CODEWITHSUNDEEP
pythonpandas
ℕʘʘḆḽḘ
ℕʘʘḆḽḘ

Reputation: 19405

how to make pandas read_csv parse dates correctly from its own csv generated file?

Consider the following very simple pandas dataset:

df = pd.DataFrame({'A': [pd.datetime(2015,02,02)]})     

df
Out[53]: 
           A
0 2015-02-02

Here, the format of the time variable A is a datetime64:

df.dtypes
Out[54]: 
A    datetime64[ns]
dtype: object

Now, I save this dataframe as a csv, and load it using read_csv() with the parse_dates option. 

df.to_csv('P:test.csv')

df2=pd.read_csv('P:test.csv',parse_dates='A')

df2.dtypes
Out[59]: 
Unnamed: 0     int64
A             object
dtype: object

The parser does not recognize the date, so I need to manually run to_datetime again (its a loss of time)...

There should be a way to make Pandas parse correctly the date Pandas itself generated, right? What is wrong here?

Thanks!

Upvotes: 1

Views: 842

Answers (1)

Mike Müller
Mike Müller

Reputation: 85622

You need to specify a list of column names:

parse_dates=['A']

Example:

>>> pd.read_csv('date.csv', parse_dates=['A']).dtypes
Unnamed: 0             int64
A             datetime64[ns]
dtype: object
dtypes: datetime64[ns](1), int64(1)
memory usage: 24.0 bytes

Upvotes: 4

Related Questions