feat: support document as startNodes in browsingContext.locateNodes

src/bidiMapper/modules/context/BrowsingContextImpl.ts

@@ -1105,12 +1105,12 @@ export class BrowsingContextImpl {
             (
               xPathSelector: string,
               maxNodeCount: number,
-              ...startNodes: HTMLElement[]
+              ...startNodes: Node[]
             ) => {
               // https://w3c.github.io/webdriver-bidi/#locate-nodes-using-xpath
               const evaluator = new XPathEvaluator();
               const expression = evaluator.createExpression(xPathSelector);
-              const locateNodesUsingXpath = (element: HTMLElement) => {
+              const locateNodesUsingXpath = (element: Node) => {
                 const xPathResult = expression.evaluate(
                   element,
                   XPathResult.ORDERED_NODE_SNAPSHOT_TYPE
@@ -1157,16 +1157,34 @@ export class BrowsingContextImpl {
               ignoreCase: boolean,
               maxNodeCount: number,
               maxDepth: number,
-              ...startNodes: HTMLElement[]
+              ...startNodes: Node[]
             ) => {
               const searchText = ignoreCase
                 ? innerTextSelector.toUpperCase()
                 : innerTextSelector;
-              const locateNodesUsingInnerText = (
-                element: HTMLElement,
+              const locateNodesUsingInnerText: (
+                node: Node,
                 currentMaxDepth: number
-              ) => {
+              ) => HTMLElement[] = (node: Node, currentMaxDepth: number) => {
                 const returnedNodes: HTMLElement[] = [];
+                if (
+                  node instanceof DocumentFragment ||
+                  node instanceof Document
+                ) {
+                  const children = [...node.children];
+                  children.forEach((child) =>
+                    returnedNodes.push(
+                      ...locateNodesUsingInnerText(child, currentMaxDepth - 1)
+                    )
+                  );
+                  return returnedNodes;
+                }
+
+                if (!(node instanceof HTMLElement)) {
+                  return [];
+                }
+
+                const element = node;
                 const nodeInnerText = ignoreCase
                   ? element.innerText?.toUpperCase()
                   : element.innerText;
@@ -1213,9 +1231,8 @@ export class BrowsingContextImpl {
                 // TODO: stop search early if `maxNodeCount` is reached.
                 return returnedNodes;
               };
-              // TODO: add maxDepth.
               // TODO: stop search early if `maxNodeCount` is reached.
-              startNodes = startNodes.length > 0 ? startNodes : [document.body];
+              startNodes = startNodes.length > 0 ? startNodes : [document];
               const returnedNodes = startNodes
                 .map((startNode) =>
                   // TODO: stop search early if `maxNodeCount` is reached.

tests/browsing_context/test_locate_nodes.py

@@ -39,20 +39,41 @@
         }
     },
 ])
+@pytest.mark.parametrize('start_node_expression',
+                         [None, 'document', 'document.body'])
 @pytest.mark.asyncio
-async def test_locate_nodes_locator_found(websocket, context_id, html,
-                                          locator):
+async def test_locate_nodes_locator_no_start_nodes_found(
+        websocket, context_id, html, locator, start_node_expression):
     await goto_url(
         websocket, context_id,
         html(
             '<div data-class="one" aria-label="test" role="button">foobarBARbaz</div><div data-class="two" aria-label="test" role="button">foobarBAR<span>baz</span></div>'
         ))
+
+    start_nodes = None
+    if start_node_expression:
+        resp = await execute_command(
+            websocket, {
+                'method': 'script.evaluate',
+                'params': {
+                    'expression': start_node_expression,
+                    'target': {
+                        'context': context_id
+                    },
+                    'awaitPromise': False
+                }
+            })
+        start_nodes = [{'sharedId': resp['result']['sharedId']}]
+
     resp = await execute_command(
         websocket, {
             'method': 'browsingContext.locateNodes',
             'params': {
                 'context': context_id,
-                'locator': locator
+                'locator': locator,
+                **({
+                    'startNodes': start_nodes
+                } if start_nodes else {})
             }
         })