Skip to content

Lightpanda: the headless browser designed for AI and automation

cleanup: Mechanical move various frame.parse* method

Karl Seguin705bc78b7276
diff --git a/src/browser/Frame.zig b/src/browser/Frame.zig
index 764e075e0..ce78f7588 100644
--- a/src/browser/Frame.zig
+++ b/src/browser/Frame.zig
@@ -73,6 +73,7 @@ const milliTimestamp = @import("../datetime.zig").milliTimestamp;
const GlobalEventHandlersLookup = @import("webapi/global_event_handlers.zig").Lookup;
+pub const parse = @import("frame/parse.zig");
pub const preload = @import("frame/preload.zig");
pub const observers = @import("frame/observers.zig");
pub const user_input = @import("frame/user_input.zig");
@@ -2867,113 +2868,6 @@ pub fn updateRangesForNodeRemoval(self: *Frame, parent: *Node, child: *Node, chi
}
}
-// TODO: optimize and cleanup, this is called a lot (e.g., innerHTML = '')
-pub fn parseHtmlAsChildren(self: *Frame, node: *Node, html: []const u8) !void {
- return self.parseHtmlAsChildrenInner(node, html, .{});
-}
-
-// setHTMLUnsafe variant: parse a fragment that may contain declarative shadow node
-pub fn parseHtmlUnsafeAsChildren(self: *Frame, node: *Node, html: []const u8) !void {
- return self.parseHtmlAsChildrenInner(node, html, .{ .allow_declarative_shadow = true });
-}
-
-// Range.createContextualFragment variant: unlike innerHTML et al., its scripts
-// are run when the fragment is inserted into a document.
-pub fn parseContextualFragment(self: *Frame, node: *Node, html: []const u8) !void {
- return self.parseHtmlAsChildrenInner(node, html, .{ .scripts_runnable = true });
-}
-
-const FragmentParseOpts = struct {
- scripts_runnable: bool = false,
- allow_declarative_shadow: bool = false,
-};
-
-fn parseHtmlAsChildrenInner(self: *Frame, node: *Node, html: []const u8, opts: FragmentParseOpts) !void {
- const previous_parse_mode = self._parse_mode;
- self._parse_mode = .fragment;
- defer self._parse_mode = previous_parse_mode;
-
- // The html5ever wrapper-unwrap below rebinds children without going
- // through the insertion path, so recompute slot assignments for any
- // shadow tree this fragment landed in (idempotent; signals only on diff).
- defer if (self._element_shadow_roots.count() != 0) {
- const root = node.getRootNode(.{});
- if (root.is(ShadowRoot) != null) {
- slotting.assignSlottablesForTree(root, self);
- }
- if (node.is(Element)) |el| {
- if (self._element_shadow_roots.get(el)) |shadow_root| {
- slotting.assignSlottablesForTree(shadow_root.asNode(), self);
- }
- }
- };
-
- const previous_scripts_runnable = self._fragment_scripts_runnable;
- self._fragment_scripts_runnable = opts.scripts_runnable;
- defer self._fragment_scripts_runnable = previous_scripts_runnable;
-
- var parser = Parser.init(self.call_arena, node, self, .{ .allow_declarative_shadow = opts.allow_declarative_shadow });
- parser.parseFragment(html);
- if (parser.terminated) {
- return error.ExecutionTerminated;
- }
-
- // html5ever wraps fragment output in an <html> element; unwrap so its
- // children land directly on `node`. See https://github.com/servo/html5ever/issues/583.
- // Because of custom element callbacks, the structure might not be what
- // we expect, and nodes might be altogether removed. We deal with this in a
- // few different places, but always the same way: leave it as-is.
- const children = node._children orelse return;
- const first = Node.linkToNode(children.first.?);
- if (first.is(Element.Html.Html) == null) {
- return;
- }
- node._children = first._children;
-
- // No mutation records for the unwrapped children either; see the comment
- // about fragment parses in _insertNodeRelative.
