def crawl(self, url=None, track=None, **kwargs):
if url is None and kwargs.get('callback'):
url = dataurl.encode(utils.text(kwargs.get('callback')))
project_data = self.processor.project_manager.get(self.project_name)
assert project_data, "can't find project: %s" % self.project_name
instance = project_data['instance']
instance._reset()
task = instance.crawl(url, **kwargs)
if isinstance(task, list):
task = task[0]
task['track'] = track
result = self.fetcher.fetch(task)
self.processor.on_task(task, result)
status = None
while not self.status_queue.empty():
status = self.status_queue.get()
newtasks = []
while not self.newtask_queue.empty():
newtasks = self.newtask_queue.get()
result = None
while not self.result_queue.empty():
_, result = self.result_queue.get()
return status, newtasks, result
def crawl(self, url=None, track=None, **kwargs):
if url is None and kwargs.get('callback'):
url = dataurl.encode(utils.text(kwargs.get('callback')))
project_data = self.processor.project_manager.get(self.project_name)
assert project_data, "can't find project: %s" % self.project_name
instance = project_data['instance']
instance._reset()
task = instance.crawl(url, **kwargs)
if isinstance(task, list):
task = task[0]
task['track'] = track
result = self.fetcher.fetch(task)
self.processor.on_task(task, result)
status = None
while not self.status_queue.empty():
status = self.status_queue.get()
newtasks = []
while not self.newtask_queue.empty():
newtasks = self.newtask_queue.get()
result = None
while not self.result_queue.empty():
_, result = self.result_queue.get()
return status, newtasks, result
def run(project):
start_time = time.time()
try:
task = utils.decode_unicode_obj(json.loads(request.form['task']))
except Exception:
result = {
'fetch_result': "",
'logs': u'task json error',
'follows': [],
'messages': [],
'result': None,
'time': time.time() - start_time,
}
return json.dumps(utils.unicode_obj(result)), \
200, {'Content-Type': 'application/json'}
project_info = {
'name': project,
'status': 'DEBUG',
'script': request.form['script'],
}
if request.form.get('webdav_mode') == 'true':
projectdb = app.config['projectdb']
info = projectdb.get(project, fields=['name', 'script'])
if not info:
result = {
'fetch_result': "",
'logs': u' in wevdav mode, cannot load script',
'follows': [],
'messages': [],
'result': None,
'time': time.time() - start_time,
}
return json.dumps(utils.unicode_obj(result)), \
200, {'Content-Type': 'application/json'}
project_info['script'] = info['script']
fetch_result = {}
try:
fetch_result = app.config['fetch'](task)
response = rebuild_response(fetch_result)
module = ProjectManager.build_module(project_info, {
'debugger': True
})
ret = module['instance'].run_task(module['module'], task, response)
except Exception:
type, value, tb = sys.exc_info()
tb = utils.hide_me(tb, globals())
logs = ''.join(traceback.format_exception(type, value, tb))
result = {
'fetch_result': fetch_result,
'logs': logs,
'follows': [],
'messages': [],
'result': None,
'time': time.time() - start_time,
}
else:
result = {
'fetch_result': fetch_result,
'logs': ret.logstr(),
'follows': ret.follows,
'messages': ret.messages,
'result': ret.result,
'time': time.time() - start_time,
}
result['fetch_result']['content'] = response.text
if (response.headers.get('content-type', '').startswith('image')):
result['fetch_result']['dataurl'] = dataurl.encode(
response.content, response.headers['content-type'])
try:
# binary data can't encode to JSON, encode result as unicode obj
# before send it to frontend
return json.dumps(utils.unicode_obj(result)), 200, {'Content-Type': 'application/json'}
except Exception:
type, value, tb = sys.exc_info()
tb = utils.hide_me(tb, globals())
logs = ''.join(traceback.format_exception(type, value, tb))
result = {
'fetch_result': "",
'logs': logs,
'follows': [],
'messages': [],
'result': None,
'time': time.time() - start_time,
}
return json.dumps(utils.unicode_obj(result)), 200, {'Content-Type': 'application/json'}
def run(project):
start_time = time.time()
try:
task = utils.decode_unicode_obj(json.loads(request.form['task']))
except Exception:
result = {
'fetch_result': "",
'logs': u'task json error',
'follows': [],
'messages': [],
'result': None,
'time': time.time() - start_time,
}
return json.dumps(utils.unicode_obj(result)), \
200, {'Content-Type': 'application/json'}
project_info = {
'name': project,
'status': 'DEBUG',
'script': request.form['script'],
}
if request.form.get('webdav_mode') == 'true':
projectdb = app.config['projectdb']
info = projectdb.get(project, fields=['name', 'script'])
