Home

GRASP/Ada - Auburn University

1. end Lf end loop Lend SearchArray Figure 2 CSD for SearchArray substantial positive effect on detailed comprehension of the software Figures 3 and 4 contain an Ada task body CONTROLLER adapted from BAR84 which loops through a priority list attempting to accept selectively a REQUEST with priority P Upon on acceptance some action is taken followed by an exit from the priority list loop to restart the loop with the first priority In typical Ada task fashion the priority list loop is contained in an outer infinite loop This short example contains two threads of control the rendezvous which enters and exists at the accept statement and the thread within the task body In addition the priority list loop contains two exits the normal exit at the beginning of the loop when the priority list has been exhausted and an explicit exit invoked task body TASK_NAME is begin loop for p in PRIOITY loop select accept REQUEST p D DATA do ACTION D end exit else null end select end loop end loop end TASK_NAME Figure 3 Ada Source for Task Body Controller task body TASK_NAME is begin loop for p in PRIOITY loop select ss accept REQUEST p D DATA do ACTION D 4 Lend lt 4 exit _ else Lo null end select end loop end loop Lend TASK NAME Figure 4 CSD for Ada Task Body Controller within the select statement While the con
2. 101 2750 143 5250 132 5250 127 3000 134 7750 160 8250 145 8000 116 1750 99 7250 115 5000 1 1 8977 3326 8885 7437 6750 7734 1161 9184 5286 3122 5197 amp P 16 11 17 13 F_0 0423 2703 1596 4552 6119 4198 3877 6070 5832 9616 4069 F _ gt F 0 a 00117 08115 00099 000011 00000871 0000112 013231 000129 000000215 0202 0 258 Table 8 ANOVA Summary for Treatments FC NS CSD Totals for Assigned Ranks of 32 Students for 3 Treatments PCH FC R NS _R CSD _R 1 SEQ 57 0 66 0 69 0 2 SEL 60 0 73 5 58 5 3 ITR 64 0 58 0 70 0 4 GEN READ 54 0 57 0 81 0 5 EXT P COD 51 0 68 5 72 5 6 CODE FROM 46 5 63 0 82 5 7 MANUAL 60 0 58 0 74 0 8 PREF MANL 53 5 61 0 77 5 9 ECONOMY 46 0 59 0 87 0 10 PREF AUTO 48 5 57 0 86 5 11 PREF GEN 50 0 58 0 84 0 Statistics for Treatments FC NS and CSD P 2k PCH SS_TOTL SS_TR MS TR SS RESID MS RESID FO 1 SEQ 59 00 2 4375 1 2188 56 5625 0 9123 1 3359 270 2 SEL 59 00 4 2656 2 1328 54 7344 0 8828 2 4159 098 3 ITR 59 50 2 2500 1 1250 57 2500 0 9234 1 2183 303 4 GEN READ 61 00 13 6875 6 8438 47 3125 0 7631 8 9683 00004 5 EXT P COD 54 00 8 1719 4 0859 45 8281 0 7392 5 5278 0062 6 CODE FROM 61 50 20 2969 10 1484 41 2031 0 6646 15 2708 000004 7 MANUAL 58 00 4 7500 2 3750 53 2500 0 8589 2 7653 071 8 PREF MANL 62 00 9 4219 4 7109 52 5781 0 8480 5 5551 006 9 ECONOMY 60
3. August 1993 D A Scanlan Structured Flowcharts Outperform Pseudocode An Experimental Comparison IEEE Software Sep 1989 28 36 R Selby et al A Comparison of Software Verification Techniques NASA Software Engineering Laboratory Series SEL 85 001 Goddard Space Flight Center Greenbelt Maryland 1985 B Shneiderman et al Experimental Investigations of the Utility of Detailed Flowcharts in Programming Communications of the ACM No 20 1977 pp 373 381 Nan C Shu Visual Programming New York NY Van Norstrand Reinhold Company Inc 1988 C Sittenauer and M R Olsem Re engineeing Tools Report Software Technology Support Center Hill Air Force Base UT 840556 July 1992 T Standish An Essay on Software Reuse IEEE Transactions on Software Engineering SE 10 9 494 497 1985 L L Tripp A Survey of Graphical Notations for Program Design An Update ACM Software Engineering Notes Vol 13 No 4 1989 39 44 A I Wasserman P A Pircher and R J Muller An Object Oriented Structured Design Method for Code Generation ACM SIGSOFT Software Engineering Notes Vol 14 No 1 January 1989 32 52 41 APPENDICES A Installation Guide B Integrating GRASP Ada with Software through Pictures StP E Evaluation Instrument 42 Appendix A GRASP Ada Installation Guide GRASP Ada 3 4 Installation Notes GRASP_ HOME sup doc README install 3 4 First compiled 10 31 91 by
4. Introduction to Software Engineering during the Fall 1992 quarter These students all had experience with FORTRAN Pascal and C in previous courses None had formal training in Ada for which the GRASP Ada tool was designed Since participation in the evaluation was optional five bonus points to be added to the final exam score were offered as an incentive All students present took part in the evaluation Each of the graphical representations included in the first part of the evaluation instrument was presented briefly in class and exercises were assigned involving the Nassi Shneiderman diagram NS and the control structure diagram CSD Most students were familiar with the flowchart FC from prior classes The GRASP Ada prototype was presented during a laboratory session and used in conjunction with the commercial CASE tool Software through Pictures StP The primary focus of the CSE 422 lab is the development of a software specification using CASE tools such as StP 6 3 The Evaluation Results An item analysis was performed on the data collected for the eleven PCHs in the first part of the evaluation instrument i e all items except the three background items at the beginning and the last item which asked for suggested improvements to the diagrams Following the item analysis an analysis of variance was performed to determine if differences in the preferences were statistically significance The results item analysis and the tests for
5. significance are presented below followed by a general summary of the responses from the second part of the evaluation instrument 6 3 1 Item Analysis of Comparison of Graphical Representations The subjects were given the following instructions regarding the eleven performance characteristics Based on the experience you have gained by using these diagramming tools to represent algorithms you are asked to assign a rating to each of the diagrams with respect to a specific comparison among the diagrams You may assign the same rating to more than one diagram for a given comparison Select your ratings from the following scale and enter them as indicated below 26 5 best most first choice 4 3 moderate 2 1 worst least last choice For each of the eleven items below the subjects used the rating scale above to complete the following FC NS WO AD CSD The eleven items describing performance characteristics were l 2 10 Compare the diagrams with respect to how well each shows sequence Compare the diagrams with respect to how well each shows selection Compare the diagrams with respect to how well each shows iteration Compare these diagrams with respect to overall readability consider reading someone else s code Each of these tools can be used with informal pseudocode as opposed to actual statements in a programming language and as such can be thought of as a graphical extension to ps
6. to make them easier to automate to provide for control flow other than sequence selection iteration etc What improvements can you suggest for any of the diagrams we used in this class Name Date optional Evaluation of GRASP Ada In CSE 422 lab you were provided the opportunity to work briefly with GRASP Ada a prototype reverse engineering tool for software written in Ada or AdaPDL The prototype is currently being evaluated prior to widespread release via the network As a prototype GRASP Ada is expected to undergo continual modification over the next year especially with respect to integration with commercially available CASE tools The current GRASP Ada prototype includes automatic generation of Control Structure Diagrams CSDs Future releases will include the generation of object diagrams Your responses to the items below are intended to provide the developers with directions for enhancements to the prototype including additional user interface requirements and overall functionality The GRASP Ada Project is supported in part by funding from George C Marshall Space Flight Center NASA Alabama 35821 Contract Number NASA NCC8 14 l Was the User Interface intuitive 23 What changes would you make to the User Interface 3 What were the major problems you encountered when using GRASP Ada Rank the following items in order of importance in the prototype Note some of these items are available in the current versio
7. 2 4 1 0 5 0 3 0 2 0 4 0 155 27 3 4 1 2 5 3 0 4 0 1 0 2 0 5 0 166 29 4 5 1 2 3 4 0 5 0 1 0 2 0 3 0 177 30 1 2 3 4 5 1 0 2 0 3 0 4 0 5 0 188 31 5 4 2 1 3 5 0 4 0 2 0 1 0 3 0 199 32 2 5 3 1 4 2 0 5 0 3 0 1 0 4 0 210 33 2 5 1 3 4 2 0 5 0 1 0 3 0 4 0 Totals 63 5 69 0 48 0 43 5 76 0 Table 5 shows that for example student number 23 ranked FC NS WO AD CSD as 4 2 3 3 and 5 respectively The Friedman s Test requires that the responses for all treatments be ranked within each block from 1 to 5 and therefore that the sum of the ranks within each block is 15 as shown below 5 5 VR j 15 Fi pi The tied ranks receive average ranks The last five columns FC_R NS_R WO_R AD_R and CSD_R of Table 9 give the ranks for each of the 20 students for the performance characteristic SEQ after tied ranks have been averaged Now the ranks assigned by student number 23 become 4 0 1 0 2 5 2 5 and 5 0 respectively Since the student ranked NS the lowest the treatment NS_R received a rank of 1 0 The student ranked WO and AD equally 30 and next lowest therefore ranks 2 and 3 were averaged and WO_R and AD_R each received 2 5 Similarly FC_R received a rank of 4 0 and CSD a rank of 5 0 Thus we have 5 YOR 4 0 1 0 2 5 2 5 5 15 jal 137 as expected for student number 23 in row 13 The 2 way treatments and blocks ANOVA is conducted on ranks as illustrated below Let SS Total SS Treatments SS Blocks and SS Residuals represent total su
8. 3 00 3 33 2 54 1 96 4 55 ITEM AVG 363 3 06 3 49 2 52 2 16 4 15 The results in Table 2 clearly indicate the overall trend in preference for the CSD For an additional perspective the averages in Table 2 were converted into percentages and are shown in Table 3 Then the differences between the percentages in Table 3 for the CSD and the other diagram types is shown in Table 4 28 Table 3 Percentage Scores For Graphical Representations PCH N items FC NS Wo AD CSD 1 SEQ 33 64 24 72 73 52 80 46 36 78 79 2 SEL 33 70 30 81 21 49 17 40 95 72 73 3 ITR 33 69 09 69 70 51 67 42 86 78 18 4 GEN READ 33 60 61 67 50 53 33 42 00 84 85 5 EXT P COD 33 56 97 75 15 47 50 49 52 78 79 6 CODE FROM 33 56 36 70 62 52 00 42 73 86 25 7 MANUAL 33 61 82 63 12 52 17 47 50 78 18 8 PREF MANL 33 60 00 66 06 48 46 43 20 83 03 9 ECONOMY 33 53 94 65 45 50 40 40 00 90 30 10 PREF AUTO 33 60 61 70 30 47 20 41 74 90 91 11 PREF GEN 33 60 00 66 67 50 77 39 20 90 91 ITEM AVG 363 61 27 69 89 50 48 43 23 82 98 Table 4 Percentage Score Differences CSD Compared to Others PCH N items FC CSD NS CSD wo CSD AD CSD CSD 1 SEQ 33 14 55 6 06 25 99 32 42 2 SEL 33 2 42 8 48 23 56 31 77 3 ITR 33 9 09 8 48 26 52 35 32 4 GEN READ 33 24 24 17 35 31 52 42 85 5 EXT P COD 33 21 82 3 64 31 29 29 26 6 CODE FROM 33 29 89 15 62 34 25 43 52 7 MANUAL 33 16 36 15 06 26 01 30 68 8 PREF MANL 33 23 03 16 97 34 57 39 83 9 ECONOMY 33 36 36 24 85
9. Percentage Scores For Graphical Representations 4 29 Table 4 Percentage Score Differences CSD Compared to Others 29 Table 5 Data for Performance Characteristic SEQ PCH 1 30 Table 6 Results of ANOVA dalle een be ae Bede es 32 Table 7 ANOVA Summary for Treatments FC NS WO AD CSD 33 Table 8 ANOVA Summary for Treatments FC NS CSD 34 1 0 Introduction Computer professionals have long promoted the idea that graphical representations of software can be extremely useful as comprehension aids when used to supplement textual descriptions and specifications of software especially for large complex systems SHU88 AOY89 SCA89 The general goal of this research has been the investigation formulation and generation of graphical representations of algorithms structures and processes for Ada GRASP Ada This specific task has focused on reverse engineering of control structure diagrams from Ada PDL or source code Reverse engineering normally includes the processing of source code to extract higher levels of abstraction for both data and processes The primary motivation for reverse engineering is increased support for software reusability verification and software maintenance all of which should be greatly facilitated by automatically generating a set of formalized diagrams to supplement the source code and other forms of existing documentation The overall
