This is a very simple and practical question. I have the feeling that it must be a silly detail and that there should be similar questions. I wasn't able to find them tho. If someone does I'll happily delete this one.
The closest I found were these: pandas: iterating over DataFrame index with loc
How to select rows within a pandas dataframe based on time only when index is date and time
anyway, the thing is, I have a datetime indexed panda dataframe as follows:
In[81]: y
Out[81]:
PETR4 CSNA3 VALE5
2008-01-01 0.0 0.0 0.0
2008-01-02 1.0 1.0 1.0
2008-01-03 7.0 7.0 7.0
In[82]: y.index
Out[82]: DatetimeIndex(['2008-01-01', '2008-01-02', '2008-01-03'], dtype='datetime64[ns]', freq=None)
Oddly enough, I can't access its values using none of the following methods:
In[83]: y[datetime.datetime(2008,1,1)]
In[84]: y['2008-1-1']
In[85]: y['1/1/2008']
I get the KeyError error.
Even more weird is that the following methods DO work:
In[86]: y['2008']
Out[86]:
PETR4 CSNA3 VALE5
2008-01-01 0.0 0.0 0.0
2008-01-02 1.0 1.0 1.0
2008-01-03 7.0 7.0 7.0
In[87]: y['2008-1']
Out[87]:
PETR4 CSNA3 VALE5
2008-01-01 0.0 0.0 0.0
2008-01-02 1.0 1.0 1.0
2008-01-03 7.0 7.0 7.0
I'm fairly new to pandas so maybe I'm missing something here?
pandas is taking what's inside the [] and deciding what it should do. If it's a subset of column names, it'll return a DataFrame with those columns. If it's a range of index values, it'll return a subset of those rows. What is does not handle is taking a single index value.
Two work around's
1.Turn the argument into something pandas interprets as a range.
df['2008-01-01':'2008-01-01']
2.Use the method designed to give you this result. loc[]
df.loc['2008-01-01']