> Python's STDlib is nice, but rarely spectacular. An old naive CSV parser I wrote for Chicken scheme (using a simple state machine) beat python's C-based one.
The stdlib is also conservative: I've run into quite a few cases where someone had reported huge gains until they realized they weren't properly handling escaping, Unicode, etc. and once updated the gap narrowed considerably.
You're basically agreeing with the above, and I have experienced the same. The only place where this makes a difference is when the JSON file is in MBs. For smaller file, it's not really worth the extra library.
Depends on how many of the small docs you deal with, I see useful differences on small documents because I process reasonable amounts of them. A 10s change on a million docs for just decoding for a single library is easily worth it for me.
Python's STDlib is nice, but rarely spectacular. An old naive CSV parser I wrote for Chicken scheme (using a simple state machine) beat python's C-based one.