10. Window fA ftmp_mnt home ese_h2 cross research nasa grasp update93 rpt_final examplel esd No syntax errors detected procedure PROCEDURE_NAME is begin null loop null if CONDITION then null while CONDITION loop null end loop for INDEX_VARIABLE in VALUE_RANGE loop null end loop end if exit when CONDITION end loop end PROCEDURE_NAME Figure 16 CSD Window with Program Structure Resulting from Clicking on procedure body if then else for loop and while loop 18 5 0 Control Structure Diagram Generator The GRASP Ada control structure diagram generator CSDgen is described in this section from a technical and developmental perspective Since display mechanisms for both the screen and printer are an integral part of the CSDgen application these are included in this section as well A more complete history and rationale for the development of the CSD is contained in CRO90a CRO90b The graphical constructs produced by CSDgen are summarized in Figure 5 Section 2 0 Examples of the CSD are presented in conjunction with the User Interface in Section 4 0 5 1 Generating the CSD The primary function of CSDgen is to produce a CSD for a corresponding Ada source or PDL file Although a complete parse is done during CSD generation CSDgen assumes the Ada source code has been previously compiled and thus is syntactically correct Currently little error recovery and error reporting are attempted when a
11. and CSD were ranked as 1 2 3 average ranks were assigned to equal ratings as before and the Friedman s Test was applied for each of the eleven PCHs The results are summarized in Table 8 32 Totals for Assigned Ranks of 20 Students for 5 Treatments 10 11 Table 7 ANOVA Summary for Treatments FC NS WO AD CSD PCH SEQ SEL ITR GEN READ EXT P COD CODE FROM MANUAL PREF MANL ECONOMY PREF AUTO PREF GEN FC_R 63 70 74 63 57 57 65 63 55 55 57 5 5 NS_R 69 85 68 67 79 70 68 69 71 70 70 0 5 WO_R AD R 48 0 43 5 39 0 38 0 50 5 40 5 47 0 39 5 41 5 46 5 45 5 43 0 48 5 46 0 42 5 45 0 49 0 38 5 42 0 40 5 46 0 38 0 Statistics for Treatments FC NS WO AD 10 11 PCH SEQ SEL ITR GEN READ EXT P COD CODE FROM MANUAL PREF MANL ECONOMY PREF AUTO PREF GEN SS _TOTL 182 188 182 191 185 194 189 196 187 194 197 50 00 50 00 00 00 S 38 86 38 58 57 59 28 50 70 94 81 S_TR 2750 7250 9750 9750 7000 7250 6750 7000 8250 2750 5000 9 21 9 14 14 14 7 12 17 23 20 MS TR 5688 6813 7438 7438 4250 9313 1688 6750 7063 5688 3750 33 CSD_R 76 67 67 83 75 84 72 79 86 92 89 and CSD 0 0 SS_RESID MS RESID 144 2250
12. documented using one of the graphical representations below Compare the diagrams with respect to which would best facilitate your task of coding from the design specification FC NS WO AD CSD 10 11 12 Compare the diagrams with respect to ease of manual use consider the initial drawing and subsequent modifications FC NS WO AD CSD Assuming you have to manually draw the diagrams in the sense that they are not automatically generated indicate your overall preference for each diagram where 5 first choice 1 last choice FC NS WO AD CSD Compare the diagrams with respect to their overall economy i e increases in comprehension versus effort to draw them manually FC NS WO AD CSD Assuming you have equivalent automated support to draw each of the diagrams in the sense that the diagrams are automatically generated either by selecting constructs from a menu or by recognizing key words in the code indicate your overall preference for each diagram where 5 first choice 1 last choice FC NS WO AD CSD All assumptions aside indicate your overall preference for each diagram where 5 first choice 1 last choice FC NS WO AD CSD It is not uncommon for individuals and organizations to introduce modifications which they consider to be improvements to standard diagramming tools These changes may be to improve readability to make the diagrams easier to work with manually
13. goal of the GRASP Ada project is to provide the foundation for a CASE computer aided software engineering environment in which reverse engineering and forward engineering development are tightly coupled In this environment the user may specify the software in a graphically oriented language and then automatically generate the corresponding Ada code ADA83 Alternatively the user may specify the software in Ada or Ada PDL and then automatically generate the graphical representations either dynamically as the code is entered or as a form of post processing The GRASP Ada project was divided into three primary development phases followed by two update phases Update 92 and Update 93 the current phase Each of these phases is briefly described below 1 1 Phase 1 The Control Structure Diagram For Ada Phase 1 focused on a survey of graphical notations for software with concentration on detailed level diagrams such as those found in MAR85 TRI89 and the development of a new algorithmic or PDL code level diagram for Ada Tentative graphical control constructs for the Control Structure Diagram CSD were created and initially prototyped in a VAX VMS environment This included the development of special diagramming fonts for both the screen and printer and the development of parser and scanner using UNIX based tools such as LEX and YACC The CSD is described in Section 2 0 1 2 Phase 2 The GRASP Ada Prototype and User Interface During Phase
14. routines which will be X Window System based CSD Screen Fonts The default CSD screen font is a bitmap 13 point Courier to which the CSD graphic characters have been added The font was defined as a bitmap distribution font BDF then converted to SNF format required by the X Window System Four additional screen fonts ranging from 9 to 24 point are user selectable These fonts were 19 later converted to OpenWindows fonts which has since become the version supported in the distribution tar file CSD Printer Fonts CSD Printer fonts were initially developed for the HP LaserJet series These were then implemented as PostScript type 3 fonts and all subsequent font development has been directed towards the PostScript font The PostScript font provides the most flexibility since its size is user selectable from 1 to 300 points 5 3 CSD Generator Future Considerations As indicated above the actual generation of the CSD and the subsequent display of it on the screen or printer are in some ways inseparable For example margins line spacing and indentation of the CSD constructs could be part of the actual generation of the CSD or they could be a function of the display mechanism tightly coupled with a CSD editor Hence there is some degree of overlap in the discussion below regarding generating and displaying the CSD 5 3 1 Generating the CSD Future Considerations Ada 9X The Ada 9X specification will impact the CSD generator in at least
15. these system and architectural level diagrams In this capacity the CSD has the potential to replace traditional non graphical pspecs PDL used in software design and textual source code listings used in implementation testing and maintenance The primary impact of reverse engineering graphical representations will be improved comprehension of software in the form of visual verification and validation V amp V To move the results of this research in the direction of visualizations to facilitate the processes of V amp V numerous additional capabilities must be explored and developed A set of graphical representations that directly support V amp V of software at the architectural and system levels of abstraction must be formulated For example the Object Diagram generator ODgen prototype described earlier is one the components of the GRASP Ada project which would address architectural and system levels of abstraction This task must include an on going investigation of visualizations reported in the literature as currently in use or in the experimental stages of research and development In particular specific applications of visualizations to support V amp V procedures must be investigated and classified Prototype software tools which generate visualizations at various levels of abstraction from source code and PDL as well as other intermediate representations must be designed and implemented Graphically oriented editors must provide capabiliti
16. two important ways 1 New 9X control constructs Additional CSD graphical constructs must be created as appropriate Addressing the new 9X control constructs should be relatively straightforward Numerous examples will need to be diagrammed and evaluated for comprehensibility and ease of implementation and integration into the current set of constructs 2 The Ada 9X specification allows all 256 ASCII character codes These are allowed to facilitate international character sets The use of all 256 character codes presents somewhat more of a challenge In the present prototype ASCII codes above 128 have been used for the CSD graphics characters Several operations in the user interface have freely filtered these character codes during the regeneration process as well as the save as Ada operation Obviously this approach will not be acceptable if the source code itself contains character codes in this range Internal Representation of the CSD Alternatives Several alternatives were considered for the internal representation of the CSD in the Version 3 prototype Each has its own merits with respect to processing and storage efficiency and is briefly described below These alternatives remain under consideration for future versions 1 Single ASCII File with CSD Characters and Text Combined This is the most direct approach and is currently used in the Version 3 prototype The primary advantage of this approach is that combining the CSD c
17. 1F8 0x0E01 0xFFE0 0x0003 0xC3FC 0x1E01 0xFFE0 0x0003 0xC606 0x1601 0x8000 0x0003 0xC606 0x0601 0x8000 0x0003 0xC006 0x0601 0x8000 0x0003 0xC006 0x0601 0x8000 0x0003 0xC006 0x0601 0x8000 0x0003 0xC006 0x0601 0x8000 0x0003 0xC07C 0x0601 0x8000 0x0003 0xC07C 0x0601 0x8000 0x0003 0xC006 0x0601 0x8000 0x0003 0xC006 0x0601 0x8000 0x0003 0xC006 0x0601 0x8000 0x0003 0xC006 0xC601 0x8000 0x0003 0xC606 0xC601 0x8000 0x0003 0xC606 0x0601 0x8000 0x0003 0xC3FC 0x1F81 0x8000 0x0003 0xC1F8 0x1F81 0x8000 0x0003 0xC000 0x0000 0x0000 0x0003 0xC000 0x0000 0x0000 0x0003 0xC000 0x0000 0x0000 0x0003 OxFFFF 0xFFFF OxFFFF 0xXFFFF 0XFFFF 0xFFFF 0xFFFF 0XFFFF help Graphical Representation of Algorithms Structures and Processes row 0 col 40 Edit_Diagrams label Edit Diagrams cmd graspada Diagram_Name msg Control Structure Diagram Editor Diagram_Name label Diagram Name s text 77 All the Hex numbers are the bitmap representation of the CSD icon The line cmd graspada Diagram_Name will specify what command to execute when the user selects CSD icon and clicks on execute The dfe_pspec_edit variable in ToolInfo file should be set as follows so that it invokes GRASP Ada instead of standard Pspec editor dfe_pspec_edit graspada amp The sce_pdl_edit variable in ToolInfo file should be set as follows so that it invokes GRASP Ada instead
18. 2 the prototype was extended and ported to a Sun UNIX environment The development of a user interface based on the X Window System represented a major part of the extension effort Verdix Ada and the Verdix DIANA interface were acquired as potential commercial tools upon which to base the GRASP Ada prototype Architectural diagrams for Ada were surveyed and the OOSD notation WAS89 was identified as having good potential for accurately representing many of the varied architectural features of an Ada software system Phase 2 also included the preliminary design for an architectural CSD DAV90 The aspects of architectural CSD are expected to be integrated into the fully operational GRASP Ada prototype during a future phase of the project 1 3 Phase 3 CSD Generation Prototype and Preliminary Object Diagram Prototype Phase 3 has had two major thrusts 1 completion of an operational GRASP Ada prototype which generates CSDs and 2 the development of a preliminary prototype which generates object diagrams directly from Ada source code Completion of the GRASP Ada CSD prototype CSDgen included the addition of substantial functionality via the User Interface to make the prototype easier to use The User Interface was reworked based on the Athena widget set CSDgen was installed and demonstrated on a Sun workstation at Marshall Space Flight Center Alabama The development of a preliminary prototype for generating architectural object diagrams O
19. 39 90 50 30 10 PREF AUTO 33 30 30 20 61 43 71 49 17 11 PREF GEN 33 30 91 24 24 40 14 51 71 ITEM AVG 363 21 72 13 09 32 50 39 76 The Table 4 shows the difference between the control structure diagram CSD percentage scores and each of the other percentage scores Negative values indicate a lack of preference for the indicated diagram type and a preference for the CSD Note that the NS SELection construct was the only item for which the CSD construct was not preferred on average Item 5 is of particular interest in that it attempts to determine perceived improvements in readability over non graphical pseudo code Tables 1 through 4 provide useful insight and suggest potential significant differences among the preferences However an analysis of variance is required to determine the presence or absence of actual statistical significance 6 3 2 Statistical Analysis for Significance of Preference of Graphical Representations Since the data were taken on a scale of 1 to 5 and thus were not continuous the ordinary Analysis of Variance ANOVA could not be applied Note that ANOVA generally assumes that the data originates from a normal population an assumption that would not be tenable in this case Therefore a nonparametric or distribution free test was used to determine if there were statistically significant differences among the five treatments FC NS WO AD and CSD The determination of significant differences would be made wi
