ENH: Add 'records' outtype to to_dict method.

[tshauck]

to_json has a records type that turns a DataFrame into a

[{'a': 1, 'b': 3}, {'a': 2, 'b': 4}]

records json string. It'd be nice if you could do a similar thing but not have to go json.loads.

[jreback]

grt....can you add a release notes entry (improvements), and a mention in v0.13.0? (use this PR ref as there is not associated issue)

[tshauck]

Updated the commit - let me know of anything else.

[jreback]

@jtratner @cpcloud any comments?

[jtratner]

This test isn't correct. It's dict and list, right? If so, you need to just use assertEqual without the for loops.

This would fail to detect a failure if recons_data were shorter than expected_records, if expected_records was missing certain keys, etc.)

[tshauck]

Good point... it's a list of dicts.

How could I test if nan == nan? I wanted to use the test case used in the other to_dict outtypes, but ran into that issue when I tried assertEqual first.

[jtratner]

I think tm.assert_almost_equal?

[tshauck]

@jtratner I updated the test to with your suggestion, seems to have done the trick.

[jreback]

ok....once more rebase and i think good 2 go (prob just release notes conflict)

[cpcloud]

you don't need the backslash here

[jreback]

thanks!

[nesterko]

Thanks for the feature!

The resulting list of dicts has NumPy data types in entries, which is unfortunate -- difficult to serialize to JSON with known packages such as json.

Is there a chance to have an option to convert entry types to native Python for this?

Thanks