BUG: to_json failing on mixed with Timedelta

[jreback]

In [6]: df = DataFrame({'A' : [pd.Timedelta('1 day 1 sec')],'B' : [5], 'C' : [pd.Timestamp('20130101') ]})

In [7]: df
Out[7]: 
                A  B          C
0 1 days 00:00:01  5 2013-01-01

In [8]: df.dtypes
Out[8]: 
A    timedelta64[ns]
B              int64
C     datetime64[ns]
dtype: object

In [9]: df[['A']].to_json()
Out[9]: '{"A":{"0":86401000}}'

In [10]: df.to_json(default_handler=lambda x: x.asm8)
Out[10]: '{"A":{"0":86401000000000},"B":{"0":5},"C":{"0":1356998400000}}'

In [11]: df.to_json()
OverflowError: Maximum recursion level reached

[jreback]

@Komnomnomnom

its currently using the datetime handler, so works in a single dtyped case.

side issue is that you are doing .values on the entire frame?

you need to go column-by-column. Much more efficient (we can do this in python an present a simpler view)