20. 50 27 4375 13 7188 33 0625 0 5333 25 7259 000000007 10 PREF AUTO 61 50 24 8594 12 4297 36 6406 0 5910 21 0324 0000001 11 PREF GEN 64 00 19 7500 9 8750 44 2500 0 7137 13 8362 000011 34 Table 8 shows that the differences among the three treatments FC NS CSD were not significant at the 05 level for the four performance characteristics SEQ SEL ITR and MANUAL but the three treatments differed very significantly with respect to each of the other seven performance characteristics Furthermore the average ranks for CSD far exceeded those for FC and NS in the case of the seven significantly different performance characteristics 6 3 3 Summary of Responses For Evaluation of GRASP Ada The second part of the evaluation instrument was specifically directed at GRASP Ada The items are presented below with a summary of the responses in italics 1 Was the User Interface intuitive Most subjects felt comfortable with the User Interface after several sessions However many expressed the desire for a User Manual 2 What changes would you make to the User Interface Most subjects stated the User Interface was acceptable as is Several expressed a desire to have stickable subwindows from which options are selected These were not available through Athena widgets from which the User Interface was constructed 3 What were the major problems you encountered when using GRASP Ada As one might expect a variety of responses were given for
21. CHM2 Additions 12 10 91 by CHM2 01 17 92 by CHM2 01 27 92 by CHM2 12 12 93 by JHC2 ARGGGGiiiiiiiiiiiiiisiiiiiziciiiiisiciiiiiziciiiiiiiciiiiiziciiiiiziciiiizsiae Purpose of this file To give the installer the proper directions for installing GRASP Ada and for instructing users in adjusting their environments as necessary ARGGGGiiiisiiiiiiizisiiiiiziciiiiisiciiiiiziiiiiiisiiiiiiisiciiiiiziciiiiisiai Installation Requirements GRASP Ada was developed on Sun SPARCstations running SunOS a derivative of UNIX BSD and was intended to run on such systems At the time of this writing it is UNKNOWN whether GRASP Ada will run under any other UNIX workstation environment The file hierarchy when unarchived consumes approximately 1 5 MB of disk space not including the tar file itself The process of making the program and associated object files should require approximately 3 MB more disk space All intermediate products of the make process will be removed almost as soon as their usefulness has expired They will not linger on until the entire make process is complete this will reduce disk space requirements The final GRASP Ada system source executable other needed files will require about 1 5 MB of disk space As stands to reason the installer must have read write and execute search permission in the directory in which he installs GRASP Ada X related requirements a The system on which GRASP Ada run
22. Dgen for Ada source PDL was an effort to determine feasibility rather than to deliver an operational prototype as was the case with CSD generator above The preliminary prototype has indicated that the development of the components to recover the information to be included in the diagram although a major effort is relatively straightforward However the research has also indicated that the major obstacle for automatic object diagram generation is the automatic layout of the diagrams in a human readable and or aesthetically pleasing format A user extensible rule base which automates the diagram layout task is expected to be formulated during future GRASP research 1 4 Update 92 Preliminary Evaluation and User Interface Enhancements Following Phase 3 the Version 3 1 prototype was used in several software engineering classes at Auburn University evaluated and enhanced to create Version 3 2 A preliminary analysis was done on data collected from students and then changes were made to the User Interface to reflect the indicated usage patterns The UNIX man page was drafted to provide online documentation and the installation guide was drafted to provide for limited distribution of the tool And finally the prototype was modified so that it could be invoked from IDE s CASE tool StP with a pspec or PDL file see Appendix B 1 5 Update 93 of the GRASP Ada Update 93 is the most recent phase of the GRASP Ada project and the one described in th
23. GRASP Ada Graphical Representations of Algorithms Structures and Processes for Ada Technical Report 93 14 Update of GRASP Ada Reverse Engineering Tools For Ada Final Report Delivery Order No 21 Basic NASA Contract No NAS8 39131 Department of Computer Science and Engineering Auburn University AL 36849 5347 Contact James H Cross II Ph D Principal Investigator 205 844 4330 cross eng auburn edu December 14 1993 Update of GRASP Ada Graphical Representations of Algorithms Structures and Processes for Ada Reverse Engineering Tools For Ada Final Report Delivery Order No 21 Basic NASA Contract No NAS8 39131 James H Cross II Ph D Principal Investigator December 14 1993 Abstract The GRASP Ada project Graphical Representations of Algorithms Structures and Processes for Ada has successfully created and prototyped a new algorithmic level graphical representation for Ada software the Control Structure Diagram CSD The primary impetus for creation of the CSD was to improve the comprehension efficiency of Ada software and as a result improve reliability and reduce costs The emphasis has been on the automatic generation of the CSD from Ada PDL or source code to support reverse engineering and maintenance The CSD has the potential to replace traditional prettyprinted Ada source code In Phase 1 of the GRASP Ada project the CSD graphical constructs were created and applied manually to several small Ad
24. Several authors have published notable books and papers that address the details of many of these MAR85 TRI88 SHN77 Tripp for example describes 18 distinct notations that have been introduced since 1977 and Aoyama et al describes the popular diagrams used in Japan In general these diagrams have been strongly influenced by structured programming and thus contain control constructs for sequence selection and iteration In addition several contain explicit EXIT structures to allow single entry multiple exit control flow through a block of code as well as PARALLEL or concurrency constructs However none the diagrams cited explicitly contains all of the control constructs found in Ada Graphical notations for representing software at the algorithmic level have been neglected for the most part by business and industry in the U S in favor of non graphical PDL A lack of automated support and the results of several studies conducted in the seventies which found no significant difference in the comprehension of algorithms represented by flowcharts and pseudo code SHN77 have been a major factors in this underutilization However automation is now available in the form of numerous CASE tools and recent empirical studies reported by Aoyami AOY89 and Scanlan SCA89 have concluded that graphical notations may indeed improve the comprehensibility and overall productivity of software Scanlan s study involved a well controlled experiment in which
25. a programs A prototype CSD generator Version 1 was designed and implemented using FLEX and BISON running under VMS on a VAX 11 780 In Phase 2 the prototype was improved and ported to the Sun 4 platform under UNIX A user interface was designed and partially implemented using the HP widget toolkit and the X Windows System In Phase 3 the user interface was extensively reworked using the Athena widget toolkit and X Windows The prototype was applied successfully to numerous Ada programs ranging in size from several hundred to several thousand lines of source code Following Phase 3 e two update phases were completed Update 92 focused on the initial analysis of evaluation data collected from software engineering students at Auburn University and the addition of significant enhancements to the user interface Update 93 the current update focused on the statistical analysis of the data collected in the previous update and preparation of Version 3 4 of the prototype for limited distribution to facilitate further evaluation The current prototype provides the capability for the user to generate CSDs from Ada PDL or source code in a reverse engineering as well as forward engineering mode with a level of flexibility suitable for practical application This report provides an overview of the GRASP Ada project with an emphasis on the current update ACKNOWLEDGEMENTS We appreciate the assistance provided by NASA personnel especially Mr Keith Shackelford w
26. ability of your source code in future software development projects extremely very moderately not very not useful Pascal extremely very moderately not very not useful C 6
27. acteristic Sequence Now that the null hypothesis has been rejected we have either made a correct decision or we have committed a Type I error note a Type I error is committed when an experimenter rejects a true hypothesis The probability of committing a Type I error or the level of significance of the test is determined from amp P F gt F 4 76 The probability or critical level of the test in this case is 0 00117 which indicates there is very little probability of a Type I error Note that the smaller the critical level is the more strongly H can be rejected and the more significant are the differences among the treatments The Friedman s Test was conducted on the complete data set of 20 students for all five treatments and the results are summarized in Table 7 The Friedman s statistic Fy showed that except in the case of performance characteristic MANUAL the differences among the 5 treatments were highly significant lt lt 0 01 In the case of MANUAL the differences among the five treatments were significant at the level amp 0 0133 Table 2 clearly shows that the average rating that the treatments WO and AD received were lower than the other three FC NS and CSD Furthermore 32 students rated FC NS and CSD as opposed to only 20 who rated all five treatments As a result the Friedman s Test was also applied to determine if the three treatments FC NS CSD differed significantly Again each student s rating of FC NS
28. been very useful especially with respect to the user interface The prototype has been prepared for limited distribution GRASP Ada V3 4 An important issue for all software tools in general and graphical representations in particular is evaluation An evaluation based on preference was conducted to provide information on user perceptions of the CSD An experiment was designed and data was collected from software engineering students Statistical analysis indicated highly significant differences among five graphical notations when compared with respect to eleven performance characteristics There was a clear preference for the CSD for seven of the eleven performance characteristics Experience indicates that empirical evaluation of the comprehensibility rather than preference of graphical notations such as data flow diagrams object diagrams structure charts and flowgraphs is difficult However such an evaluation for the CSD and GRASP Ada tool would provide further insight into the role that graphical notations play in the comprehension of software and as a result their potential impact on the overall cost of software The CSD generation component of GRASP Ada has been loosely integrated with IDE s Software though Pictures to replace non graphical process specifications pspecs for data flow diagrams and module PDL for structure charts and object diagrams see Appendix B In fact the CSD becomes a natural detailed level graphical extension for