- var it = node.childrenIterator();
- while (it.next()) |child| {
- child._parent = node;
- }
-}
-
-// Build a detached XMLDocument from `xml` (DOMParser.parseFromString and
-// XMLHttpRequest.responseXML). Returns null when the input isn't well-formed
-// XML. The parse borrows the fragment parse-mode so frame-side hooks triggered
-// from `Build.created` / `nodeIsReady` (external stylesheet fetches, script
-// execution, mutation-observer fan-out, default-script injection) treat the
-// parsed nodes as detached and skip side effects on the live document.
-pub fn parseXmlDocument(self: *Frame, xml: []const u8) !?*Document.XMLDocument {
- const arena = try self.getArena(.medium, "Frame.parseXmlDocument");
- defer self.releaseArena(arena);
-
- const previous_parse_mode = self._parse_mode;
- self._parse_mode = .fragment;
- defer self._parse_mode = previous_parse_mode;
-
- const doc = try self._factory.document(Document.XMLDocument{ ._proto = undefined });
- const doc_node = doc.asNode();
- var parser = Parser.init(arena, doc_node, self, .{});
- parser.parseXML(xml);
- if (parser.terminated) {
- return error.ExecutionTerminated;
- }
-
- if (parser.err != null or doc_node.firstChild() == null) {
- return null;
- }
-
- // If first node is a `ProcessingInstruction` (e.g. the <?xml?>
- // declaration), skip it.
- const first_child = doc_node.firstChild().?;
- if (first_child.getNodeType() == 7) {
- _ = try doc_node.removeChild(first_child, self);
- }
-
- return doc;
-}
-
// Runs the "ready" work for an inserted node and, when it's an element with
// children, for its descendants in tree order: appending a subtree
// containing scripts must execute them all, after the whole insertion.
diff --git a/src/browser/forms.zig b/src/browser/forms.zig
index 18d068734..ce6de36b1 100644
--- a/src/browser/forms.zig
+++ b/src/browser/forms.zig
@@ -282,7 +282,7 @@ fn testForms(html: []const u8) ![]FormInfo {
const doc = frame.window._document;
const div = try doc.createElement("div", null, frame);
- try frame.parseHtmlAsChildren(div.asNode(), html);
+ try Frame.parse.htmlAsChildren(frame, div.asNode(), html);
return collectForms(frame.call_arena, div.asNode(), frame);
}
diff --git a/src/browser/frame/parse.zig b/src/browser/frame/parse.zig
new file mode 100644
index 000000000..526fc872d
--- /dev/null
+++ b/src/browser/frame/parse.zig
@@ -0,0 +1,129 @@
+// Copyright (C) 2023-2026 Lightpanda (Selecy SAS)
+//
+// Francis Bouvier <[email protected]>
+// Pierre Tachoire <[email protected]>
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as
+// published by the Free Software Foundation, either version 3 of the
+// License, or (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program. If not, see <https://www.gnu.org/licenses/>.
+
+const Frame = @import("../Frame.zig");
+const Parser = @import("../parser/Parser.zig");
+
+const Node = @import("../webapi/Node.zig");
+const Element = @import("../webapi/Element.zig");
+const Document = @import("../webapi/Document.zig");
+const ShadowRoot = @import("../webapi/ShadowRoot.zig");
+const slotting = @import("../webapi/element/slotting.zig");
+
+pub fn htmlAsChildren(frame: *Frame, node: *Node, html: []const u8) !void {
+ return htmlAsChildrenInner(frame, node, html, .{});
+}
+
+// setHTMLUnsafe variant: parse a fragment that may contain declarative shadow node
+pub fn htmlUnsafeAsChildren(frame: *Frame, node: *Node, html: []const u8) !void {
+ return htmlAsChildrenInner(frame, node, html, .{ .allow_declarative_shadow = true });
+}
+
+// Range.createContextualFragment variant: unlike innerHTML et al., its scripts
+// are run when the fragment is inserted into a document.
