up vote 2 down vote favorite
share [g+] share [fb]

I have python dict object with key as datetime.date object and values as tuple objects:

>>> data_dict
{datetime.date(2006, 1, 1): (5, 3),
 datetime.date(2006, 1, 2): (8, 8),
 datetime.date(2006, 1, 3): (8, 5),
 datetime.date(2006, 1, 4): (3, 3),
 datetime.date(2006, 1, 5): (3, 3),
 datetime.date(2006, 1, 6): (4, 3),
...

and I want to convert it to numpy array object in this format:

dtype([('date', '|O4'), ('high', '<i1'), ('low', '<i1')])

so that I could store it on disk and later work with it, and learn, in numpy, matplotlib...

As a matter of fact, I thought to use this format after looking at this matplotlib examples: http://matplotlib.sourceforge.net/users/recipes.html but can't find my way out how to get there.

link|improve this question
88% accept rate
feedback

1 Answer

up vote 2 down vote accepted

The following will do it:

arr = np.array([(k,)+v for k,v in data_dict.iteritems()], \
         dtype=[('date', '|O4'), ('high', '<f8'), ('low', '<f8')])

If you then wish to use arr as a recarray, you could use:

arr = arr.view(np.recarray)

This will enable you to reference fields by name, e.g. arr.date.

link|improve this answer
Indeed. Thanks. Had no idea that array type can be declared like that. Cheers – zetah Sep 19 '11 at 13:54
Small side note: sorted(data_dict.iteritems()) seems needed for dict – zetah Sep 19 '11 at 14:06
feedback

Your Answer

 
or
required, but never shown
discard

Not the answer you're looking for? Browse other questions tagged or ask your own question.