29. compilation and installation more straightforward The Installation Guide see Appendix A includes sections on the system requirements installation procedure and getting started Although the window based User Interface is relatively intuitive one of items requested most by the students that evaluated the prototype was a User Manual The UNIX Man Page has been updated to provide all necessary user information and an online User Manual which allows the user to view individual topics via the User Interface has been included Investigation of Object Diagrams Existing CASE tools that utilize object diagrams in one form or another were reviewed ObjectMaker by Mark V Systems was acquired for evaluation purposes as an example of currently available software tools that support object diagrams and other architectural level graphical representations The results of the review indicate that the GRASP Ada tool can play an important role as a natural extension to these existing object diagrams and their supporting CASE tools The control structure diagram provides a detailed algorithmic level graphical representation which has statistically significant advantages over other graphical notations PDL and source code The entire spectrum of re engineering oriented CASE tools with respect to functionality and availability has recently been well documented in OLS93 SIT92 Presentation of Update 93 Specifications for this update and the expected results of t
30. currency and multiple exits are useful in modeling the ssolution they do increase the effort required of the reader to comprehend the code The CSD in Figure 4 uses intuitive graphical constructs to depict the point of rendezvous the two nested loops the select statement guarding the accept statement for the task the unconditional exit from the inner loop and the overall control flow of the task When reading the code without the diagram as shown in Figure 3 the control constructs and control paths are much less visible although the same structural and control information is available With additional levels of nesting and increased physical separation of sequential components the visibility of control constructs and control paths becomes increasingly obscure and the effort required of the reader dramatically increases in the absence of the CSD Now that the CSD has been briefly introduced the various CSD constructs for Ada are presented in Figure 5 Each of the CSD constructs should be relatively self explanatory since the CSD is designed to supplement the semantics of the underlying Ada 2 3 Observations The control structure diagram is a new graphical tool which maps directly to Ada and Ada PDL The CSD offers advantages over previously available diagrams in that it is combines the best features of PDL and code with simple intuitive graphical constructs The potential of the CSD can be best realized during detailed design implementation ve
31. d The Control Structure Diagram An Automated Graphical Representation For Software Proceedings of the 21st Hawaii International Conference on Systems Sciences Kailui Kona HA Jan 5 8 IEEE Computer Society Press Washington D C 1988 Vol 2 pp 446 454 J H Cross K I Morrison C H May Generation of Graphical Representations From Source Code Proceedings of the Southeast Regional ACM Computer Science Conference April 18 20 1990 Greenville South Carolina 54 62 J H Cross S V Sheppard and W H Carlisle Control Structure Diagrams for Ada Journal of Pascal Ada and Modula 2 Vol 9 No 5 Sep Oct 1990 J H Cross E J Chikofsky and C H May Reverse Engineering Advances in Computers Vol 35 1992 199 254 F W Calliss J H Cross V Rajlich Panel on Reverse Engineering Proceedings of 5th International Conference on Software Engineering and Knowledge Engineering San Francisco CA June 16 18 1993 544 545 R A Davis A Reverse Engineering Architectural Level Control Structure Diagram M S Thesis Auburn University December 14 1990 J Martin and C McClure Diagramming Techniques for Analysts and Programmers Englewood Cliffs NJ Prentice Hall 1985 40 OLS93 SCA89 SEL85 SHN77 SHU88 SIT92 STA85 TRI89 WAS89 M R Olsem and C Sittenauer Reengineeing Technology Report Volume 1 Software Technology Support Center Hill Air Force Base UT 840556
32. ded into two parts 1 the evaluation of graphical representations of algorithms and 2 the evaluation of GRASP Ada see Appendix C In the first part the first three items solicited background information with respect to familiarity with the five diagram types The next eleven items indicated PCHs by which the subjects were asked to compare the diagrams with respect to a how well each represented sequence selection and iteration b overall readability c improvement in readability as an extension to pseudo code d ease of coding from e ease of manual use f overall preference if drawn manually g overall economy h overall preference with equivalent automated support and finally h overall preference all assumptions aside These eleven items are described in more detail blow in the discussion of results The first part of the instrument concluded with an open ended question soliciting suggestions on how to improve any of the diagrams compared 25 The second part of the evaluation instrument was directed specifically at the GRASP Ada prototype Questions were designed to solicit information regarding the User Interface major problems encountered modifications enhancements desired and the level of coverage provided for Ada during the presentation of GRASP Ada 6 2 The Subjects The evaluation instrument was administered to 33 junior senior computer science and engineering students at Auburn University in the course CSE 422
33. deeply nested if then else constructs represented in structured flowcharts and pseudo code were read by intermediate level students Scores for the flowchart were significantly higher than those of the PDL The statistical studies reported by Aoyami et al involved several tree structured diagrams e g PAD YACC II and SPD widely used in Japan which in combination with their environments have led to significant gains in productivity The results of these recent studies suggest that the use of a graphical notation with appropriate automated support for Ada PDL and Ada should provide significant increases productivity over current non graphical approaches 2 2 The Control Structure Diagram Illustrated Two examples are presented below to illustrate the CSD The first shows the basic control constructs of sequence selection and iteration in Ada These three control constructs are common to all structured procedural languages such as Ada C and Pascal The second example illustrates a more complex control construct the task rendezvous in Ada Figure contains an Ada procedure called SearchArray that searches an array A of elements and counts the number of elements above below and or equal to a specified element Figure 2 contains the CSD for SearchArray which includes the three basic control constructs sequence selection and iteration Although this is a very simple example the CSD clearly indicates the levels of control inherent in the nest
34. e and Preliminary Object Diagram Prototype ubi lele ie ai 2 14 Update 92 Preliminary Evaluation and User Interface Enhancements Qh kee on ahs a le UNA Mane ole seh ob ata NODI NA 2 1 5 Update 93 of the GRASP Ada Concise vetting wet eee Skewes 2 2 0 The Control Structure Diagram 0 0 00 00 0c eee eee 4 2a Backeroundi Aaa 254 Lane ee ie Pe ee ea ee ease ee aw a 4 2 2 The Control Structure Diagram Illustrated 5 29 Observations rien BS 7 2 4 Control Structure Diagram Future Directions 7 3 0 The GRASP Ada System Model 22 curs petra 9 AO User Interface Une inte ea at kae whale a a a R 11 4 1 System WVIMCUOW araisa hn E A e E E 11 4 2 Control Structure Diagram Window 00055 14 4 3 User Interface Future Directions LL 17 5 0 Control Structure Diagram Generator 19 Sl Generating the CSD Lordo irritanti arteatro 19 5 2 Displaying the CSD Screen and Printer 00 4 19 5 3 CSD Generator Future Considerations LL 20 Internal Representation of the CSD Alternatives 0 20 6 0 Evaluation of the Control Structure Diagram and GRASP Ada 25 6 1 The Design of the Experiment use la dia 25 G2 She puDIECIS ck eee eet tees eee RR Le 26 6 3 The Evaluation Results cassettone erences oh eS 8 26 64 Future Directions for Evaluation iii ia a 37 Or CONCIISIONG ats crudi ireland Related iaeee sted 38 REFERENCES ieri IEEE wanton hes de OI 40 APPENDICES oiae ob
35. e included in the file These would be generated by the screen display and print routines as required The advantages of this approach would be most beneficial in an editing mode since the diagram could grow and shrink automatically as additional text source code is inserted into the diagram The extent of required modifications to text edit windows must be considered with this alternative Ada Comments Currently the location of Ada comments is not preserved by the CSD generator Future CSD generators should attempt to preserve the original location to the degree possible Ada Coding Standards Future CSD generators should provide the user with the capability to generate CSDs together with prettyprinted source code according to a prescribed standard This may include conventions for keywords identifiers indentation layout for compound statements placement of comments etc Syntax Error Messages Currently the CSD generator assumes the Ada source code is syntactically correct and makes little attempt to recover from encountered errors or to display meaningful error messages This has been a major source of complaints from users who have used the CSD window to develop software in a forward direction rather than simply reverse engineering existing software Future versions of the CSD generator could address this issue in one of two ways 1 CSD Generator could generate messages during the parse 2 Capabilities of current Ada development en
36. e remainder of this report Since Update 92 the Version 3 2 prototype has undergone continual upgrades which have resulted in Version 3 4 of the prototype The CSD evaluation data collected in the previous update was formally analyzed during Update 93 and the User Manual and Installation Guide have been completed Each of these tasks is briefly described below 1 The GRASP Ada tool was evaluated As part of the ongoing evaluation of GRASP Ada the tool has been and continues to be used in CSE 422 Introduction to 2 2 3 4 5 Software Engineering An evaluation instrument was developed and administered to collect feedback from the students during the Fall 1992 quarter Both the CSD and the GRASP Ada tool were evaluated A preliminary analysis of the data was done during Update 92 During Update 93 a formal statistical analysis was applied to the CSD evaluation data to determine statistical significance if any in the preference for the CSD over other common graphical representations such as the flowchart The results were in fact statistically significant and are described in Section 6 Modification of the GRASP Ada tool As a result of the GRASP Ada tool evaluation described above and in Section 6 numerous modifications and enhancements were made to the User Interface The current User Interface is described in Section 4 The Installation Guide Man Page and online User Manual The make files were streamlined to make re
37. en integrated with an environment such as Apex 10 4 0 User Interface GRASP Ada user interface was developed using the X Window System Version 11 Release 4 X11R4 The X Window System or simply X meets the GRASP Ada user interface requirements of an industry standard window based environment which supports portable graphical user interfaces for application software Some of the key features which make X attractive for this application are its availability on a wide variety of platforms unique device independent architecture adaptability to various user interface styles support from a consortium of major hardware and software vendors and low acquisition cost With its unique device independent architecture X allows programs to display windows on any hardware that supports the X Protocol X does not define any particular user interface style or policy but provides mechanisms to support many various interface styles The specifications and figures that follow are intended to define the look and feel of the GRASP Ada User Interface as well as indicate much of the current and planned functionality of the CSD generator The Man Page provides additional information 4 1 System Window The System window shown in Figure 7 provides the user with the overall organization and structure of the GRASP Ada tool Option buttons include General and Control Structure Diagram These are briefly described below A future button is planned for Object Dia
38. end case s EXCEPTION HANDLER exception when Errl L_ S when Err2 L_ s when Err3 s Lend FOR _ Ss l_ for F in R loop s s end loop Figure 5 Control Structure Diagram Constructs for Ada GUARDED SELECT Si select ps when C1 gt A accept M do L_ s Lend Lo null or when C2 gt accept N do L_ s Lend end select INFINITE LOOP S loop L_ Sr L_ s end loop S LOOP EXIT __ exit when C L_ s end loop S PACKAGE package Y procedure Z Lend Y PROCEDURE procedure X is raise Err function Z return Boolean RENDEZVOUS RECEIVER s Spena accept C do L lend L_ s SELECT s m select ane accept I do L_ s Lend accept J do end select SEQUENCE TASK SPECIFICATION task Y task body Y is accept F do L_ s end loop 3 0 The GRASP Ada System Model 3 1 Overview The major system components of the GRASP Ada system are shown in the GRASP Ada block diagram in Figure 6 Currently the entire prototype is written in the programming language C The User User Interface x Interface was built using the X Window System and includes a special CSD window modified text editor and provides general contro
