public/browser
1
0
Fork 0
mirror of https://github.com/lightpanda-io/browser.git synced 2026-03-22 04:34:44 +00:00
Code Issues Actions 5 Packages Projects Releases Wiki Activity

SemanticTree: filter computed names for generic containers

Browse Source 
This prevents token bloat in JSON/text dumps and ensures that StaticText leaf nodes are not incorrectly pruned when structural containers (like none, table) hoist their text.
This commit is contained in:
Adrià Arrufat
2026-03-11 10:22:40 +09:00
parent a6ccc72d15
commit 4f262e5bed
1 changed files with 19 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files

26
src/SemanticTree.zig
View File

@@ -146,7 +146,21 @@ fn walk(self: @This(), node: *Node, xpath_buffer: *std.ArrayList(u8), parent_nam
try appendXPathSegment(node, xpath_buffer.writer(self.arena), index); try appendXPathSegment(node, xpath_buffer.writer(self.arena), index);
const xpath = xpath_buffer.items; const xpath = xpath_buffer.items;
const name = try axn.getName(self.page, self.arena); var name = try axn.getName(self.page, self.arena);
const has_explicit_label = if (node.is(Element)) |el|
el.getAttributeSafe(.wrap("aria-label")) != null or el.getAttributeSafe(.wrap("title")) != null
else
false;
const structural = isStructuralRole(role);
// Filter out computed concatenated names for generic containers without explicit labels.
// This prevents token bloat and ensures their StaticText children aren't incorrectly pruned.
// We ignore interactivity because a generic wrapper with an event listener still shouldn't hoist all text.
if (name != null and structural and !has_explicit_label) {
name = null;
}
var data = NodeData{ var data = NodeData{
.id = cdp_node.id, .id = cdp_node.id,
@@ -162,12 +176,6 @@ fn walk(self: @This(), node: *Node, xpath_buffer: *std.ArrayList(u8), parent_nam
var should_visit = true; var should_visit = true;
if (self.prune) { if (self.prune) {
const structural = isStructuralRole(role);
const has_explicit_label = if (node.is(Element)) |el|
el.getAttributeSafe(.wrap("aria-label")) != null or el.getAttributeSafe(.wrap("title")) != null
else
false;
if (structural and !is_interactive and !has_explicit_label) { if (structural and !is_interactive and !has_explicit_label) {
should_visit = false; should_visit = false;
} }
@@ -368,6 +376,10 @@ fn isStructuralRole(role: []const u8) bool {
std.mem.eql(u8, role, "main") or std.mem.eql(u8, role, "main") or
std.mem.eql(u8, role, "list") or std.mem.eql(u8, role, "list") or
std.mem.eql(u8, role, "listitem") or std.mem.eql(u8, role, "listitem") or
std.mem.eql(u8, role, "table") or
std.mem.eql(u8, role, "rowgroup") or
std.mem.eql(u8, role, "row") or
std.mem.eql(u8, role, "cell") or
std.mem.eql(u8, role, "region"); std.mem.eql(u8, role, "region");
// zig fmt: on // zig fmt: on
} }