Working with deeply nested data structures is a pain. Sometimes the easiest way to figure out the control structure to get the data that you want is to play around with the data the REPL. My problem is comes from an API, it’s a pain to figure out how to quickly evaluate the structure, make lasting changes to the script, and then jump back into figuring out the structure. In the past, I’ve sucked it up and used the REPL, I’ve written scripts and eval’d integrated the data one change at a time using print, and occasionally, I’ll use jupyter-notebook. None of these solutions seemed to be very good though.

I still haven’t figured out a great way, but yesterday when working with json data from Reddit, the thought occurred to me that I might be able to jump straight into the REPL and just pass in the scope that I was working in. I’d seen this done in the irc3 code, but I couldn’t figure out how that would be useful at the high level that it was done at (basically interrogating the final bot instance).

Sure enough, you can definitely do it. Using the interact method from the code module.

from code import interact


# some expensive calls here to get....
# ...a deeply nested data structures
data = [[{}, {}, {}], [{}, {}], [{}, {}, {}, {}]]
for value in data:
    for deeper_value in value:
        interact(local=locals())

Dropping into the REPL like this gives me the ability to query deeper_value interactively. This is often useful when dealing with JSON to figure out what keys are available.