39. er should have the following line setenv GRASP_HOME name of installation directory graspada Check LD LIBRARY PATH environment variable to ensure it contains the path for the X Athena Widgets libraries i e Xaw Xmu The default is usr lib There should be a link directory in usr lib to the X libraries if they are not in the default library OPTIONAL There is an applications default file in GRASP_HOME lib called XGrasp which determines things like background color button font etc To activate these defaults set the XENVIRONMENT variable as follows A 2 setenv XENVIRONMENT GRASP_HOME lib XGrasp At this point the installer should source the cshrc file initializing the environment variables set above source cshrc To ensure that others will not tamper with various source and other files the installer should change the modes of the files and directories in this way chmod R 755 GRASP HOME or if installer is especially trusting of those in his group chmod R 775 GRASP HOME The installer can now proceed to install GRASP Ada The installation is done via a hierarchy of makefiles so there will be a lot of fast moving messages The installer need not be disconcerted by this Make and install GRASP Ada in this way Note currently only configured for OpenWindows cd GRASP HOME make e all GRASP_VERSION 3 4 WINDOWING SYSTEM 1 The following actions are performed in the make proce
40. es for dynamic reconstruction of the diagrams as changes are made to other diagrams at various levels These graphical representations should provide immediate visual feedback to the user in an incremental 38 fashion as individual structural and control constructs are completed Future directions and specific tasks for the GRASP Ada project have been been described at the end of each of the sections above The current prototype of the CSD generator while only one of set of required visualization tools has clearly indicated the utility of the CSD Future enhancements will only increase its effectiveness as a tool for improving the comprehensibility of software 39 ADA83 AOY89 BAR84 CHI90 CON80 CRO88 CRO90a CRO90b CRO92 CAL93 DAV90 MAR85 REFERENCES The Programming Language Ada Reference Manual ANSI MIL STD 1815A 1983 Approved 17 February 1983 In Lecture Notes in Computer Science Vol 155 G Goos and J Hartmanis eds Berlin Springer Verlag M Aoyama et al Design Specification in Japan Tree Structured Charts IEEE Software Mar 1989 31 37 J G P Barnes Programming in Ada Second Edition Addison Wesley Publishing Co Menlo Park CA 1984 E J Chikofsky and J H Cross Reverse Engineering and Design Recovery A Taxonomy IEEE Software Jan 1990 13 17 W J Conover Practical Nonparametric Statistics Joh Wiley and Sons New York 1980 J H Cross and S V Sheppar
41. eudocode with possibly some spatial rearrangement Rate the diagrams on the extent to which they increase readability over non graphical pseudocode Suppose as a programmer you are given a design specification in which the program logic has been documented using one of the graphical representations below Compare the diagrams with respect to which would best facilitate your task of coding from the design specification Compare the diagrams with respect to ease of manual use consider the initial drawing and subsequent modifications Assuming you have to manually draw the diagrams in the sense that they are not automatically generated indicate your overall preference for each diagram where 5 first choice 1 last choice Compare the diagrams with respect to their overall economy i e increases in comprehension versus effort to draw them manually Assuming you have equivalent automated support to draw each of the diagrams in the sense that the diagrams are automatically generated either by selecting constructs from a menu or by recognizing key words in the code 27 indicate your overall preference for each diagram where 5 first choice 1 last choice 11 All assumptions aside indicate your overall preference for each diagram where 5 first choice 1 last choice The results of the item analysis for these eleven items are summarized in Tables 1 and 2 below Table 1 shows the number of responses fro
42. gram x graspada GRASP Ada V3 4 Options General Control Structure Dagram i f For information or comments contact Version 3 4 James H Cross II Project Director Computer Science and Engineering Auburn University 1993 107 Dunstan Hall Auburn University AL 36849 5347 Funded in part by 205 844 4330 Marshall Space Flight Center crost engaubumedu Figure 7 GRASP Ada System Window 11 General Provides for selection of a printer access to the user manual see Figure 8 Print Allows the user to select among several printers or the user may enter a printer name The list of printers is contained in the fileGRASP printers located in the diectory GRASP_HOME lib User Manual Allows the user to open the GRASP Ada online User Manual which provides an interactive index of topics This information is also contained in the UNIX man page Quit Allows the user to exit the GRASP Ada system fa graspada GRASPjAda 3 4 Options General Control Structure Dagram Print User manual Quit For information or comments contact Version 3 4 James H Cross II Project Director Computer Science and Engineering Aubum University 1993 107 Dunstan Hall Aubum University AL 36849 5347 Funded in part by 205 844 4330 Marshall Space Flight Center crost engaubumedu Figure 8 General Options 12 Control Structure Diagram Allows the user to open one or m
43. haracters with text in a single file eliminates the need for elaborate transformation and thus enables the rapid 20 implementation of prototypes as was the case in the previous phases of this project The major disadvantages of this approach are that it does not lend itself to incremental changes during editing and the CSD characters have to be filtered if the user wants to regenerate the CSD or send the file to a compiler 2 Separate ASCII Files for CSD Characters and Text In this approach the file containing the CSD characters along with placement information would be merged with the prettyprinted source file The primary advantage of the this approach is that the CSD characters would not have to be stripped out if the user wants to send the file to a compiler The major disadvantage of this approach is that coordinating the two files would add complexity to generation and editing routines with little or no benefit As a result this approach would be more difficult to implement than the single file approach and not provide the advantages of the next alternative 3 Single ASCII File Without Hard coded CSD Characters This approach represents a compromise between the previous two While it uses a single file only begin construct and end construct codes are actually required for each CSD graphical construct in the CSD file rather than all CSD graphics characters that compose the diagram In particular no continuation characters would b
44. his update Update 93 were presented informally during a panel session at the 5th International Conference on Software Engineering and Knowledge Engineering SEKE 93 June 16 18 1993 in San Francisco CAL93 The actual results are expected to be presented and published at appropriate professional meetings and or in an appropriate journal following this update phase The following sections describe the control structure diagram the GRASP Ada system model the user interface the control structure diagram generator evaluation of the CSD and prototype and future requirements The overall rationale for the development of the CSD is described in CRO90a CRO90b which were written during Phase 1 A taxonomy and extensive literature review of reverse engineering can be found in CHI90 CRO92 which were written during Phases 2 and 3 2 0 The Control Structure Diagram Advances in hardware and software particularly high density bit mapped monitors and window based user interfaces have led to a renewed interest in graphical representation of software Although much of the research activity in the area of software visualization and computer aided software engineering CASE tools has focused on architectural level charts and diagrams the complex nature of the control constructs and control flow defined by programming languages such as Ada and C and their associated PDLs makes source code and detailed design specifications attractive candidates for g
45. hose guidance has been of great value The following is an alphabetical listing of the team members who have participated in various phases of the project An asterisk indicates the team member worked on the Update of GRASP Ada addressed in this report Principal Investigator Dr James H Cross II Associate Professor Computer Science and Engineering Graduate Research Assistants Richard A Davis Charles H May Kelly I Morrison Timothy A Plunkett Narayana S Rekapalli Darren Tola Statistician Dr Saeed Magssoodloo Professor Industrial Engineering The following trademarks are referenced in the text of this report Ada is a trademark of the United States Government Ada Joint Program Office AdaVision is a trademark of Sun Microsystems Inc Apex is a trademark of Rational ObjectMaker is a trademark of Mark V Systems Inc PostScript is a trademark of Adobe Systems Inc Software through Pictures StP Ada Development Environment ADE and IDE are trademarks of Interactive Development Environments VAX and VMS are trademarks of Digital Equipment Corporation VERDIX and VADS are trademarks of Verdix Corporation UNIX is a trademark of AT amp T TABLE OF CONTENTS LO Jntrod ction ho tiie ee adres oe ant Gn praga elas ve den dd 1 1 1 Phase 1 The Control Structure Diagram For Ada 1 1 2 Phase 2 The GRASP Ada Prototype and User Interface 1 LI Phase 3 CSD Generation Prototyp
46. ich is called by the CSD generation routines during the parse of the Ada source 2 Using the Rational Apex or Verdix VADS library system along with ASIS 3 Using the relational database system of a commercial CASE tool such as StP TROLL USE Of these alternatives the first one may be the most direct approach since it would be the easiest to control The Apex VADS and StP library approaches may be more useful with the addition of object diagram generation and also with future integration of GRASP with commercial CASE tools 24 6 0 Evaluation of the Control Structure Diagram and GRASP Ada An important aspect of any research project is the evaluation of the results In the GRASP Ada project the two primary results were 1 the development of the Control Structure Diagram CSD as a new algorithmic level graphical representation for Ada software and 2 the development of a prototype that automatically generates the CSD from existing Ada PDL or source code Formal statistically based controlled experiments dealing with the comprehensibility of graphical representations of software are difficult to design and conduct Similar difficulties are encountered when attempting design controlled experiments to evaluate CASE tools with respect to improvements in productivity that result from their use The primary difficulty arises from the learning curve that users subjects must overcome For example a year or more may be required to become proficie
47. ile and sends it to the selected printer Quit Closes the CSD window Edit Allows the user to do traditional text editing functions of cut copy paste and search see Figure 12 Cut Allows the user to deleted highlighted text This option is currently disabled and the user must press CTRL W to accomplish a cut Copy Copies the highlighted text to buffer This option is currently disabled and the user must click the right mouse button to accomplish a copy 15 Paste Inserts the previously cut or copied text from the buffer This option is currently disabled and the user must click the middle mouse button to accomplish a copy Search Opens a textedit search window Currently this option is disabled and the user must press CTL s to activate the search window View Currently allows the user to select one of several window font sizes ranging from 9 points to 24 points see Figure 13 The default font size for the CSD window is 13 points Option aA Aa described below also facilitates the selection of font sizes Misc Allows the user to show or hide The CSD character panel set compiler future option invoke interactive User Manual and load window fonts With the CSD character panel visible CSD characters can be inserted directly into the current window primarily for the purpose of experimentation see Figure 14 Show CSD character panel Displays the CSD character set in a panel
