Altering a global var within function - more specifically a pandas series

Question

thanks in advance to anyone looking to enlighten me here a little bit since I have been struggling for a solid couple of hours :-(

def pullQueue(eventQueue, barLength):
    # Setting start and End times for extractions
    startExtract = dt.time(8, 00, 00, 0)
    endExtract = dt.time(22, 00, 00, 0)
    while dt.datetime.now().time() >  startExtract and dt.datetime.now().time() < endExtract: 
        try:
            event = eventQueue.get(False)
        except Queue.Empty:
            pass
        else:
            if event is not None:
                if event.type == 'TICK':
                    global ts
                    ts.append(pd.Series([event.ask], index=[event.time]))
        time.sleep(heartbeat)

The above function (an indentation error occurs when copying into Stackoverflow - please simply ignore) pulls a custom object (event) out of a Queue object - if the event is a tick event, it is meant to append the ask variable and the corresponding timestamp (as an index value) to a pandas Series object (see last nested if statement in function).

However, it simply doesn't work, when I try to access the Series ts - it always logs the 'old' Series excluding the appended values. Originally when I tried to build this function I used this link(stackoverflow question) as guidance. I also constructed the nested if statement (below) on a standalone basis and it did work - any ideas?

if event.type == 'TICK':
    global ts
    ts.append(pd.Series([event.ask], index=[event.time]))

Artem asked a valid question here - I am using threading - see below: evalThread = threading.Thread(target=pullQueue,args=(eventQueue, barLength)) evalThread.start()

I never had shared objects with threading, I would rather used queue to communicate back the result. I will hack a bit to look for solution. — Artem Fedosov, Sep 03 '15 at 13:12
@ArtemFedosov Thanks for looking into this - to be honest I am not stubborn here - if you reocgn it would be better to look at an entirely different approach altogether that would be fine as well - any suggestions are of gr8 help — Philipp, Sep 03 '15 at 13:32

score 1 · Answer 1 · answered Sep 03 '15 at 13:34

1

The method pandas.Series.append doesn't change the original series in-place, it returns a new series, so all you need to do is:

ts = ts.append(pd.Series([event.ask], index=[event.time]))

answered Sep 03 '15 at 13:34

Daniel Boline

506
4
4

score 0 · Accepted Answer · answered Sep 03 '15 at 13:31

Here is almost identical example. You may see that thread worker modifies the variable.

I think the ts variable is not modified by you thread at all. You have several conditional statements (if/while) to reach the globas ts statement and they probably lead to avoiding the block.

import time
import Queue
import threading
import pandas as pd
from collections import namedtuple

Event = namedtuple('Event', ['type', 'ask', 'time'])

eventQueue = Queue.Queue()

for i in xrange(150):
    eventQueue.put(Event(type='TICK', ask=i, time=i))

ts = []

def pullQueue(eventQueue):
    while True:
        try:
            event = eventQueue.get(False)
        except Queue.Empty:
            pass
        else:
            if event is not None:
                if event.type == 'TICK':
                    global ts
                    ts.append(pd.Series([event.ask], index=[event.time]))
        time.sleep(0.01)
evalThread = threading.Thread(target=pullQueue,args=(eventQueue, ))
evalThread.start()

time.sleep(0.01)
print ts

time.sleep(0.01)
print ts

Thanks Artem! that's correct! Daniel Boline answer sums up my silliness! — Philipp, Sep 03 '15 at 13:59

Altering a global var within function - more specifically a pandas series

2 Answers2