selection of data subsets in spatial classes

thanks  Roger Bivand, this is really powerful and much less cumbersome than
I feared.

I ran an inverse distance interpolation on some of the screened data.  In
order to do so I created a destination file from scratch with a 2 x 2 m
grid. The destination file is of rectangular shape and is larger than the
field for which a I have the aerial image data which is a polygon.  As a
result the kriging procedure extends beyond the field.

Is there a way to either create a sampling grid that is the same shape as
the field, ie from a shape file,  or is there a way to subtract the two
layers such that only the field shape is retained (something similar to
mapcalc in GRASS)?