48. in the lower section of the CSD Window This option allows the user to position the curser in the edit window and then click on a CSD character in the panel to insert it at the cursor position This is useful for experimenting with alternative graphical constructs and diagram layouts Regenerating the CSD results in the removal of all added CSD characters CSD Window Edit View Mise Cut dev null Copy Paste Search Figure 12 CSD Window Edit Options CSD Window File Edit View Misc Ada esd09 csd11 csd13 esd15 csd18 csd24 dev null Figure 13 CSD Window View Options CSD Window File Edit View Misc Ada Show CSD character panel Hide CSD character panel Set compiler User manual Load window fonts Figure 14 CSD Window Misc Options Hide CSD character panel Removes the panel of CSD characters if it is currently dispalayed Compile Allows the user to enter the name of an Ada compiler to be called from the CSD window Currently the CSD window file options do not support the invocation of a compiler Experience has indicated that it is more practical to invoke 16 the compiler from a separate window In particular GRASP Ada can be called as the Ada editor from an Ada development tool such as Sun s AdaVisio
49. ing Generate CSD under File options In the current version of GRASP Ada generation of the CSD is done on a file level basis where each file contains one or more units When changes are made to the source code the entire CSD for the file involved is regenerated Future versions of GRASP Ada will address incremental regeneration of the CSD in conjunction with editing capabilities in the CSD window The CSD window Options ata CSD Window File Allows the user to select from jera Yiew ME Fi Fano ew ev nu numerous options see Figure 10 including Loaa the following Generate CSD Save Save as New Removes existing file if any ee from CSD window Currently this qQui option is disabled and the user must load a new file or explicitly delete the contents of the CSD window to accomplish the function of new Figure 10 CSD Window File Options Load Loads a CSD file A window is presented that allows the user to navigate among directories and select a file Generate CSD Generates a CSD from source code and or regenerates a CSD after modification When the CSD window is opened and loaded with a source file without a csd extension a separate CSD window is automatically opened to display the CSD when it is generated Note that CSD graphics characters if any are filtered prior to the parse or reparse Currently this option is the same as Regenerate CSD
50. ing of control statements For example at level 1 there are four statements executed in sequence the three assignment statements and the for loop The for loop defines a second level of control which contains a single statement the if statement which in turn defines three separate level 3 sequences each of which contains one assignment statement It is noteworthy that even the CSDs for most production strength procedures rarely contain more than ten statements at level 1 or in any of the subsequences defined by control constructs for selection and iteration This graphical chunking on the basis of functionality and level of control appears to have a 5 procedure SearchArray A in ArrayType Element in ElementType Above Below EqualTo out integer is begin Above 0 Below 0 EqualTo 0 for index in A first A last loop if Element gt A index then Below Below 1 elsif Element lt A index then Above Above 1 else EqualTo EqualTo 1 end if end loop end SearchArray Figure 1 Ada Source for SearchArray procedure SearchArray A in ArrayType Element in ElementType Above Below EqualTo out integer is begin Above Below t EqualTo 0 H for index in A first A last loop if Element gt A index then L Below Below 1 0 0 elsif Element lt A index then Above Above 1 else EqualTo EqualTo 1
51. l and coordination among the other components The control structure diagram generator CSDgen inputs Ada PDL or source code and produces a CSD CSDgen has its own parser scanner built GRASPlib using FLEX and BISON successors of UNIX File System LEX and YACC It also includes its own printer utilities As such CSDgen is a source code graphical reps self sufficient component which can be executed from the user interface or the command line without the commercial components When changes are made to Figure 6 GRASP Ada System Block Diagram the Ada PDL or source code the entire file must be reparsed to produce an updated CSD A CSD editor which will provide for dynamic incremental modification of the CSD is currently in the planning stages The object diagram generation component ODgen is in the analysis phase and has been implemented as a separate preliminary prototype The dashed lines indicate future integration The feasibility of automatic diagram layout remains under investigation Beyond automatic diagram layout several design alternatives have been identified The major alternatives include the decision of whether to attempt to integrate GRASP Ada directly with commercial components For example the Verdix Ada development system VADS and DIANA interface could be used for extraction of diagram information and 2 IDE s Software through Pictures Ada Development Environment IDE S
52. m of squares treatment sum of squares block sum of squares and residual sum of squares respectively Then 20 5 SS Tota Y Y R i 1 j l 300 _ 100 17 2 57 4 900 182 50 where the correction factor is 20 5 2 20 2 R 15 TER so a _ 300 100 100 100 63 5 69 0 48 0sup2 43 5 76 0sup2 SS Treatments R7120 900 900 38 275 SS Block F Ry _ 300 _ 15 15 15 _ggg 20025 _900 0 Biot DZ 5 2100 5 5 j Note that since each block subtotal is SS Blocks SS Students 5 is identically zero Hence we have the following R R 15 then erie j l SS Residual SS Total SS Treatments SS Blocks 144 225 31 Table 6 Results of ANOVA Source df SS MS F Degrees of Sum of Mean Squares SS 4 Freedom Squares SS df SS 76 Total 99 182 50 Tr Stu 1 Treatments 4 38 275 9 56875 5 0423 Tr 1 Blocks 19 0 0 0 0 Stu 1 76 144 225 1 8977 Totl Tr Blk 99 4 19 Lk Residuals E The results of the ANOVA are shown in Table 6 Having computed the F statistic we must now determine if it is sufficiently large to reject the null hypothesis Since the 1 percentage point of the F distribution with 4 and 76 degrees of freedom df is F 4 76 3 577 which is less than F 5 0423 we reject the null hypothesis that there were no significant differences among the FC NS WO AD and CSD with respect to the performance char
53. m the 33 students for each PCH and diagram type FC NS WO AD CSD A number less than 33 indicates the performance characteristic for that particular diagram type was left blank Students were advised orally to leave an item blank if they were unfamiliar with the notation or a particular construct Note that 20 students responded to all performance characteristics for all five diagrams and 32 students responded to all performance characteristics for three of the diagrams FC NS CSD Table 2 contains the averages for the responses computed on the basis of only those items completed Table 1 Item Response Frequencies PCH N items FC NS wo AD CSD 1 SEQ 33 33 33 25 22 33 2 SEL 33 33 33 24 21 33 3 ITR 33 33 33 24 21 33 4 GEN READ 33 33 32 24 20 33 5 EXT P COD 33 33 33 24 21 33 6 CODE FROM 33 33 32 25 22 32 7 MANUAL 33 33 32 23 24 33 8 PREF MANL 33 33 33 26 25 33 9 ECONOMY 33 33 33 25 24 33 10 PREF AUTO 33 33 33 25 23 33 11 PREF GEN 33 33 33 26 25 33 Table 2 Item Analysis for Graphical Representations PCH N items FC NS wo AD CSD 1 SEQ 33 3 21 3 64 2 64 2 32 3 94 2 SEL 33 3 52 4 06 2 46 2 05 3 64 3 ITR 33 3 45 3 48 2 58 2 14 3 91 4 GEN READ 33 3 03 3 38 2 67 2 10 4 24 5 EXT P COD 33 2 85 3 76 2 38 2 48 3 94 6 CODE FROM 33 2 82 3 53 2 60 2 14 4 31 7 MANUAL 33 3 09 3 16 2 61 2 38 3 91 8 PREF MANL 33 3 00 3 30 2 42 2 16 4 15 9 ECONOMY 33 2 70 3 27 2 52 2 00 4 52 10 PREF AUTO 33 3 03 3 52 2 36 2 09 4 55 11 PREF GEN 33
54. n This allows the CSD to be utilized for viewing and editing the Ada source code while taking full advantage of the Ada development environment Ada Displays Ada control constructs and pona OT enables the user to insert them directly into P Ftutie ke Ade fg Reeenerate CSD Quiet monete the current window at the location of the block with declarations j bi case curser see Figure 15 A syntactically correct eee program can be constructed quickly using this for loop gt a guarded select option Figure 16 shows a program structure sla resulting from four clicks on the Ada iffthen else n O infinite loop constructs procedure body loop if then else one for loop and while loop The template package body placeholders can be modified or replaced as sd i procedure body necessary rendezvous select task spec AA AA Allows the user to increase or i decrease the font size for the current window thus shrinking or expanding the overall size of Figure 15 CSD Window with Ada the CSD Constructs Regenerate CSD Allows the user to quickly regenerate a CSD after making changes Currently this option is the same as Generate CSD on the File submenu Quiet mode Verify mode Allows the user to toggle between to modes of regeneration Quiet mode assumes that the existing files should be overwritten during generation or regeneration and Verify mode queries the user before co
55. n and others are under consideration as modifications enhancements Also feel free to comment on each in the space provided 1 least important 7 most important a Integration of CSD generation editing capabilities with a CASE tool such as StP to facilitate development of process pspecs and or module PDL b GRASP Ada User s Manual Error messages resulting from CSD generation d Integration of CSD editing generation with automatic generation of object diagrams to show software architectural design i e the object diagrams indicate the dependencies among a set of CSDs e Spatial options line spacing amount of indentation etc f Direct access to a compiler from the User Interface to facilitate use of the CSD during implementation g Extension of the CSD editor and generator to handle other languages such as C and Pascal i Rate your knowledge of Ada excellent good moderate very little virtually none How useful was the Ada template feature in the CSD Window in producing Ada PDL CSDs extremely very moderately not very not useful 8 What modifications improvements should be made to this feature The time in class spent on Ada and or AdaPDL should have been increased was about right should have been decreased Comments CSD editors and generators are planned for C and Pascal If these tools were available on the network how useful would they be to you with respect to improving the read
56. nce survey instrument should be refined with respect to the performance characteristics The diagram types FC NS WO AD CSD that were compared to the CSD should be reassessed as to whether they are currently used widely in practice Non graphical PDL should also be considered as a treatment in the next evaluation of preference While the analysis of preference data in this research clearly indicated statistically significant differences which heavily favored the CSD a controlled experiment should be done to evaluate actual increases and or decreases in comprehension due to the use of a particular graphical notation or PDL In fact since PDL is in such widespread use as a detailed design language an experiment comparing the comprehensibility of PDL versus the CSD would be perhaps even more appropriate than attempting to compare numerous graphical notations 37 7 0 Conclusions The GRASP Ada project has provided a strong foundation for the automatic generation of graphical representations from existing Ada software The current prototype provides the capability for the user to generate the Control Structure Diagram CSD from syntactically correct Ada PDL or source code in a reverse engineering mode with a level of flexibility suitable for practical application The prototype is being used in two software engineering courses at Auburn University on student projects in conjunction with other CASE tools The feedback provided by the students has