+pub fn contextualFragment(frame: *Frame, node: *Node, html: []const u8) !void {
+ return htmlAsChildrenInner(frame, node, html, .{ .scripts_runnable = true });
+}
+
+const FragmentParseOpts = struct {
+ scripts_runnable: bool = false,
+ allow_declarative_shadow: bool = false,
+};
+
+fn htmlAsChildrenInner(frame: *Frame, node: *Node, html: []const u8, opts: FragmentParseOpts) !void {
+ const previous_parse_mode = frame._parse_mode;
+ frame._parse_mode = .fragment;
+ defer frame._parse_mode = previous_parse_mode;
+
+ // The html5ever wrapper-unwrap below rebinds children without going
+ // through the insertion path, so recompute slot assignments for any
+ // shadow tree this fragment landed in (idempotent; signals only on diff).
+ defer if (frame._element_shadow_roots.count() != 0) {
+ const root = node.getRootNode(.{});
+ if (root.is(ShadowRoot) != null) {
+ slotting.assignSlottablesForTree(root, frame);
+ }
+ if (node.is(Element)) |el| {
+ if (frame._element_shadow_roots.get(el)) |shadow_root| {
+ slotting.assignSlottablesForTree(shadow_root.asNode(), frame);
+ }
+ }
+ };
+
+ const previous_scripts_runnable = frame._fragment_scripts_runnable;
+ frame._fragment_scripts_runnable = opts.scripts_runnable;
+ defer frame._fragment_scripts_runnable = previous_scripts_runnable;
+
+ var parser = Parser.init(frame.call_arena, node, frame, .{ .allow_declarative_shadow = opts.allow_declarative_shadow });
+ parser.parseFragment(html);
+ if (parser.terminated) {
+ return error.ExecutionTerminated;
+ }
+
+ // html5ever wraps fragment output in an <html> element; unwrap so its
+ // children land directly on `node`. See https://github.com/servo/html5ever/issues/583.
+ // Because of custom element callbacks, the structure might not be what
+ // we expect, and nodes might be altogether removed. We deal with this in a
+ // few different places, but always the same way: leave it as-is.
+ const children = node._children orelse return;
+ const first = Node.linkToNode(children.first.?);
+ if (first.is(Element.Html.Html) == null) {
+ return;
+ }
+ node._children = first._children;
+
+ // No mutation records for the unwrapped children either; see the comment
+ // about fragment parses in _insertNodeRelative.
+ var it = node.childrenIterator();
+ while (it.next()) |child| {
+ child._parent = node;
+ }
+}
+
+// Build a detached XMLDocument from `xml` (DOMParser.parseFromString and
+// XMLHttpRequest.responseXML). Returns null when the input isn't well-formed
+// XML.
+pub fn xmlDocument(frame: *Frame, xml: []const u8) !?*Document.XMLDocument {
+ const arena = try frame.getArena(.medium, "parse.xmlDocument");
+ defer frame.releaseArena(arena);
+
+ const previous_parse_mode = frame._parse_mode;
+ frame._parse_mode = .fragment;
+ defer frame._parse_mode = previous_parse_mode;
+
+ const doc = try frame._factory.document(Document.XMLDocument{ ._proto = undefined });
+ const doc_node = doc.asNode();
+ var parser = Parser.init(arena, doc_node, frame, .{});
+ parser.parseXML(xml);
+ if (parser.terminated) {
+ return error.ExecutionTerminated;
+ }
+
+ if (parser.err != null or doc_node.firstChild() == null) {
+ return null;
+ }
+
+ // If first node is a `ProcessingInstruction` (e.g. the <?xml?>
+ // declaration), skip it.
