Setting values of a CDMS unlimited file axis, or a file variable with an unlimited axis

It is often necessary to replace the values of an unlimited axis of a file. Ordinarily this is the time axis of a spatiotemporal dataset. An unlimited axis has the property that data can be written beyond the current length of the axis, in order to extend a datafile. To replace the values of a time axis it is natural to try the following:

>>> import cdms
>>> f = cdms.open('sample.nc')
>>> t = f['time']
>>> t.isUnlimited()
True
>>> x = t[:] + 10000.
>>> x
[ 83063.5, 83093.5, 83123.5, 83154. ,]
>>> t[:] = x[:]
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
  File "/export/data/cdat/cdat-4.0b7/lib/python2.4/site-packages/cdms/axis.py", line 1756, in __setslice__
    return apply(self._obj_.setslice,(low,high,MA.filled(value)))
ValueError: shapes are not aligned
>>>

What's going on?

The expression t[:] is shorthand for t[0:lastindex] where lastindex-1 is the largest valid index for t. But if t is unlimited, there IS no largest index. In fact, lastindex resolves to a system-dependent value, typically 2^31 = 2147483648. So internally the statement

t[:] = x[:]

becomes

t[0:2147483648] = x[0:4]

Since the shapes do not match, an exception is raised. The correct way is to use the fact that len(t) returns the length of t as currently written to the file:

>>> t[:len(t)] = x[:]
>>> 

Similar considerations apply to setting the values of a file variable having an unlimited axis in its domain. For example:

>>> import cdms, MV
>>> f = cdms.open('hfogo_O1.nc','r+')
>>> hfogo = f['hfogo']
>>> hfogo.getAxisIds()
['time', 'region', 'lat']
>>> time = f['time']
>>> time.isUnlimited()
True
>>> hfogo.shape
(1812, 4, 61)
>>> len(hfogo)  # Length of the first dimension
1812
#
# First try. Result is a double, which must be recast
#
>>> hfogo[:] = 1000*hfogo[:]  
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
  File "/export/data/cdat/cdat-4.0b7/lib/python2.4/site-packages/cdms/fvariable.py", line 100, in __setslice__
    apply(self._obj_.setslice,(low,high,MA.filled(value)))
TypeError: Array can not be safely cast to required type
#
# Second try, casting the result to single precision
#
>>> hfogo[:] = (1000*hfogo[:]).astype(MV.Float32)
Traceback (most recent call last):
  File "<stdin>", line 1, in ?
  File "/export/data/cdat/cdat-4.0b7/lib/python2.4/site-packages/cdms/fvariable.py", line 100, in __setslice__
    apply(self._obj_.setslice,(low,high,MA.filled(value)))
ValueError: shapes are not aligned
#
# The correct way
#
>>> hfogo[:len(hfogo)] = (1000*hfogo[:]).astype(MV.Float32)
>>>