Added function to convert Excel date to datetime.

[jmcnamara]

This relates to issue #76 and the initial pull request #77.

I've added a function, xldate.xldate_as_datetime(), to return a datetime.datetime object from an Excel serial date/time.

This would be useful for anyone interested in preserving milli/microseconds from an Excel date and would allow us to close Pandas issue: pandas-dev/pandas/issues/5945.

The PR includes tests based on dates and time extracted from Excel files. The function is very slightly slower than the xldate.xldate_as_tuple() function:

$ python
Python 2.7.2 (default, Feb 19 2013, 23:47:31)
[GCC 4.2.1 (Based on Apple Inc. build 5658) (LLVM build 2336.11.00)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>>
>>> from xlrd import xldate
>>> import timeit
>>>
>>> xldate.xldate_as_tuple(41657.75165, 0)
(2014, 1, 18, 18, 2, 23)
>>>
>>> xldate.xldate_as_datetime(41657.75165, 0)
datetime.datetime(2014, 1, 18, 18, 2, 22, 560000)
>>>
>>> timeit.timeit('xldate.xldate_as_tuple(   41657.75165, 0)', setup='from xlrd import xldate', number=170000)
1.045557975769043
>>>
>>> timeit.timeit('xldate.xldate_as_datetime(41657.75165, 0)', setup='from xlrd import xldate', number=170000)
1.0717370510101318
>>>

I didn't update the docs because I wasn't sure if they are auto-generated or not. If you need me to update them I can do that too.

John

[cjw296]

Can we lose this boilerplate? nose is the "blessed" way to run these and the boilerplate isn't needed.

[jmcnamara]

No problem. I was copying the format of the existing date test cases.

[cjw296]

feel free to clean them up too if you in there ;-)

[jmcnamara]

No thanks. :-)

[jmcnamara]

Hi Chris,

I've updated the PR with the 3 initial test classes merged into 1 and with reduced and documented testcases.

I also profiled the int() and subtraction versus divmod() and the former is faster so I've left that as is:

>>> import timeit
>>>
>>> timeit.timeit('days = int(60376.2345); fraction = 60376.2345 - days', number=10**7)
3.014816999435425
>>>
>>> timeit.timeit('days, fraction = divmod(60376.2345, 1)',               number=10**7)
3.755660057067871
>>>

Hope it passes muster now.

John