Question

I use Multiprocessing in Python in order to do several requests to a database (and other stuff):

po = multiprocessing.Pool()
for element in setOfElements:
    results.append(po.apply_async(myDBRequestModule, (element, other stuff...)))                    
po.close()
po.join()
for r in results:
    newSet.add(r.get())

myDBRequestModule returns an object I defined, made of a list and two numbers. I redefined the hash function, in order to define what I mean by equality in my sets of these objects:

class myObject:
    def __init__(self, aList, aNumber, anotherNumber):
        self.list = aList
        self.number1 = aNumber
        self.number2 = anotherNumber
    def __hash__(self):
        # turn elements of list into a string, in order to hash the string
        hash_text = ""
        for element in self.list:
            hash_text += str(element.x.id) # I use the ID of the element of my list...
        return hash(hash_text)
    def __eq__(self, other):
        self_hash_text = ""
        other_hash_text = ""
        for element in self.list:
            self_hash_text += str(element.x.id)
        for element in other.listDest:
            other_hash_text += str(element.x.id)
        return self_hash_text == other_hash_text 

And in most cases it works as it should. Twice, for no known reason and in exactly the same context, I had a bug:

newSet.add(r.get())
File "/usr/lib/python2.6/multiprocessing/pool.py", line 422, in get
    raise self._value
TypeError: 'str' object does not support item assignment

It comes from the get method (last line):

def get(self, timeout=None):
    self.wait(timeout)
    if not self._ready:
        raise TimeoutError
    if self._success:
        return self._value
    else:
        raise self._value

Since I had this mistake only once and it disappeared, I decided to give up earlier, but it created a second problem recently, and I really don’t know how to fight this bug.
In particular, it’s difficult for me to tell why it happens almost never, and usually works perfectly fine.

Answer 1

multiprocessing is not the issue here.

You have not given us the right code to diagnose the issue. At some point you have assigned a caught exception to self._value. That is where the error is occurring. Look at everywhere that self._value is assigned and you will be on your way to finding this error.