Extract Series from Series of Dictionaries

Question

I have a Pandas Series object with a date-index and dict-values like this:

timeSeries = pd.Series({'2014-05-01': {'property1': 1, 'property2': 2},
                        '2014-05-02': {'property1': 3, 'property2': 4}})

I know that every dict contains the same keys ( property1 and property2 ). Is there a way to get a Series without a loop with just property1 as value.

Ie I want:

 propertySeries = pd.Series({'2014-05-01': 1,
                             '2014-05-02': 3})

Answer 1

You can convert Series to numpy array by values and then use DataFrame constructor:

print (timeSeries.values.tolist())
[{'property1': 1, 'property2': 2}, {'property1': 3, 'property2': 4}]

df = pd.DataFrame(timeSeries.values.tolist(), index=timeSeries.index)
print (df)
            property1  property2
2014-05-01          1          2
2014-05-02          3          4

print (df['property1'])
2014-05-01    1
2014-05-02    3
Name: property1, dtype: int64

print (df['property2'])
2014-05-01    2
2014-05-02    4
Name: property2, dtype: int64

Another slowier solution:

print (timeSeries.apply(lambda x: x['property1'])) 
2014-05-01    1
2014-05-02    3
dtype: int64

print (timeSeries.apply(lambda x: x['property2'])) 
2014-05-01    2
2014-05-02    4
dtype: int64

If you created the time series yourself use DataFrame.from_dict :

timeSeries = pd.DataFrame.from_dict({'2014-05-01': {'property1': 1, 'property2': 2},
                                     '2014-05-02': {'property1': 3, 'property2': 4}}, 
                                      orient='index')


print (timeSeries) 
            property1  property2
2014-05-01          1          2
2014-05-02          3          4

Answer 2

If you created the time series yourself, you could create a DataFrame instead:

timeSeries = pd.DataFrame({'2014-05-01': {'property1': 1, 'property2': 2},
                           '2014-05-02': {'property1': 3, 'property2': 4}}).T
timeSeries['property1']
# 2014-05-01    1
# 2014-05-02    3
# Name: property1, dtype: int64