Ground rules for spatial operations algorithm development (cont’d)
GR 9: Software verification & computational error assessment:
- it is good practice to use authoritative data to confirm results, but do not
be overly surprised that such data bases themselves are too sparse and
are sometimes not accurate enough,
- all formulations in the SRM have closed form solutions in at least the
forward direction. These can be used to automatically generate a dense
- forming localized dense test sets near singular points is a good practice,
- one result of dense testing is to verify that the implementation is working
properly over its intended area of application. Formulation and coding
errors will be evident when dense test sets are used,
- converting a point, then converting that back to the original and then
repeating the process a large number of times is called circular testing.
Circular testing will most likely yield a divergent sequence of results. This
is because round-off error, truncation error and other approximation
errors are not symmetric. Circular testing is not a valid test procedure.