Generation is primarily thought of as focusing on code, but may also include codelike
or declarative formats such as SQL or XML, documentation ?¬?les, and simple text
?¬?les. It can also be used as the basis for model checking reports, although the modeling
language may be a better place for rules that should always hold, and for certain
specialized analysis tasks it may be better to use existing external programs.
When generating code, the kind of code you want is determined by what you would
write by hand, with perhaps some concessions to ease of generation. While this would
appear to leave little in the way of generic patterns for code generation, experience
shows that almost all code can be seen as either a ?¬‚ow model or a state model. While
simple ?¬‚ow models can be generated with a na?±??ve recursive approach, anything more
complicated tends to use functions, called either dynamically or statically, or then
switch??“case statements working similarly to functions.
Having the generator not only create the output ?¬?les but also compile them and
run them is a key ingredient in raising the abstraction level. Such an autobuild
approach allows the modeler to go straight from models to the ?¬?nished product,
remaining at a high level of abstraction and thinking in the problem domain all the
time.
Pages:
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580