2. Atomic Components 3. Conclusion Applying File-based Information Flow and Ase Identification to a Legacy System

Note that, because archsq.bat is generated, the files it shares are not represented here. The name *.tbl refers to a group of 8 files with extension " tbl " which are used by the program in ENTER.bat. This corresponds to the service required to enter the next precinct, then run its canvass worksheet. The preceding section illustrated that, within a batch file, the flow of information between programs through shared files could help understanding the functionality of the batch file. This section presents results of techniques which help understand what happens inside individual programs. We applied two heuristics to identify abstract state encapsulations (ASE-a group of global variables together with the routines which access them): same module heu-ristic and delta-ic heuristic [Gira97]. Seven ASE candidates were identified. One correponds to the function and data required to print the headers in unofficial result summary. The other ASE candidates can be categorized as two groups of clone ASEs: a group of 2 ASEs and another of 3 ASEs. This reflects the fact that many functions have been developed by " copy and modify ". The first group of ASEs corresponds to the high-level functions and data needed to produce canvass worksheets from the database of unofficial election results. The second prepares and prints reports (unofficial results summary and the state and local government reports). The ASEs are useful in that they focus the attention of the analyst trying to understand these program to a smaller portion of the code. If one was looking at only one program , the contribution of the ASEs to understanding could be considered marginal since the programs of WELTAB are relatively small. However, when one looks at the complete systems, the ASEs help focus the attention on similar functionalities across the programs. This lead to another unexpected benefit, by comparing the name of functions, the names of global variables and their relation to their context (data files accessed, functions called, and global variables accessed), one can quickly recognize the clones and their differences. Further techniques for architectural components recovery , like dominance tree analysis to identify large components or ADT identification, were not appropriate for this system. The dominance tree was not deep enough to recognize useful patterns due to the limited depth of the call graph. There was no ADT because no record or union was defined in the WELTAB system. Two complementary approaches to understanding …