Models used to
generate code are usually poor sources from which to generate tests: would such tests
really prove anything? Data from models can however be used to select from existing
tests those that are applicable to the developed system. This is especially relevant for
large systems where thousands of tests can exist.DSMcan also be used speci?¬?cally for
testing: test engineers can de?¬?ne tests using testing and product concepts to generate
test cases and applications running the tests.
Language Use and Re?¬?nement Information We can also shift the focus
from reporting about models to reporting about metamodels. Generators can produce
information about how the language and generator are used. This helps to reveal
patterns of language use, which concepts are not used, which models use older version
of the language, and so on. Modeling languages can also include some elements with
open semantics that provide possibilities for modelers to express things that can??™t be
captured with the current language. A generator can report on possible uses of these
concepts to ?¬?nd out where users of the language ?¬?nd it lacking and identify areas for
further improvement. We discuss this more in Chapter 10 together with language
evolution and maintenance.
With multiple generators, we start to enjoy having a single source but multiple
targets: developers need to change only one place and the generators take care of the
rest.
Pages:
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185