Method parse_my_html

webarena/browser_env/processors.py:358–438

Parse the html tree into a string text

(dom_tree: DOMTree)

    @staticmethod
    def parse_my_html(dom_tree: DOMTree) -> tuple[str, str, dict[str, Any], Any]:
        """Parse the html tree into a string text"""

        obs_nodes_info = {}
        nodeid_to_cursor = {
            node["nodeId"]: idx for idx, node in enumerate(dom_tree)
        }

        def dfs(node_cursor: int, depth: int) -> tuple[str, list[str]]:
            tree_str, labeled_elems = '', []
            node = dom_tree[node_cursor]
            valid_node = True
            pure_text = False
            try:
                if node['nodeName'] == '#text':
                    node['nodeName'] = 'text'
                
                node_str = f"<{node['nodeName']}"
                if node["attributes"]:
                    node_str += f" {node['attributes']}"
                node_str += f" backend-id=\"bid-{node['backendNodeId']}\"> {node['nodeValue']}"
                
                # if node['nodeName'] == '#text':
                #     pure_text = True
                #     node_str = node['nodeValue']
                    
                valid_node = bool(node["attributes"] or node["nodeValue"] or pure_text)
                    
                if valid_node:                    
                    node_html = lxml.html.fromstring(node_str)
                    label = node_html.attrib.get('data-testid', '')
                    if len(label) > 0:
                        labeled_elems.append(node["backendNodeId"])
                    obs_nodes_info[str(node_cursor)] = {
                        "backend_id": node["backendNodeId"],
                        "union_bound": node["union_bound"],
                        "text": node['nodeValue'],
                        "label": label,
                    }
                    tree_str += f"{node_str}"

            except Exception as e:
                valid_node = False

            for child_ids in node["childIds"]:
                child_cursor = nodeid_to_cursor[child_ids]
                child_depth = depth + 1 if valid_node else depth
                child_str, elems = dfs(child_cursor, child_depth)
                tree_str += child_str
                labeled_elems.extend(elems)
            
            if valid_node and not pure_text:
                tree_str += f"</{node['nodeName']}>"

            return tree_str, labeled_elems

        html, labeled_elems = dfs(0, 0)