gc.collect(2) call shouldn't be forced by pandas

[sylvinus]

We have lots of objects allocated in memory and the explicit call to gc.collect(2) in generic.py is killing our app's performance (taking ~0.5s each time!)
https://github.com/pydata/pandas/blob/master/pandas/core/generic.py#L1278

Ideally pandas shouldn't call the garbage collector manually, or at the very list there should be a way to disable this call.

Thanks!

[shoyer]

Ouch!

I believe you can avoid these checks by setting the mode.chained_assignment option:

pd.set_option('mode.chained_assignment', None)

[sylvinus]

Thanks @shoyer! I can confirm this option avoids the calls.

[jreback]

note that setting single items repeatedly is probably not the performs my way to go

further if you simply copy the original object that is sliced this won't even be checked