+ const first_child = doc_node.firstChild().?;
+ if (first_child.getNodeType() == 7) {
+ _ = try doc_node.removeChild(first_child, frame);
+ }
+
+ return doc;
+}
diff --git a/src/browser/interactive.zig b/src/browser/interactive.zig
index 8c9c51e31..d6f092842 100644
--- a/src/browser/interactive.zig
+++ b/src/browser/interactive.zig
@@ -501,7 +501,7 @@ fn testInteractive(html: []const u8) ![]InteractiveElement {
const doc = frame.window._document;
const div = try doc.createElement("div", null, frame);
- try frame.parseHtmlAsChildren(div.asNode(), html);
+ try Frame.parse.htmlAsChildren(frame, div.asNode(), html);
return collectInteractiveElements(div.asNode(), frame.call_arena, frame);
}
diff --git a/src/browser/links.zig b/src/browser/links.zig
index 45695923a..95624ce0f 100644
--- a/src/browser/links.zig
+++ b/src/browser/links.zig
@@ -94,7 +94,7 @@ fn testLinks(html: []const u8) ![]Link {
const doc = frame.window._document;
const div = try doc.createElement("div", null, frame);
- try frame.parseHtmlAsChildren(div.asNode(), html);
+ try Frame.parse.htmlAsChildren(frame, div.asNode(), html);
return collectLinks(frame.call_arena, div.asNode(), frame);
}
diff --git a/src/browser/markdown.zig b/src/browser/markdown.zig
index ae702e2d4..86cdd45f4 100644
--- a/src/browser/markdown.zig
+++ b/src/browser/markdown.zig
@@ -599,7 +599,7 @@ fn testMarkdownHTML(html: []const u8, expected: []const u8) !void {
const doc = frame.window._document;
const div = try doc.createElement("div", null, frame);
- try frame.parseHtmlAsChildren(div.asNode(), html);
+ try Frame.parse.htmlAsChildren(frame, div.asNode(), html);
var aw: std.Io.Writer.Allocating = .init(testing.allocator);
defer aw.deinit();
@@ -800,7 +800,7 @@ test "browser.markdown: resolve links" {
const doc = frame.window._document;
const div = try doc.createElement("div", null, frame);
- try frame.parseHtmlAsChildren(div.asNode(),
+ try Frame.parse.htmlAsChildren(frame, div.asNode(),
\\<a href="b">Link</a>
\\<img src="../c.png" alt="Img">
\\<a href="/my page">Space</a>
@@ -839,7 +839,7 @@ test "browser.markdown: max_bytes leaves output untouched when under cap" {
const doc = frame.window._document;
const div = try doc.createElement("div", null, frame);
- try frame.parseHtmlAsChildren(div.asNode(), "<p>Short</p>");
+ try Frame.parse.htmlAsChildren(frame, div.asNode(), "<p>Short</p>");
var aw: std.Io.Writer.Allocating = .init(testing.allocator);
defer aw.deinit();
@@ -855,7 +855,7 @@ test "browser.markdown: max_bytes truncates with marker" {
const doc = frame.window._document;
const div = try doc.createElement("div", null, frame);
- try frame.parseHtmlAsChildren(div.asNode(), "<p>" ++ ("AAAA " ** 100) ++ "</p>");
+ try Frame.parse.htmlAsChildren(frame, div.asNode(), "<p>" ++ ("AAAA " ** 100) ++ "</p>");
var aw: std.Io.Writer.Allocating = .init(testing.allocator);
defer aw.deinit();
@@ -878,11 +878,11 @@ fn testMarkdownShadow(light: []const u8, shadow: []const u8, expected: []const u
const host = try doc.createElement("div", null, frame);
if (light.len > 0) {
- try frame.parseHtmlAsChildren(host.asNode(), light);
+ try Frame.parse.htmlAsChildren(frame, host.asNode(), light);
}
const sr = try host.attachShadow(.{ .mode = .open }, frame);
- try frame.parseHtmlAsChildren(sr.asNode(), shadow);
+ try Frame.parse.htmlAsChildren(frame, sr.asNode(), shadow);
var aw: std.Io.Writer.Allocating = .init(testing.allocator);
defer aw.deinit();
diff --git a/src/browser/structured_data.zig b/src/browser/structured_data.zig
index 8e5aac614..0090f81eb 100644
--- a/src/browser/structured_data.zig
+++ b/src/browser/structured_data.zig
