def prepare_response(self, grab):
if self.body_file:
self.body_file.close()
response = Document()
response.head = b''.join(self.response_header_chunks)
if self.body_path:
response.body_path = self.body_path
else:
response.body = b''.join(self.response_body_chunks)
# Clear memory
self.response_header_chunks = []
self.response_body_chunks = []
response.code = self.curl.getinfo(pycurl.HTTP_CODE)
response.total_time = self.curl.getinfo(pycurl.TOTAL_TIME)
response.connect_time = self.curl.getinfo(pycurl.CONNECT_TIME)
response.name_lookup_time = self.curl.getinfo(pycurl.NAMELOOKUP_TIME)
response.download_size = self.curl.getinfo(pycurl.SIZE_DOWNLOAD)
response.upload_size = self.curl.getinfo(pycurl.SIZE_UPLOAD)
response.download_speed = self.curl.getinfo(pycurl.SPEED_DOWNLOAD)
response.remote_ip = self.curl.getinfo(pycurl.PRIMARY_IP)
response.url = self.curl.getinfo(pycurl.EFFECTIVE_URL)
response.parse(charset=grab.config['document_charset'])
response.cookies = CookieManager(self.extract_cookiejar())
# We do not need anymore cookies stored in the
# curl instance so drop them
self.curl.setopt(pycurl.COOKIELIST, 'ALL')
return response
def __init__(self, document_body=None, transport=None, **kwargs):
"""
Create Grab instance
"""
self.meta = {}
self._doc = None
self.config = default_config()
self.config['common_headers'] = self.common_headers()
self.cookies = CookieManager()
self.proxylist = ProxyList()
# makes pylint happy
self.request_counter = None
self.request_head = None
self.request_body = None
self.request_method = None
self.transport_param = transport
self.transport = None
self.reset()
if kwargs:
self.setup(**kwargs)
if document_body is not None:
self.setup_document(document_body)
def prepare_response(self, grab):
response = Response()
try:
response.body = MOCK_REGISTRY[self.request_url]['body']
except KeyError:
raise GrabMockNotFoundError(
'Mock registry does not have information about '\
'following URL: %s' % self.request_url)
now_str = datetime.now().strftime('%a, %d %B %Y %H:%M:%S')
response.head = '\r\n'.join((
'Accept-Ranges:bytes',
'Content-Length:%d' % len(response.body),
'Content-Type:text/plain',
'Date:%s GMT' % now_str,
'Last-Modified:%s GMT' % now_str,
'Vary:Accept-Encoding',
))
response.code = 200
response.total_time = 0
response.name_lookup_time = 0
response.connect_time = 0
response.url = self.request_url
response.parse()
response.cookies = CookieManager(self.extract_cookiejar())
return response
def prepare_response(self, grab):
#if self.body_file:
# self.body_file.close()
response = Response()
head = ''
for key, val in self._response.getheaders().items():
head += '%s: %s\r\n' % (key, val)
head += '\r\n'
response.head = make_str(head, encoding='latin', errors='ignore')
#if self.body_path:
# response.body_path = self.body_path
#else:
# response.body = b''.join(self.response_body_chunks)
if self._request._response_path:
response.body_path = self._request._response_path
# Quick dirty hack, actullay, response is fully read into memory
self._request._response_file.write(self._response.read()) #data)
self._request._response_file.close()
else:
if self._request.body_maxsize is not None:
#if self.response_body_bytes_read > self.config_body_maxsize:
# logger.debug('Response body max size limit reached: %s' %
# self.config_body_maxsize)
response.body = self._response.read(self._request.body_maxsize)
else:
response.body = self._response.read() #data
# Clear memory
#self.response_header_chunks = []
response.code = self._response.status
#response.total_time = self.curl.getinfo(pycurl.TOTAL_TIME)
#response.connect_time = self.curl.getinfo(pycurl.CONNECT_TIME)
#response.name_lookup_time = self.curl.getinfo(pycurl.NAMELOOKUP_TIME)
#response.download_size = self.curl.getinfo(pycurl.SIZE_DOWNLOAD)
#response.upload_size = self.curl.getinfo(pycurl.SIZE_UPLOAD)
#response.download_speed = self.curl.getinfo(pycurl.SPEED_DOWNLOAD)
#response.remote_ip = self.curl.getinfo(pycurl.PRIMARY_IP)
response.url = self._response.get_redirect_location(
) or self._request.url
import email.message
hdr = email.message.Message()
for key, val in self._response.getheaders().items():
hdr[key] = val
response.parse(charset=grab.config['document_charset'], headers=hdr)
jar = self.extract_cookiejar(self._response, self._request)
response.cookies = CookieManager(jar)
# We do not need anymore cookies stored in the
# curl instance so drop them
#self.curl.setopt(pycurl.COOKIELIST, 'ALL')
return response
def custom_prepare_response_func(transport, g):
response = Response()
response.head = cache_item['head']
response.body = body
response.code = cache_item['response_code']
response.download_size = len(body)
response.upload_size = 0
response.download_speed = 0
response.url = cache_item['response_url']
response.parse()
