[Numpy-discussion] numpy.ndarray.astype conversion

Orest Kozyar orest.kozyar@gmail....
Sat May 24 11:12:18 CDT 2008


> > The following line:  array([6.95e-5]).astype('S')
> > returns: array(['6'], dtype='|S1')

> I don't know what it should do. Issue a warning, maybe?
> What are you trying to do?

At the very least a warning would not hurt.  When I first started
using the numpy arrays it looked like conversion to string would
ensure that the minimum string length needed to accurately represent
the numbers in the array would be used.  It wasn't until I noticed
some odd outliers in my plotted data that I realized what was going on
-- when exponential notation is used, the conversion gets truncated.
I use numpy arrays to store data from SQL queries, and this data
sometimes has mixed types (string, number, etc.).  I now realize that
I was setting up the arrays wrong and should have been using
dtype=object instead.   It just seems like the above might be
something that might "bite" new users of numpy unintentionally.

Orest


More information about the Numpy-discussion mailing list