@@ -494,7 +494,7 @@ fn testStructuredData(html: []const u8) !StructuredData {
const doc = frame.window._document;
const div = try doc.createElement("div", null, frame);
- try frame.parseHtmlAsChildren(div.asNode(), html);
+ try Frame.parse.htmlAsChildren(frame, div.asNode(), html);
return collectStructuredData(div.asNode(), frame.call_arena, frame);
}
@@ -714,7 +714,7 @@ test "structured_data: link headers from response" {
const doc = frame.window._document;
const div = try doc.createElement("div", null, frame);
- try frame.parseHtmlAsChildren(div.asNode(), "<title>Example</title>");
+ try Frame.parse.htmlAsChildren(frame, div.asNode(), "<title>Example</title>");
const data = try collectStructuredData(div.asNode(), frame.call_arena, frame);
diff --git a/src/browser/tests/custom_elements/parser_wrapper_removed.html b/src/browser/tests/custom_elements/parser_wrapper_removed.html
index 77c637eb5..c2ca9f62b 100644
--- a/src/browser/tests/custom_elements/parser_wrapper_removed.html
+++ b/src/browser/tests/custom_elements/parser_wrapper_removed.html
@@ -2,7 +2,7 @@
<body>
<script src="../testing.js"></script>
-<!-- Frame.parseHtmlAsChildren extracts children from the temporary
+<!-- Frame.parse.htmlAsChildren extracts children from the temporary
<html> element the fragment parser leaves at the top of the parse
result. A custom element's connectedCallback fires synchronously
during that parse, so it can reach two levels up (this.parentNode is
diff --git a/src/browser/webapi/DOMParser.zig b/src/browser/webapi/DOMParser.zig
index f30467608..3a8088bd1 100644
--- a/src/browser/webapi/DOMParser.zig
+++ b/src/browser/webapi/DOMParser.zig
@@ -55,7 +55,7 @@ pub fn parseFromString(
defer frame.releaseArena(arena);
// DOMParser builds a detached Document. Borrow the same fragment
- // parse-mode that `parseHtmlAsChildren` uses so frame-side hooks
+ // parse-mode that `Frame.parse` uses so frame-side hooks
// triggered from `Build.created` / `nodeIsReady` (external
// stylesheet fetches, script execution, mutation-observer fan-out,
// default-script injection) treat the parsed nodes as detached and
@@ -89,9 +89,9 @@ pub fn parseFromString(
return doc.asDocument();
},
else => {
- const doc = (try frame.parseXmlDocument(html)) orelse blk: {
+ const doc = (try Frame.parse.xmlDocument(frame, html)) orelse blk: {
// Return a document with a <parsererror> element per spec.
- break :blk (try frame.parseXmlDocument("<parsererror xmlns=\"http://www.mozilla.org/newlayout/xml/parsererror.xml\">error</parsererror>")).?;
+ break :blk (try Frame.parse.xmlDocument(frame, "<parsererror xmlns=\"http://www.mozilla.org/newlayout/xml/parsererror.xml\">error</parsererror>")).?;
};
return doc.asDocument();
},
diff --git a/src/browser/webapi/Element.zig b/src/browser/webapi/Element.zig
index e44c1f066..a81b82b66 100644
--- a/src/browser/webapi/Element.zig
+++ b/src/browser/webapi/Element.zig
@@ -484,7 +484,7 @@ pub fn setOuterHTML(self: *Element, html: []const u8, frame: *Frame) !void {
var fragment: ?*Node = null;
if (html.len > 0) {
const frag = (try Node.DocumentFragment.init(frame)).asNode();
- try frame.parseHtmlAsChildren(frag, html);
+ try Frame.parse.htmlAsChildren(frame, frag, html);
fragment = frag;
}
diff --git a/src/browser/webapi/HTMLDocument.zig b/src/browser/webapi/HTMLDocument.zig
index be47ff982..08befb559 100644
--- a/src/browser/webapi/HTMLDocument.zig
+++ b/src/browser/webapi/HTMLDocument.zig
@@ -71,7 +71,7 @@ pub fn setBody(self: *HTMLDocument, html: []const u8, frame: *Frame) !void {
// parsing strips any <html>/<body>/<head> wrappers the author included.