response.cookies = CookieManager(transport.extract_cookiejar())
return response
def custom_prepare_response_func(transport, grab):
response = Response()
response.head = cache_item['head']
response.body = body
response.code = cache_item['response_code']
response.download_size = len(body)
response.upload_size = 0
response.download_speed = 0
response.url = cache_item['response_url']
response.parse(charset=grab.config['document_charset'])
response.cookies = CookieManager(transport.extract_cookiejar())
response.from_cache = True
return response
def custom_prepare_response_func(transport, grab):
doc = Document()
doc.head = cache_item['head']
doc.body = body
doc.code = cache_item['response_code']
doc.download_size = len(body)
doc.upload_size = 0
doc.download_speed = 0
doc.url = cache_item['response_url']
doc.parse(charset=grab.config['document_charset'])
doc.cookies = CookieManager(transport.extract_cookiejar())
doc.from_cache = True
return doc
def __init__(self, document_body=None, transport='pycurl', **kwargs):
"""
Create Grab instance
"""
self.meta = {}
self._doc = None
self.config = default_config()
self.config['common_headers'] = self.common_headers()
self.cookies = CookieManager()
self.proxylist = ProxyList()
self.setup_transport(transport)
self.reset()
if kwargs:
self.setup(**kwargs)
if document_body is not None:
self.setup_document(document_body)
def __init__(self, grab=None):
if grab is None:
self.grab = None
else:
if isinstance(grab, weakref.ProxyType):
self.grab = grab
else:
self.grab = weakref.proxy(grab)
self.status = None
self.code = None
self.head = None
self.headers = None
self.url = None
self.cookies = CookieManager()
self.charset = 'utf-8'
self.bom = None
self.timestamp = datetime.utcnow()
self.name_lookup_time = 0
self.connect_time = 0
self.total_time = 0
self.download_size = 0
self.upload_size = 0
self.download_speed = 0
self.error_code = None
self.error_msg = None
# Body
self.body_path = None
self._cached_body = None
self._unicode_body = None
self._runtime_body = None
self._unicode_runtime_body = None
# DOM Tree
self._lxml_tree = None
self._strict_lxml_tree = None
# Pyquery
self._pyquery = None
# Form
self._lxml_form = None
self._file_fields = {}
def custom_prepare_response_func(transport, g):
response = Response()
response.head = cache_item['head']
response.body = body
response.code = cache_item['response_code']
response.download_size = len(body)
response.upload_size = 0
response.download_speed = 0
# Hack for deprecated behaviour
if 'response_url' in cache_item:
response.url = cache_item['response_url']
else:
logger.debug('You cache contains items without `response_url` key. It is depricated data format. Please re-download you cache or build manually `response_url` keys.')
response.url = cache_item['url']
response.parse()
response.cookies = CookieManager(transport.extract_cookiejar())
return response
def __init__(self, grab=None):
self._grab_config = {}
self.grab = None
if grab:
self.process_grab(grab)
self.status = None
self.code = None
self.head = None
self.headers = None
self.url = None
self.cookies = CookieManager()
self.charset = 'utf-8'
self.bom = None
self.timestamp = datetime.utcnow()
self.name_lookup_time = 0
self.connect_time = 0
self.total_time = 0
self.download_size = 0
self.upload_size = 0
self.download_speed = 0
self.error_code = None
self.error_msg = None
self.from_cache = False
# Body
self.body_path = None
self._bytes_body = None
self._unicode_body = None
# DOM Tree
self._lxml_tree = None
self._strict_lxml_tree = None
# Pyquery
self._pyquery = None
# Form
self._lxml_form = None
self._file_fields = {}
def __init__(self,
document_body=None,
transport='grab.transport.curl.CurlTransport',
**kwargs):
"""
Create Grab instance
"""
self._doc = None
self.config = default_config()
self.config['common_headers'] = self.common_headers()
self._request_prepared = False
self.cookies = CookieManager()
self.proxylist = ProxyList()
self.setup_transport(transport)
self.reset()
if kwargs:
self.setup(**kwargs)
self.clone_counter = 0
if document_body is not None:
self.setup_document(document_body)
def prepare_response(self, grab):
# Information about urllib3
# On python2 urllib3 headers contains original binary data
# On python3 urllib3 headers are converted to unicode
# using latin encoding
try:
#if self.body_file:
# self.body_file.close()
response = Document()
head = ''
for key, val in self._response.getheaders().items():
if six.PY2:
key = key.decode('utf-8', errors='ignore')
val = val.decode('utf-8', errors='ignore')
if six.PY3:
key = key.encode('latin').decode('utf-8', errors='ignore')
val = val.encode('latin').decode('utf-8', errors='ignore')
head += '%s: %s\r\n' % (key, val)
head += '\r\n'
response.head = make_str(head, encoding='utf-8')
#if self.body_path:
# response.body_path = self.body_path
#else:
