lzy/sci-gui-agent-benchmark
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

insure no exception (if failed, return 0) and change 'load' to 'networkidle'

Browse Source
This commit is contained in:
Jason Lee
2024-02-26 22:07:08 +08:00
parent 889c22975e
commit 0edbcf404d
5 changed files with 47 additions and 14 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
desktop_env/evaluators/getters/chrome.py
View File

@@ -444,7 +444,7 @@ def get_open_tabs_info(env, config: Dict[str, str]):
for page in context.pages:
try:
# Wait for the page to finish loading, this prevents the "execution context was destroyed" issue
page.wait_for_load_state('load') # Wait for the 'load' event to complete
page.wait_for_load_state('networkidle') # Wait for the 'load' event to complete
title = page.title()
url = page.url
tabs_info.append({'title': title, 'url': url})
@@ -486,12 +486,17 @@ def get_active_url_from_accessTree(env, config):
# first, use accessibility API to get the active tab URL
at: _Element = lxml.etree.fromstring(accessibility_tree)
arch = platform.machine()
if "arm" in arch:
selector = CSSSelector(", application[name=Chromium] entry[name=Address\\ and\\ search\\ bar]",
namespaces=_accessibility_ns_map)
else:
selector = CSSSelector("application[name=Google\\ Chrome] entry[name=Address\\ and\\ search\\ bar]",
namespaces=_accessibility_ns_map)
print("Your architecture is: {}".format(arch))
try:
if "arm" in arch:
selector = CSSSelector("application[name=Chromium] entry[name=Address\\ and\\ search\\ bar]",
namespaces=_accessibility_ns_map)
else:
selector = CSSSelector("application[name=Google\\ Chrome] entry[name=Address\\ and\\ search\\ bar]",
namespaces=_accessibility_ns_map)
except:
logger.error("Failed to parse the selector for active tab URL")
return None
elements: List[_Element] = selector(at)
# if "xpath" in config:
# elements: List[_Element] = at.xpath(config["xpath"], namespaces=_accessibility_ns_map)
@@ -519,6 +524,9 @@ def get_active_tab_info(env, config: Dict[str, str]):
}
"""
active_tab_url = get_active_url_from_accessTree(env, config)
if active_tab_url is None:
logger.error("Failed to get the url of active tab")
return None
host = env.vm_ip
port = 9222 # fixme: this port is hard-coded, need to be changed from config file
@@ -533,7 +541,11 @@ def get_active_tab_info(env, config: Dict[str, str]):
active_tab_info = {}
# go to the target URL page
page = browser.new_page()
page.goto(active_tab_url)
try:
page.goto(active_tab_url)
except:
logger.error("Failed to go to the target URL page")
return None
page.wait_for_load_state('load') # Wait for the 'load' event to complete
active_tab_info = {
'title': page.title(),
@@ -972,6 +984,8 @@ def get_active_tab_html_parse(env, config: Dict[str, Any]):
}
"""
active_tab_url = get_active_url_from_accessTree(env, config)
if not isinstance(active_tab_url, str):
return None
host = env.vm_ip
port = 9222 # fixme: this port is hard-coded, need to be changed from config file
@@ -1002,9 +1016,11 @@ def get_active_tab_html_parse(env, config: Dict[str, Any]):
target_page = None
for context in browser.contexts:
for page in context.pages:
page.wait_for_load_state("load")
page.wait_for_load_state("networkidle")
if page.url == active_tab_url:
target_page = page
print("tartget page url: ", target_page.url)
print("tartget page title: ", target_page.title())
break
if target_page is None:
return {}