This is needed because it usually does not make sense, or
is not even possible, to put all the rules in the metamodel and check them during each
modeling action. This is especially true when checking partial models, when there are
multiple models, or when integrating models made by different developers.
Generators for model analysis can also be used for guiding modeling work and
informing about actions needed. A typical such scenario is to look if a model, or
models, is incomplete and report possible actions needed to make the model complete.
Such model checking can be run similarly to generators: when needed or after
conducting certain modeling actions.
Metrics When moving toward model-based development, code-based metrics can
still be applied; they are now calculated from the generated code instead of from the
manually written code. As platform code is already available, the metrics may also
cover platform functions or libraries used instead of focusing only on the generated
application code. Use of code metrics based on models is easy as it does not require
much change to earlier practices, but it is not likely to be the most effective use of
metrics.
In DSM, code-based metrics no longer measure the amount of human work
needed. Metrics, like function point analysis (FPA, Albrecht and Gaffney, 1983), to
analyze program size and to estimate required development effort are no longer
relevant: the application is often already ready when we can calculate these metrics.
Pages:
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182