def fetcher(ctx,
xmlrpc,
xmlrpc_host,
xmlrpc_port,
poolsize,
proxy,
user_agent,
timeout,
Fetcher=Fetcher):
g = ctx.obj
fetcher = Fetcher(inqueue=g.scheduler2fetcher,
outqueue=g.fetcher2processor,
poolsize=poolsize,
proxy=proxy)
fetcher.phantomjs_proxy = g.phantomjs_proxy
if user_agent:
fetcher.user_agent = user_agent
if timeout:
fetcher.default_options = dict(fetcher.default_options)
fetcher.default_options['timeout'] = timeout
g.instances.append(fetcher)
if g.get('testing_mode'):
return fetcher
if xmlrpc:
run_in_thread(fetcher.xmlrpc_run, port=xmlrpc_port, bind=xmlrpc_host)
fetcher.run()
def run_fetcher(g=g):
from pyspider.fetcher.tornado_fetcher import Fetcher
fetcher = Fetcher(inqueue=g.scheduler2fetcher, outqueue=g.fetcher2processor)
fetcher.phantomjs_proxy = g.phantomjs_proxy
run_in_thread(fetcher.xmlrpc_run, port=g.fetcher_xmlrpc_port, bind=g.webui_host)
fetcher.run()
def webui(ctx,
host,
port,
cdn,
scheduler_rpc,
fetcher_rpc,
max_rate,
max_burst,
username,
password,
need_auth,
app=app):
g = ctx.obj
app.config['taskdb'] = g.taskdb
app.config['projectdb'] = g.projectdb
app.config['resultdb'] = g.resultdb
app.config['cdn'] = cdn
if max_rate:
app.config['max_rate'] = max_rate
if max_burst:
app.config['max_burst'] = max_burst
if username:
app.config['webui_username'] = username
if password:
app.config['webui_password'] = password
# fetcher rpc
if isinstance(fetcher_rpc, six.string_types):
fetcher_rpc = connect_rpc(ctx, None, fetcher_rpc)
if fetcher_rpc is None:
fetcher = Fetcher(inqueue=None, outqueue=None, async=False)
fetcher.phantomjs_proxy = g.phantomjs_proxy
app.config['fetch'] = lambda x: fetcher.fetch(x)[1]
else:
import umsgpack
app.config['fetch'] = lambda x: umsgpack.unpackb(
fetcher_rpc.fetch(x).data)
if isinstance(scheduler_rpc, six.string_types):
scheduler_rpc = connect_rpc(ctx, None, scheduler_rpc)
if scheduler_rpc is None and os.environ.get('SCHEDULER_NAME'):
app.config['scheduler_rpc'] = connect_rpc(
ctx, None, 'http://%s/' %
(os.environ['SCHEDULER_PORT_23333_TCP'][len('tcp://'):]))
elif scheduler_rpc is None:
app.config['scheduler_rpc'] = connect_rpc(ctx, None,
'http://localhost:23333/')
else:
app.config['scheduler_rpc'] = scheduler_rpc
app.debug = g.debug
g.instances.append(app)
if g.get('testing_mode'):
return app
app.run(host=host, port=port)
def fetcher(ctx, xmlrpc, xmlrpc_host, xmlrpc_port):
g = ctx.obj
from pyspider.fetcher.tornado_fetcher import Fetcher
fetcher = Fetcher(inqueue=g.scheduler2fetcher, outqueue=g.fetcher2processor)
fetcher.phantomjs_proxy = g.phantomjs_proxy
g.instances.append(fetcher)
if xmlrpc:
run_in_thread(fetcher.xmlrpc_run, port=xmlrpc_port, bind=xmlrpc_host)
fetcher.run()
def run_fetcher(g=g):
from pyspider.fetcher.tornado_fetcher import Fetcher
fetcher = Fetcher(inqueue=g.scheduler2fetcher,
outqueue=g.fetcher2processor)
fetcher.phantomjs_proxy = g.phantomjs_proxy
run_in_thread(fetcher.xmlrpc_run,
port=g.fetcher_xmlrpc_port,
bind=g.webui_host)
fetcher.run()
def webui(ctx, host, port, cdn, scheduler_rpc, fetcher_rpc,
max_rate, max_burst, username, password):
g = ctx.obj
from pyspider.webui.app import app
app.config['taskdb'] = g.taskdb
app.config['projectdb'] = g.projectdb
app.config['resultdb'] = g.resultdb
app.config['cdn'] = cdn
if max_rate:
app.config['max_rate'] = max_rate
if max_burst:
app.config['max_burst'] = max_burst
if username:
