memory leak?

[datoug]

Our python application uses babel 1.3 and it has a memory leak. I used heapy to dump the objects that use the most memory, and it seems the strings are referenced from babel(core.py):

Index Count % Size % Cumulative % Referred Via:
0 14 0 642896 3 642896 3 "['languages']"
1 15 0 274536 1 917432 4 "['territories']"
2 14 0 224672 1 1142104 5 "['currency_names']"
3 804 1 219640 1 1361744 7 "['long']"
4 15 0 213096 1 1574840 8 "['meta_zones']"
5 14 0 195920 1 1770760 9 "['time_zones']"
6 19 0 157792 1 1928552 9 "['scripts']"
7 909 1 113712 1 2042264 10 "['other']"
8 832 1 103144 0 2145408 10 "['one']"
9 788 1 93512 0 2238920 11 "['standard']"

[akx]

That's most definitely locale data.

I could see that happening if you repeatedly construct new Locale objects (either via the direct constructor Locale() or Locale.parse() (or any other method that might build Locales)) and stores references to them. Can you use Heapy to find out where Locale objects get referenced?

What sort of application is this? For instance, if it's a webapp, is it possible that each request is getting a new Locale in, say, a threadlocal dict which never gets cleaned up?

As an aside, PR #305 should make this impossible.

[akx]

@datoug Still experiencing this problem? :)

[akx]

Closing as abandoned.