The uv routines allow the correlation data to be stored on disk, either
as floating point numbers, or as 16 bit integers with an associated scale
factor. The 16 bit format roughly halves the disk space required, but slows
the read and write operation, and can cause precision problems. On the
first call to uvwrite, the uv routines decide on the format to
use, using a simple rule. The uvset call
can be used to override this rule. To be of use, it must be called before
the first call to uvwrite. The type argument is a single
character, being `r' or `j', which instructs floating
point or scaled integers, respectively, to be used.
This conrtols whether UVWRITE writes the
data to the corr or wcorr variable. The default is to write it
to the corr item (i.e. it assumes that the data is spectral, rather
than continuum data).