/4 bits are normally 0. */
+	const size_t rol_amount = (SIZEOF_ZEND_LONG == 8) ? 4 : 3;
+	return (value >> rol_amount) | (value << (sizeof(value) * 8 - rol_amount));
+}
+
 #endif