ENH: add to_offset method to Timedelta

[jreback]

from SO

This works, but I think td.to_offset() makes sense

In [1]: td = Timedelta('1h')

In [2]: td
Out[2]: Timedelta('0 days 01:00:00')

In [7]: pd.DateOffset(hours=td.hours)
Out[7]: <DateOffset: kwds={'hours': 1}>

[tvyomkesh]

Does this look like an acceptable ENH (as suggested in SO) or is it expected to introduce a to_offset() method in timedelta which returns offsets.Second?

diff --git a/pandas/tseries/frequencies.py b/pandas/tseries/frequencies.py
index 54b29b1..8694436 100644
--- a/pandas/tseries/frequencies.py
+++ b/pandas/tseries/frequencies.py
@@ -1,4 +1,4 @@
-from datetime import datetime
+from datetime import datetime,timedelta
 from pandas.compat import range, long, zip
 from pandas import compat
 import re
@@ -298,6 +298,10 @@ def to_offset(freqstr):
             name, stride = stride, name
         name, _ = _base_and_stride(name)
         delta = get_offset(name) * stride
+
+    elif isinstance(freqstr, timedelta):
+        delta = offsets.Second(freqstr.total_seconds())
+
     else:
         delta = None
         stride_sign = None

[jreback]

create a test that fails w/o the fix and passes with

[tvyomkesh]

Done. Created pull request #9226. Thanks.

[tvyomkesh]

CI build failed.

ERROR: pandas.tseries.tests.test_frequencies.test_to_offset_timedelta

Traceback (most recent call last):
File "/home/travis/miniconda/envs/pandas/lib/python2.6/site-packages/nose/case.py", line 197, in runTest
self.test(*self.arg)
File "/home/travis/build/pydata/pandas/pandas/tseries/tests/test_frequencies.py", line 86, in test_to_offset_timedelta
result = frequencies.to_offset(td)
File "/home/travis/build/pydata/pandas/pandas/tseries/frequencies.py", line 303, in to_offset
delta = offsets.Second(freqstr.total_seconds())

AttributeError: 'datetime.timedelta' object has no attribute 'total_seconds'

I will need to revisit and add existence check for method etc.

[jorisvandenbossche]

@tvyomkesh the reason is that total_seconds was only added in python 2.7/3.2 (it only fails the tests for 2.6)
But we better continue the discussion on the PR

[tvyomkesh]

@jorisvandenbossche for now I have fixed this by introducing a check if the method is available or not (https://github.com/pydata/pandas/pull/9226/files). For older versions which do not have the method, it will use the equivalent calculation mentioned in the Python 2.7 docs (https://docs.python.org/2/library/datetime.html#datetime.timedelta.total_seconds).
Wondering if it would be a good idea to also wrap it up with try/catch. Happy to discuss more.