# response.body = b''.join(self.response_body_chunks)
def read_with_timeout():
if self._request.config_nobody:
return b''
maxsize = self._request.config_body_maxsize
chunks = []
default_chunk_size = 10000
if maxsize:
chunk_size = min(default_chunk_size, maxsize + 1)
else:
chunk_size = default_chunk_size
bytes_read = 0
while True:
chunk = self._response.read(chunk_size)
if chunk:
bytes_read += len(chunk)
chunks.append(chunk)
if maxsize and bytes_read > maxsize:
# reached limit on bytes to read
break
else:
break
if self._request.timeout:
if (time.time() - self._request.op_started >
self._request.timeout):
raise GrabTimeoutError
data = b''.join(chunks)
if maxsize:
data = data[:maxsize]
return data
if self._request.response_path:
response.body_path = self._request.response_path
# FIXME: Quick dirty hack, actullay, response is fully
# read into memory
self._request.response_file.write(read_with_timeout())
self._request.response_file.close()
else:
response.body = read_with_timeout()
# Clear memory
#self.response_header_chunks = []
response.code = self._response.status
#response.total_time = self.curl.getinfo(pycurl.TOTAL_TIME)
#response.connect_time = self.curl.getinfo(pycurl.CONNECT_TIME)
#response.name_lookup_time = (self.curl
# .getinfo(pycurl.NAMELOOKUP_TIME))
#response.download_size = self.curl.getinfo(pycurl.SIZE_DOWNLOAD)
#response.upload_size = self.curl.getinfo(pycurl.SIZE_UPLOAD)
#response.download_speed = self.curl.getinfo(pycurl.SPEED_DOWNLOAD)
#response.remote_ip = self.curl.getinfo(pycurl.PRIMARY_IP)
response.url = (self._response.get_redirect_location()
or self._request.url)
import email.message
hdr = email.message.Message()
for key, val in self._response.getheaders().items():
if six.PY2:
key = key.decode('utf-8', errors='ignore')
val = val.decode('utf-8', errors='ignore')
if six.PY3:
key = key.encode('latin').decode('utf-8', errors='ignore')
val = val.encode('latin').decode('utf-8', errors='ignore')
#if key == 'Location':
# import pdb; pdb.set_trace()
hdr[key] = val
response.parse(charset=grab.config['document_charset'],
headers=hdr)
jar = self.extract_cookiejar() # self._response, self._request)
response.cookies = CookieManager(jar)
# We do not need anymore cookies stored in the
# curl instance so drop them
#self.curl.setopt(pycurl.COOKIELIST, 'ALL')
return response
finally:
self._response.release_conn()
def prepare_response(self, grab):
try:
#if self.body_file:
# self.body_file.close()
response = Response()
head = ''
for key, val in self._response.getheaders().items():
head += '%s: %s\r\n' % (key, val)
head += '\r\n'
response.head = make_str(head, encoding='latin', errors='ignore')
#if self.body_path:
# response.body_path = self.body_path
#else:
# response.body = b''.join(self.response_body_chunks)
def read_with_timeout():
if self._request.config_nobody:
return b''
maxsize = self._request.config_body_maxsize
chunks = []
default_chunk_size = 10000
if maxsize:
chunk_size = min(default_chunk_size, maxsize + 1)
else:
chunk_size = default_chunk_size
total_size = 0
while True:
chunk = self._response.read(chunk_size)
if chunk:
total_size += len(chunk)
chunks.append(chunk)
if maxsize and total_size > maxsize:
logger.debug(
'Response body max size limit reached: %s' %
maxsize)
else:
break
if self._request.timeout:
if time.time(
) - self._request.op_started > self._request.timeout:
raise GrabTimeoutError
data = b''.join(chunks)
if maxsize:
data = data[:maxsize]
return data
if self._request._response_path:
response.body_path = self._request._response_path
# FIXME: Quick dirty hack, actullay, response is fully read into memory
self._request._response_file.write(read_with_timeout())
self._request._response_file.close()
else:
response.body = read_with_timeout()
# Clear memory
#self.response_header_chunks = []
response.code = self._response.status
#response.total_time = self.curl.getinfo(pycurl.TOTAL_TIME)
#response.connect_time = self.curl.getinfo(pycurl.CONNECT_TIME)
#response.name_lookup_time = self.curl.getinfo(pycurl.NAMELOOKUP_TIME)
#response.download_size = self.curl.getinfo(pycurl.SIZE_DOWNLOAD)
#response.upload_size = self.curl.getinfo(pycurl.SIZE_UPLOAD)
#response.download_speed = self.curl.getinfo(pycurl.SPEED_DOWNLOAD)
#response.remote_ip = self.curl.getinfo(pycurl.PRIMARY_IP)
response.url = self._response.get_redirect_location(
) or self._request.url
import email.message
hdr = email.message.Message()
for key, val in self._response.getheaders().items():
hdr[key] = val
response.parse(charset=grab.config['document_charset'],
headers=hdr)
jar = self.extract_cookiejar() #self._response, self._request)
response.cookies = CookieManager(jar)
# We do not need anymore cookies stored in the
# curl instance so drop them
#self.curl.setopt(pycurl.COOKIELIST, 'ALL')
return response
finally:
self._response.release_conn()