57. nt enough with a software tool to actually realize gains in productivity Thus it may be difficult to compare two CASE tools in a controlled experiment without introducing bias based on familiarity or in many cases the lack of familiarity As a result most evaluation of CASE tools is based on preference surveys in which the user subject is asked to make mental assessments or comparisons of various aspects of the tool s under study This section describes the design of the experiment the subjects that participated in the evaluation of the CSD and GRASP Ada the preference survey instrument that was developed and administered and the results of the analysis of the data collected 6 1 The Design of the Experiment The primary objective of the evaluation was to determine preference if any for the CSD over other similar graphical representations for algorithms The ANSI flowchart FC the Nassi Shneiderman diagram NS the Warnier Orr diagram WO the action diagram AD were selected for comparison MAR84 The experiment was set up as a block design in which each subject was considered a block and each diagram type FC NS WO AD CSD represented a treatment The subjects were to compare the treatments with respect to eleven performance characteristics PCHs A secondary objective of the evaluation was to collect constructive criticisms of the GRASP Ada prototype to set priorities for future enhancements The evaluation instrument was divi
58. ntinuing 43 User Interface Future Directions The User Interface is expected to continue to evolve especially as new functionality is added In particular implementation based on alternate widget sets is under consideration as well as utilization with other window manager The requirements definition and design of the current version were done in a learning mode under a schedule that required an operational prototype to be implemented quickly As a result many of the features such as placement of options are expected to be streamlined considerably However the current prototype is suitable for limited practical application and information collected from current users is expected to have a positive effect on the overall evolution of the prototype Two rather immediate future tasks are described below 1 Move to the Motif widget set Currently the user interface is implemented using the Athena widget set which is provided with the MIT distribution tape of the X Windows System Recently the Motif widget set was adopted as an industry 17 2 standard By moving to Motif the GRASP Ada user interface will have the same look and feel as most commercial CASE tools in the near future Simplification of the User Interface In order to streamline the integration of GRASP Ada with commercial CASE tools the user interface should be simplified The appropriate time to do this would be during the move to Motif described above CSD
59. of standard PDL editor sce_pdl_edit graspada amp SUMMARY OF INTEGRATION STEPS l Copy the ToolInfo and STPmenu spec files from StP library to user s home directory Set the ToolInfo environment variable in cshre file to refer to the ToolInfo file in user s home directory Load the ToolInfo file which is copied into user s home directory into an editor and modify the ToolInfo variable STPMenuSpec as follows STPMenuSpec STPmenu spec To invoke GRASP Ada in place of Pspec editor replace the ToolInfo variable dfe_pspec_edit variable as follows dfe_pspec_edit graspada To invoke GRASP Ada in place of PDL editor replace the ToolInfo variable sce_pdl_edit variable as follows sce_pdl_edit graspada To invoke GRASP Ada as an independent application like DFE SCE etc copy the information provided above in the Structure of the Specification File to STPmenu spec file in user s home directory Appendix C GRASP Ada Evaluation Instrument Name Date optional Evaluation of Graphical Representations for Algorithms Several of the following graphical representations were briefly presented in class flowcharts FC Nassi Shneiderman diagrams NS Warnier Orr diagrams WO action diagrams AD and control structure diagrams CSD During this course which of the above diagrams were presented Check the appropriate responses FC NS WO AD CSD Prior to this course which of these diagrams had you used Check
60. on the menu selection bar of the CSD window described below Figure 11 shows an Ada procedure provided by NASA after the CSD has been generated Save Saves the CSD file with the same name that was loaded Save as Saves the CSD file with a new name This options allows the user to selected the directory and name from a list 14 CSD Window tmp_mnt home cse_h2 cross research nasa_code res_hip csd procedure RCS_HIP subtype TEMP_NJETS_TYPE is INTEGER range 1 16 type thruster_type is array 1 16 of ON_OR_OFF thrusters thruster_type others gt OFF thrusters two_byte_var others gt false thruster_data arr_64 bc_interrupt_status unsigned_word 16 75 function convert_two_byte_var is new UNCHECKED_CONVERSION SOURCE gt TWO_BYTE_VAR TARGET gt UNSIGNED_WORD for INDEX in RCS_ON range loop JET_CMNDC INDEX RCS_ONCINDEX if RCS_ONCINDEX ON then THRUSTERS INDEX 1 true else H THRUSTERS INDEX 1 false end if end loop 1553B thruster data message thruster_data 1 16 0000 Figure 11 CSD Window with Procedure Provided by NASA after CSD Generation Save as Ada Filters the CSD characters from the CSD file and writes to a file with a a extension Print Presents a window which allows the user to select various print options such as point size page numbers and header and then generates a PostScript file ps from the csd f
61. ore CSD windows close all CSD windows that are currently open and generate CSDs in a batch mode see Figure 9 In addition a list of all CSD windows currently open is presented to the user Open CSD window Opens a CSD window and adds it to the list of open CSD windows This list is displayed at the end of the options pop up menu by appending the name of the respective file in each CSD window In Figure 9 the file dev null has been appended which indicates that a CSD window has been opened but no file has been loaded or saved This list allows the user to see quickly how many CSD windows are active and what file is associated with each Close all CSD windows Closes all CSD windows Currently this quits each CSD window without querying the user regarding unsaved changes When a CSD window is quit from the CSD window file option menu the user is queried if changes have been made since the last save Generate CSD Allows the user to generate a set of CSDs by entering the file name using standard wildcard notation For example entering a would generate a CSD for each file with a a extension A CSD generation summary window displays the progress of the generation by listing each file as it is being processed and any resulting error messages The summary concludes with number of files processed and the number of errors encountered The default for each CSD file name is the source file name with csd appended If an error is encoun
62. raphical representation In particular source code should benefit from the use of an appropriate graphical notation since it must be read many times during the course of initial development testing and maintenance The control structure diagram CSD is a notation intended specifically for the graphical representation of algorithms in detailed designs as well as actual source code The primary purpose of the CSD is to reduce the time required to comprehend software by clearly depicting the control constructs and control flow at all relevant levels of abstraction The CSD is a natural extension to existing architectural graphical representations such as data flow diagrams structure charts and object diagrams The CSD which was initially created for Pascal PDL CRO88 has been extended significantly so that the graphical constructs of the CSD map directly to the constructs of Ada The rich set of control constructs in Ada e g task rendezvous and the wide acceptance of Ada PDL by the software engineering community as a detailed design language made Ada a natural choice for the basis of a graphical notation A major objective in the philosophy that guided the development of the CSD was that the graphical constructs should supplement the code and or PDL without disrupting their familiar appearance That is the CSD should appear to be a natural extension to the Ada constructs and similarly the Ada source code should appear to be a natural extension of
63. rea a nre eni pla i CRA nen 42 A Tnst llation G de ridato o la De eh og oe ee eS A 0 B Integrating GRASP Ada with Software through Pictures StP B 0 C Evaluation Instrument Son ene E Se ee Pee a C 0 ill LIST OF FIGURES Figure 1 Ada Source for SearchArray 0 2 20 e ee eee eee ene eae 6 Figure 2 CSD for Ada Task Body Controller LL 6 Figure 5 CSD for SearcliArtay x fen tue eee oa 6 Figure 4 Ada Source for Task Body Controller 0 0 0 0 00002 6 Figure 5 Control Structure Diagram Constructs for Ada 8 Figure 6 GRASP Ada System Block Diagram LL 9 Figure 7 GRASP Ada System Window LL 11 Figure g General Opuons lt 4 4 vee sE SER ara 12 Figure 9 CSD Options o 3 bac besos ea dae eee RS Ree be RES 13 Figure 10 CSD Window File Options LL 14 Figure 11 CSD Window with Procedure Provided by NASA after CSD Generations scalari e RA VOTE ES YS CASA Sees pe 15 Figure 12 CSD Window Edit Options sox vis ove eee ee bee or og 16 Figure 13 CSD Window View Options 16 Figure 14 CSD Window Misc Options cere ee 16 Figure 15 CSD Window with Ada Constructs LL 17 Figure 16 CSD Window with Program Structure Resulting from Clicking on procedure body if then else for loop and while loop 18 iv LIST OF TABLES Table 1 Item Response Frequencies iui pai ee baba ag paia 28 Table 2 Item Analysis for Graphical Representations 4 28 Table 3
64. rification and maintenance The CSD can be used as a natural extension to popular architectural level representations such as data flow diagrams object diagrams and structure charts The GASP Ada prototype described in Sections 4 and 5 provides for the automatic generation of the CSD from Ada or Ada PDL A preliminary statistical evaluation of the CSD is presented in Section 6 2 4 Control Structure Diagram Future Directions The CSD constructs shown in Figure 5 are expected to continue to evolve especially with Ada 9X on the horizon Suggestions for improvements to the individual CSD graphical constructs are continually solicited from users While most of these suggested changes appear to be minor when considered individually their aggregate implementation in the current prototype represents a major rework Theoretically the CSD and its individual constructs are a separate issue from the automatic generation of the diagrams in a production environment However in practice unless CSDs or any other diagrams can be automatically generated they will not be utilized Additional future considerations concerning the overall system model the user interface and the automatic generation of the CSD can be found at the end of Sections 3 4 and 5 respectively ABORT task body P is begin L_ S abort P BLOCK WITH DECLARATIONS s declare INTEGER CASE s case D is when Cl tese when C2
65. ry calls may be optionally suppressed to make the diagram more compact Collapsing the CSD The CSD window should provide the user with the capability to collapse the CSD based on all control constructs as well as complete diagram entities e g procedures functions tasks and packages This capability directly combines the ideas of chunking with control flow which are major aids to comprehension of software An architectural CSD ArchCSD DAV90 can be facilitated by collapsing the CSD based on procedure function and task entry calls and the control constructs that directly affect these 22 calls In future versions of the prototype the ArchCSD will be generated by the display routines from the single intermediate representation of the CSD Color Although general color options such as background and foreground may be selected via the X Windows system color options within a specific diagram were only briefly investigated for both the screen and printer It was decided that these will not be pursued in the Version 3 prototype Printing An Entire Set of CSDs Printing an entire set of CSDs in an organized and efficient manner is an important capability when considering the typically large size of Ada software systems A book format is under consideration which would include a table of contents and or index In the event GRASP Ada is fully integrated with IDE StP ADE the StP Document Preparation System may be utilized for this function Na
66. ry little How useful was the Ada template feature in the CSD Window in producing Ada PDL CSDs extremely very moderately not very not useful 3 53 indicates usefulness of the Ada template was between moderately and very useful What modifications improvements should be made to this feature Many subjects indicated that additional Ada construct templates were needed Only control structures are included presently The time in class spent on Ada and or AdaPDL should have been increased was about right should have been decreased 2 64 indicates the class time spent on Ada was between about right and should have been increased Comments Some subjects indicated that the course CSE 422 should have a more formal emphasis on Ada Other indicated an emphasis on Pascal or C was preferred since prior required courses cover these languages CSD editors and generators are planned for C and Pascal If these tools were available on the network how useful would they be to you with respect to improving the readability of your source code in future software development projects 36 extremely very moderately not very not useful 4 09 indicates a CSD editor generator for C would be between very and extremely useful Pascal extremely very moderately not very not useful 3 36 indicates a CSD editor generator for Pascal would be between moderately and very useful 6 4 Future Directions for Evaluation The current prefere