app.config['webui_username'] = username
if password:
app.config['webui_password'] = password
# fetcher rpc
if isinstance(fetcher_rpc, basestring):
fetcher_rpc = connect_rpc(ctx, None, fetcher_rpc)
if fetcher_rpc is None:
from pyspider.fetcher.tornado_fetcher import Fetcher
fetcher = Fetcher(inqueue=None, outqueue=None, async=False)
fetcher.phantomjs_proxy = g.phantomjs_proxy
app.config['fetch'] = lambda x: fetcher.fetch(x)[1]
else:
import umsgpack
app.config['fetch'] = lambda x: umsgpack.unpackb(fetcher_rpc.fetch(x).data)
if isinstance(scheduler_rpc, basestring):
scheduler_rpc = connect_rpc(ctx, None, scheduler_rpc)
if scheduler_rpc is None and os.environ.get('SCHEDULER_NAME'):
app.config['scheduler_rpc'] = connect_rpc(ctx, None, 'http://%s/' % (
os.environ['SCHEDULER_PORT_23333_TCP'][len('tcp://'):]))
elif scheduler_rpc is None:
app.config['scheduler_rpc'] = connect_rpc(ctx, None, 'http://localhost:23333/')
else:
app.config['scheduler_rpc'] = scheduler_rpc
app.debug = g.debug
if g.get('testing_mode'):
return app
app.run(host=host, port=port)
def fetcher(ctx, xmlrpc, xmlrpc_host, xmlrpc_port, poolsize, proxy, user_agent, timeout, Fetcher=Fetcher):
g = ctx.obj
fetcher = Fetcher(inqueue=g.scheduler2fetcher, outqueue=g.fetcher2processor,
poolsize=poolsize, proxy=proxy)
fetcher.phantomjs_proxy = g.phantomjs_proxy
if user_agent:
fetcher.user_agent = user_agent
if timeout:
fetcher.default_options = dict(fetcher.default_options)
fetcher.default_options['timeout'] = timeout
g.instances.append(fetcher)
if g.get('testing_mode'):
return fetcher
if xmlrpc:
run_in_thread(fetcher.xmlrpc_run, port=xmlrpc_port, bind=xmlrpc_host)
fetcher.run()
def run_webui(g=g):
import cPickle as pickle
from pyspider.fetcher.tornado_fetcher import Fetcher
fetcher = Fetcher(inqueue=None, outqueue=None, async=False)
fetcher.phantomjs_proxy = g.phantomjs_proxy
from pyspider.webui.app import app
app.config['taskdb'] = g.taskdb
app.config['projectdb'] = g.projectdb
app.config['resultdb'] = g.resultdb
app.config['fetch'] = lambda x: fetcher.fetch(x)[1]
app.config['scheduler_rpc'] = g.scheduler_rpc
#app.config['cdn'] = '//cdnjs.cloudflare.com/ajax/libs/'
if g.demo_mode:
app.config['max_rate'] = 0.2
app.config['max_burst'] = 3.0
if 'WEBUI_USERNAME' in os.environ:
app.config['webui_username'] = os.environ['WEBUI_USERNAME']
app.config['webui_password'] = os.environ.get('WEBUI_PASSWORD', '')
if not getattr(g, 'all_in_one', False):
app.debug = g.debug
app.run(host=g.webui_host, port=g.webui_port)
def run_webui(g=g):
import cPickle as pickle
from pyspider.fetcher.tornado_fetcher import Fetcher
fetcher = Fetcher(inqueue=None, outqueue=None, async=False)
fetcher.phantomjs_proxy = g.phantomjs_proxy
from pyspider.webui.app import app
app.config['taskdb'] = g.taskdb
app.config['projectdb'] = g.projectdb
app.config['resultdb'] = g.resultdb
app.config['fetch'] = lambda x: fetcher.fetch(x)[1]
app.config['scheduler_rpc'] = g.scheduler_rpc
#app.config['cdn'] = '//cdnjs.cloudflare.com/ajax/libs/'
if g.demo_mode:
app.config['max_rate'] = 0.2
app.config['max_burst'] = 3.0
if 'WEBUI_USERNAME' in os.environ:
app.config['webui_username'] = os.environ['WEBUI_USERNAME']
app.config['webui_password'] = os.environ.get('WEBUI_PASSWORD', '')
if not getattr(g, 'all_in_one', False):
app.debug = g.debug
app.run(host=g.webui_host, port=g.webui_port)
