Once we ensure that our index is a DatetimeIndex, we can take advantage of all the methods in this recipe. It is impossible to do selection or slicing based on just the time component of a Timestamp with the .loc indexer. To select all dates by a range of time, you must use the between_time method, or to select an exact time, use at_time. Make sure that the passed string for start and end times consists of at least the hour and minute. It is also possible to use time objects from the datetime module. For instance, the following command would yield the same result as in step 2:

>>> import datetime
>>> crime.between_time(datetime.time(2,0), datetime.time(5,0),
                       include_end=False)

In step 4, we begin using the simple first method, but with a complicated parameter offset. It must be a DateOffset object or an offset alias as a string. To help understand DateOffset objects, it's best to see what they do to a single Timestamp. For example, let's take the first element of the index and add six months to it in two different ways:

>>> first_date = crime_sort.index[0]
>>> first_date
Timestamp('2012-01-02 00:06:00')

>>> first_date + pd.offsets.MonthBegin(6)
Timestamp('2012-07-01 00:06:00')

>>> first_date + pd.offsets.MonthEnd(6)
Timestamp('2012-06-30 00:06:00')

Both the MonthBegin and MonthEnd offsets don't add or subtract an exact amount of time but effectively round up to the next beginning or end of the month regardless of what day it is. Internally, the first method uses the very first index element of the DataFrame and adds the DateOffset passed to it. It then slices up until this new date. For instance, step 4 is equivalent to the following:

>>> step4 = crime_sort.first(pd.offsets.MonthEnd(6))

>>> end_dt = crime_sort.index[0] + pd.offsets.MonthEnd(6)
>>> step4_internal = crime_sort[:end_dt]
>>> step4.equals(step4_internal)
True

Steps 5 through 7 follow from this preceding equivalence directly. In step 8, offset aliases make for a much more compact method of referencing DateOffsets.