const new_body_node = try Frame.node_factory.createElementNS(frame, .html, "body", null);
if (html.len > 0) {
- try frame.parseHtmlAsChildren(new_body_node, html);
+ try Frame.parse.htmlAsChildren(frame, new_body_node, html);
}
const document_node = document_element.asNode();
diff --git a/src/browser/webapi/Node.zig b/src/browser/webapi/Node.zig
index 898e28f57..0bd37dbcc 100644
--- a/src/browser/webapi/Node.zig
+++ b/src/browser/webapi/Node.zig
@@ -1557,9 +1557,9 @@ pub fn setHTML(self: *Node, html: []const u8, allow_declarative_shadow: bool, fr
if (html.len > 0) {
if (allow_declarative_shadow) {
- try frame.parseHtmlUnsafeAsChildren(self, html);
+ try Frame.parse.htmlUnsafeAsChildren(frame, self, html);
} else {
- try frame.parseHtmlAsChildren(self, html);
+ try Frame.parse.htmlAsChildren(frame, self, html);
}
}
diff --git a/src/browser/webapi/Range.zig b/src/browser/webapi/Range.zig
index cc443566b..9339fde76 100644
--- a/src/browser/webapi/Range.zig
+++ b/src/browser/webapi/Range.zig
@@ -703,7 +703,7 @@ pub fn createContextualFragment(self: *const Range, html: []const u8, frame: *Fr
else
try Frame.node_factory.createElementNS(frame, .html, "div", null);
- try frame.parseContextualFragment(temp_node, html);
+ try Frame.parse.contextualFragment(frame, temp_node, html);
// Move all parsed children to the fragment
// Keep removing first child until temp element is empty
diff --git a/src/browser/webapi/element/Html.zig b/src/browser/webapi/element/Html.zig
index c2776a1f0..699e98e13 100644
--- a/src/browser/webapi/element/Html.zig
+++ b/src/browser/webapi/element/Html.zig
@@ -290,7 +290,7 @@ pub fn insertAdjacentHTML(
) !void {
const DocumentFragment = @import("../DocumentFragment.zig");
const fragment = (try DocumentFragment.init(frame)).asNode();
- try frame.parseHtmlAsChildren(fragment, html);
+ try Frame.parse.htmlAsChildren(frame, fragment, html);
const target_node, const prev_node = try self.asNode().findAdjacentNodes(position, .html);
diff --git a/src/browser/webapi/net/XMLHttpRequest.zig b/src/browser/webapi/net/XMLHttpRequest.zig
index d59c6cc26..52b8539b1 100644
--- a/src/browser/webapi/net/XMLHttpRequest.zig
+++ b/src/browser/webapi/net/XMLHttpRequest.zig
@@ -26,6 +26,7 @@ const Transfer = @import("../../../network/HttpClient.zig").Transfer;
const URL = @import("../../URL.zig");
const Mime = @import("../../Mime.zig");
const Page = @import("../../Page.zig");
+const Frame = @import("../../Frame.zig");
const Node = @import("../Node.zig");
const Event = @import("../Event.zig");
@@ -438,14 +439,14 @@ pub fn getResponse(self: *XMLHttpRequest, exec: *const Execution) !?Response {
.frame => |frame| {
const final: Mime = self._override_mime orelse self._response_mime orelse .{ .content_type = .text_xml };
if (final.isXML()) {
- const document = (try frame.parseXmlDocument(data)) orelse return null;
+ const document = (try Frame.parse.xmlDocument(frame, data)) orelse return null;
break :blk .{ .document = document.asDocument() };
}
if (!final.isHTML()) {
return null;
}
const document = try exec._factory.node(Node.Document{ ._proto = undefined, ._type = .generic });
- try frame.parseHtmlAsChildren(document.asNode(), data);
+ try Frame.parse.htmlAsChildren(frame, document.asNode(), data);
break :blk .{ .document = document };
},
.worker => return error.NotSupportedInWorker,
@@ -487,7 +488,7 @@ pub fn getResponseXML(self: *XMLHttpRequest, exec: *const Execution) !?*Node.Doc
switch (exec.js.global) {
.frame => |frame| {
- const xml_document = (try frame.parseXmlDocument(self._response_data.items)) orelse return null;
+ const xml_document = (try Frame.parse.xmlDocument(frame, self._response_data.items)) orelse return null;
const document = xml_document.asDocument();
self._response_xml = document;
return document;