Re: assigning a fill value

Hi Gary,

Ordinary NumPy arrays do not support the concept of missing values.
To get support for missing data you need to use the NumPy MaskedArray
subclass. The paradigm is a little different from the NetCDF
_FillValue concept. Associated with the masked array is a 'mask'
array, a Boolean array with the same dimensions as the data array. It
is set to True for values that need to be masked and false otherwise.
It also includes a fill_value attribute, which is only used when
converting to a 'filled' array.

In your case one way you could get a masked array is as follows
(assume that you don't know that the array actually contains no
values equal to the filled value):

from numpy import ma

y= numpy.zeros((plot_len+15),'f')
y = ma.masked_where(y == -999.9, y, copy = 0) # avoids data copy

# set the fill value

y.set_fill_value(-999.9)

You can also create a masked array directly using ma.array() or its
equivalent ma.MaskedArray().

The function signature is
x = MaskedArray(data, mask=nomask, dtype=None, copy=True,
         fill_value=None, keep_mask=True, hard_mask=False, shrink=True)

PyNgl (and PyNio) understand masked arrays and will behave
appropriately when given one.
PyNIo has various options when reading data from a file concerning
how or whether to create a masked array.

To get a pure NumPy array that has the masked elements replaced by
the fill value use:

y_notmasked = y.filled(fill_value)

Hope this helps.
  -dave

On Dec 8, 2008, at 5:12 PM, Gary Bates wrote:

> How does one assign a fill value to an array in PyNGL?
>
> Here's what I tried and the error I got:
>
> y= numpy.zeros((plot_len+15),'f')
> y._FillValue = -999.9
>
> AttributeError: 'numpy.ndarray' object has no attribute '_FillValue'
>
> -Gary
> _______________________________________________
> pyngl-talk mailing list
> List instructions, subscriber options, unsubscribe:
> http://mailman.ucar.edu/mailman/listinfo/pyngl-talk

_______________________________________________
pyngl-talk mailing list
List instructions, subscriber options, unsubscribe:
http://mailman.ucar.edu/mailman/listinfo/pyngl-talk
Received on Mon Dec 08 2008 - 17:53:06 MST