Performance issue about python-benedict HOT 8 CLOSED

In the screenshot above, you can see that 0.21.0 took 60 seconds ish to run. The 0.18.0 code took just 2 seconds. Code to reproduce this is attached above. I just used python 3.7, and in two different environs had 0.21.0 and 0.18.0. This was spotted during a huge performance slow down in our kubernetes cluster - I have now locked the version (should have done that before, that's on me) - but thought you would want to know about this in case it's fixable.

from python-benedict.

I removed the "get" of missing elements and it still takes a long time - I wondered if it might just be that bit that's slow, but it still took the same time. Ideally the timer would only start after the dictionary has been created too - so I reran the test and 0.21.0 took 30 seconds - so perhaps both setting and getting is affected?

from python-benedict.

@alexespencer thank you for reporting this, I'm confident that this slowdown is caused by the base layer introduced here.

from python-benedict.

@fabiocaccamo was that from 0.20.0? I just double checked 0.19.0 and that's all good, the slow down starts from 0.20.0

from python-benedict.

Yes, it has been released in 0.20.0 
Do you pass an existing dict to the constructor or you start with an empty dict?

from python-benedict.

@fabiocaccamo in the test code above I start with an empty one, passing "benedict({})" to initialise it, then populate via .set. However, in our production code we usually read JSON into a dict, then pass that in.

from python-benedict.

@alexespencer I figured out the problem, it is the automatic casting to benedict of retrieved dict values.

The problem occurs because some internal utilities call methods like __getitem__ or get (for example this one) that potentially could return dict values that will be casted to benedict.

from python-benedict.

@alexespencer You can upgrade to 0.21.1 version.

from python-benedict.