if not info:
result = {
'fetch_result': "",
'logs': u' in wevdav mode, cannot load script',
'follows': [],
'messages': [],
'result': None,
'time': time.time() - start_time,
}
return json.dumps(utils.unicode_obj(result)), \
200, {'Content-Type': 'application/json'}
project_info['script'] = info['script']
fetch_result = {}
try:
fetch_result = app.config['fetch'](task)
response = rebuild_response(fetch_result)
module = ProjectManager.build_module(project_info, {'debugger': True})
ret = module['instance'].run_task(module['module'], task, response)
except Exception:
type, value, tb = sys.exc_info()
tb = utils.hide_me(tb, globals())
logs = ''.join(traceback.format_exception(type, value, tb))
result = {
'fetch_result': fetch_result,
'logs': logs,
'follows': [],
'messages': [],
'result': None,
'time': time.time() - start_time,
}
else:
result = {
'fetch_result': fetch_result,
'logs': ret.logstr(),
'follows': ret.follows,
'messages': ret.messages,
'result': ret.result,
'time': time.time() - start_time,
}
result['fetch_result']['content'] = response.text
if (response.headers.get('content-type', '').startswith('image')):
result['fetch_result']['dataurl'] = dataurl.encode(
response.content, response.headers['content-type'])
try:
# binary data can't encode to JSON, encode result as unicode obj
# before send it to frontend
return json.dumps(utils.unicode_obj(result)), 200, {
'Content-Type': 'application/json'
}
except Exception:
type, value, tb = sys.exc_info()
tb = utils.hide_me(tb, globals())
logs = ''.join(traceback.format_exception(type, value, tb))
result = {
'fetch_result': "",
'logs': logs,
'follows': [],
'messages': [],
'result': None,
'time': time.time() - start_time,
}
return json.dumps(utils.unicode_obj(result)), 200, {
'Content-Type': 'application/json'
}
def api_debug_run(project):
start_time = time.time()
try:
task = utils.decode_unicode_obj(json.loads(request.form['task']))
except Exception:
result = {
'fetch_result': "",
'logs': u'task json error',
'follows': [],
'messages': [],
'result': None,
'time': time.time() - start_time,
}
return json.dumps({
'Status': 1,
'Result': utils.unicode_obj(result)
}), 200, cors_resp_header
project_info = {
'name': project,
'status': 'DEBUG',
'script': request.form['script'],
}
if request.form.get('webdav_mode') == 'true':
projectdb = app.config['projectdb']
info = projectdb.get(project, fields=['name', 'script'])
if not info:
result = {
'fetch_result': "",
'logs': u' in wevdav mode, cannot load script',
'follows': [],
'messages': [],
'result': None,
'time': time.time() - start_time,
}
return json.dumps({
'Status': 1,
'Result': utils.unicode_obj(result)
}), 200, cors_resp_header
project_info['script'] = info['script']
fetch_result = {}
try:
module = ProjectManager.build_module(
project_info, {
'debugger': True,
'process_time_limit': app.config['process_time_limit'],
})
# The code below is to mock the behavior that crawl_config been joined when selected by scheduler.
# but to have a better view of joined tasks, it has been done in BaseHandler.crawl when `is_debugger is True`
# crawl_config = module['instance'].crawl_config
# task = module['instance'].task_join_crawl_config(task, crawl_config)
fetch_result = app.config['fetch'](task)
response = rebuild_response(fetch_result)
ret = module['instance'].run_task(module['module'], task, response)
except Exception:
type, value, tb = sys.exc_info()
tb = utils.hide_me(tb, globals())
logs = ''.join(traceback.format_exception(type, value, tb))
result = {
'fetch_result': fetch_result,
'logs': logs,
'follows': [],
'messages': [],
'result': None,
'time': time.time() - start_time,
}
else:
result = {
'fetch_result': fetch_result,
'logs': ret.logstr(),
'follows': ret.follows,
'messages': ret.messages,
'result': ret.result,
'time': time.time() - start_time,
}
result['fetch_result']['content'] = response.text
if (response.headers.get('content-type', '').startswith('image')):
result['fetch_result']['dataurl'] = dataurl.encode(
response.content, response.headers['content-type'])
try:
# binary data can't encode to JSON, encode result as unicode obj
# before send it to frontend
return json.dumps({
'Status': 1,
'Result': utils.unicode_obj(result)
}), 200, cors_resp_header
except Exception:
type, value, tb = sys.exc_info()
tb = utils.hide_me(tb, globals())
logs = ''.join(traceback.format_exception(type, value, tb))
result = {
'fetch_result': "",
'logs': logs,
'follows': [],
'messages': [],
'result': None,
'time': time.time() - start_time,
}
return json.dumps({
'Status': 1,
'Result': utils.unicode_obj(result)
}), 200, cors_resp_header
