Assuming the code will be run in a multi threaded environment, is it necessary to implement locks, semaphores, etc. when overriding a magic method? For example, should code like this use a lock to prevent new values from being overwritten by two threads accessing the same missing key?
import time
class TimesDict(dict):
def __missing__(self, key):
self[key] = value = [time.time()]
return value
The confusion comes from the fact that many sources online (ie. http://effbot.org/zone/thread-synchronization.htm#atomic-operations) claim that something like dict_['key'] is atomic, yet the byte code for the given __missing__ example method uses multiple instructions:
dis.dis(TimesDict.__missing__)
5 0 LOAD_GLOBAL 0 (time)
2 LOAD_ATTR 0 (time)
4 CALL_FUNCTION 0
6 BUILD_LIST 1
8 DUP_TOP
10 LOAD_FAST 0 (self)
12 LOAD_FAST 1 (key)
14 STORE_SUBSCR
16 STORE_FAST 2 (value)
6 18 LOAD_FAST 2 (value)
20 RETURN_VALUE
