Reputation: 129
I am working with python2.7 with futures module installed.
I am trying to implement multithreading in tornado using ThreadPoolExecutor.
Here is the code that I have implemented.
from __future__ import absolute_import
from base_handler import BaseHandler
from tornado import gen
from pyrestful import mediatypes
from pyrestful.rest import get, post, put, delete
from bson.objectid import ObjectId
from spark_map import Map
from concurrent import futures
import tornado
class MapService(BaseHandler):
MapDB = dict()
executor = futures.ProcessPoolExecutor(max_workers=3)
@tornado.web.asynchronous
@gen.coroutine
@post(_path='/map', _type=[str, str])
def postMap(self, inp, out):
db = self.settings['db']
function = lambda (x,y): (x,y[0]*2)
future = yield db.MapInfo.insert({'input': inp, 'output': out, 'input_function': str(function)})
response = {"inserted ID": str(future)}
self.write(response)
m = Map(inp, out, function, appName=str(future))
futuree = self.executor.submit(m.operation())
self.MapDB[str(future)] = {'map_object': m, 'running_process_future_object': futuree}
self.finish()
@tornado.web.asynchronous
@gen.coroutine
@delete(_path='/map/{_id}', _types=[str])
def deleteMap(self, _id):
db = self.settings['db']
future = yield db.MapInfo.find_one({'_id': ObjectId(_id)})
if future is None:
raise AttributeError('No entry exists in the database with the provided ID')
chk = yield db.MapInfo.remove(future)
response = { "Succes": "OK" }
self.write(response)
self.MapDB[_id]['map_object'].stop()
del self.MapDB[_id]
self.finish()
In the above code, I receive two inputs using the post request in inp and out. Then I perform some operation with them. This operation should last until a delete request is received to stop and remove the process.
The problem I am facing is with the multiple requests. It only executes the first request while other requests wait for the first one to complete thus blocking the main IOLoop.
So, I want to run each process in a separate thread. How should I implement it?
Upvotes: 0
Views: 1443
Reputation: 24017
It appears that m.operation()
is blocking, so you need to run it on a thread. The way you're doing it blocks the main thread while calling m.operation()
, and spawns a thread after:
self.executor.submit(m.operation())
You want, instead, to pass the function to a thread which will execute it:
self.executor.submit(m.operation)
No parens.
Upvotes: 1