67. s must have X11R4 or OpenWindows b The system must have the following link libraries Athena Widget Set Xaw Xmu Xt X11 HEEEEAHEEEEEEAEAE EERE EEEEEETTEEEEEETTETEEEEETEEEEERETEEEEERETETEEETT TREE Installation Directions The installer should have an archive file named grasp 3 4 tar or possibly grasp 3 4 tar Z the compressed version He should move this archive file to the directory in which he intends to install GRASP Ada called logically enough the installation directory A typical example directory would be usr local thus the archive file will be in the usr local directory The installer then should unarchive the archive file in this way if compressed version uncompress grasp 3 4 tar Z tar xvBhmof grasp 3 4 tar if uncompressed version tar xvBhmof grasp 3 4 tar This will unarchive an hierarchy of files whose root is a directory called graspada graspada will be a subdirectory of the installation directory i e usr local in the above example ergo usr local graspada The installer will be the owner of all the files and directories in the hierarchy The permissions mode of the files and directories will be set according to the installer s umask code see umask 1 At this point the installer needs to define an environment variable known as GRASP_HOME This variable is necessary to the installation and use of GRASP Ada In his cshrc file the install
68. should include full syntax checking with appropriate error messages Appropriate hooks to and from the compiler and debugger should be considered 2 Appropriate hooks to from the GRASP Ada tool are necessary to facilitate integration with commercially available CASE tools and programming environments 3 GRASP Ada system model should include the capability of running in a stand alone mode in which several CSD windows are to be coordinated are coordinated by a main window similar to the current Version 4 3 prototype 4 GRASP Ada system model should include the capability of running in a single window mode similar to opening an X Windows textedit application 5 Finally GRASP Ada system model should include the capability of running in a single window mode as an extension of a commercial CASE tool For example when clicking on an object or module in an architectural diagram a CSD window could be opened with the PDL or source code represented in a CSD rather than simply text Many commercial tools are competing in this market One of particular interest is Rational s Ada programming development environment for UNIX Apex which has recently been made available to universities Apex is a state of the art environment which supports the Ada Semantic Interface Specification ASIS to facilitate tool integration The GRASP Ada tool with a CSD editor generator could play an integral role in software development maintenance and reengineering wh
69. ss the program executable is constructed the man page s is are constructed the appropriate screen CSD character fonts are constructed and the printer definitions file is constructed The whole process should take about 5 10 minutes ARGGGGiiiisiiiiiiizisiiiiiziiiiiiisiciiiziziciiiiisiciiiiiziciiiizziciiiiiziai User Instructions Once installation is complete no error messages the installer should inform potential users of the following environmental changes The following should appear in the cshrc files of all GRASP users in this order 1 An environment variable GRASP_HOME which is where GRASP the current version is installed namely A 3 setenv GRASP HOME name of installation directory graspada 2 The path wherein is located the GRASP executable namely GRASP_HOME bin in their PATH list 3 The path containing the GRASP Ada man page s GRASP_HOME man in their MANPATH list 4 If users wish to override the X11R4 application defaults for GRASP Ada or whatever defaults that might be in effect then they should set the environment variable XENVIRONMENT to their own custom application defaults file Invocation is by the command graspada or graspada lt filename gt VARA ii iii iii iii cicci ciccio Appendix B GRASP Ada Integrating GRASP Ada with Software through Pictures StP Integrating GRASP Ada with Software through Pictures StP Introduction Software through Pict
70. syntax error is encountered The diagram is simply generated down to the point of the error In the case of Ada PDL non Ada statements must be valid Ada identifiers so that they are treated like procedure calls For example the PDL for search array for largest element could be represented as Search_array_for _largest_element so that the phrase becomes a single identifier As indicated in Section 3 CSDgen was constructed using the UNIX tools LEX to build the lexical analyzer and YACC to build the parser for Ada An Ada grammar was converted to YACC syntax and seeded with calls to action routines which generate the CSD as tokens in the source code are recognized as individual productions in the grammar The current CSDgen prototype builds the diagram directly during the parse by inserting CSD graphics characters into a file along with text To increase efficiency and improve extensibility future versions of the CSDgen prototype may use a more abstract intermediate representation 5 2 Displaying the CSD Screen and Printer Basic display capabilities to the screen and printer were implemented during Phase 2 Screen display is facilitated by sending the CSD file to a CSD window opened under an X Window manager Printing is accomplished by converting the CSD file to a PostScript file and then sending it to a printer Moving to a more abstract intermediate representation in future versions would necessitate the development of a new set of display
71. tP ADE for the display of the object diagrams ObjectMaker by Mark V Systems has also been reviewed and is a strong contender as a basis for generating and displaying object diagrams ObjectMaker was recently ported to UNIX from a PC platform and unfortunately it appeared to be somewhat unstable in its current state The GRASP Ada library component GRASPIlib allows for coordination of all generated items with their associated source code The current file organization uses standard UNIX directory conventions as well as default naming conventions For example all Ada source files end in a the corresponding CSD files end in a csd and the corresponding print files end in a csd ps In the present prototype library complexity has been keep to a minimum by relying on the UNIX directory organization Its purpose is to facilitate navigation among the diagrams and the production of sets of diagrams 3 2 System Model Future Directions The GRASP Ada tool was conceived to be self sufficient reverse engineering tool that would generate control structure diagrams primarily and then architectural level diagrams e g object diagrams secondarily all from Ada source code or PDL An alternative to the model presented in Figure 6 would be to concentrate on issues that improve the integration capabilities of GRASP Ada with commercially available CASE tools and programming environments The following are potential tasks 1 A CSD editor generator
72. tered an extension of err is used As the CSDs are generated the GRASP library is updated which currently consists of populating a specified directory with file images of the CSDs Since GRASP Ada is expected to be used to process and analyze large existing Ada 5j graspada GRASP Ada V3 4 Options General Control Structure Dagram A PT we A 7 Open CSD window Close all CSD windows Generate CSD jdevjnull For information or comments contact Version 3 4 James H Cross II Project Director Computer Science and Engineering Auburn University 1993 107 Dunstan Hall Auburn University AL 36849 5347 Funded in part by 205 844 4330 Marshall Space Flight Center cross eng auburn edu Figure 9 CSD Options 13 software systems consisting of perhaps hundreds of files the option to generate a set of CSDs in batch mode is particularly useful 42 Control Structure Diagram Window The CSD window shown in Figure 10 with file options displayed provides the user with capabilities for generating and viewing a CSD for an Ada PDL or source file Multiple CSD windows may be opened to access several CSD files at once CSD file names and their associated directory paths are selected under the File option and displayed at the top of each window Figure 11 shows a CSD window after a procedure provided by NASA has been loaded and the CSD generated by clicking Regenerate CSD on the menu selection bar or by click
73. th respect to all eleven PCHs Each student responded to at least 3 of 5 treatments i e each student ranked at least 3 of the treatments FC NS WO AD and CSD Therefore each student is considered as 29 a block and the relevant model is the randomized complete block design The appropriate ANOVA nonparametric test is the Friedman s Test CON80 This test analyzes ranked data for a complete randomized block design Since only 20 students responded to all 5 treatments initially the test for significance was done for significant differences among all five treatments The null hypothesis is as follows Ho There are no significant differences among the 5 treatments versus the alternative H There are significant differences For the sake of illustration the data for the PCH SEQ is provided in Table 5 below Table 5 Data for Performance Characteristic SEQ PCH 1 Student Responses Assigned Ranks Record STU_NO FC NS WO AD CSD FC R NS_R WOR AD RCSD R 1 2 1 2 5 4 3 1 0 2 0 5 0 4 0 3 0 12 3 5 1 2 4 3 5 0 1 0 2 0 4 0 3 0 23 7 5 4 4 2 3 5 0 3 5 3 5 1 0 2 0 34 8 5 3 2 1 4 5 0 3 0 2 0 1 0 4 0 45 11 5 3 3 3 5 4 5 2 0 2 0 2 0 4 5 56 14 3 3 1 2 4 3 5 3 5 1 0 2 0 5 0 67 15 5 4 3 3 3 5 0 4 0 2 0 2 0 2 0 78 16 4 2 3 1 5 4 0 2 0 3 0 1 0 5 0 89 17 3 4 2 2 3 3 5 5 0 1 5 1 5 3 5 100 19 1 4 2 2 5 1 0 4 0 2 5 2 5 5 0 111 20 1 5 4 2 3 1 0 5 0 4 0 2 0 3 0 122 21 3 3 3 3 3 3 0 3 0 3 0 3 0 3 0 133 Z3 4 2 3 3 5 4 0 1 60 2 5 265 pil 144 25 1 5 3
74. the appropriate responses FC NS WO AD CSD Were any of the diagrams used in a professional setting Check the appropriate responses FC NS WO AD CSD Based on the experience you have gained by using these diagramming tools to represent algorithms you are asked to assign a rating to each of the diagrams with respect to a specific comparison among the diagrams You may assign the same rating to more than one diagram for a given comparison Select your ratings from the following scale and enter them as indicated below 5 best most first choice 4 3 moderate 2 1 worst least last choice Compare the diagrams with respect to how well each shows sequence FC NS WO AD CSD Compare the diagrams with respect to how well each shows selection FC NS WO AD CSD Compare the diagrams with respect to how well each shows iteration FC NS WO AD CSD Compare these diagrams with respect to overall readability consider reading someone else s code FC NS WO AD CSD Each of these tools can be used with informal pseudocode as opposed to actual statements in a programming language and as such can be thought of as a graphical extension to pseudocode with possibly some spatial rearrangement Rate the diagrams on the extent to which they increase readability over non graphical pseudocode FC NS WO AD CSD Suppose as a programmer you are given a design specification in which the program logic has been
75. the diagram This has resulted in a concise compact graphical notation which attempts to combine the best features of diagraming with those of well indented PDL or source code 2 1 Background Graphical representations have been recognized as having an important impact in communicating from the perspective of both the writer and the reader For software this includes communicating requirements between users and designers and communicating design specifications between designers and implementors However there are additional areas where the potential of graphical notations have not been fully exploited These include communicating the semantics of the actual implementation represented by the source code to personnel for the purposes of testing and maintenance each of which are major resource sinks in the software life cycle In particular Selby SEL85 found that code reading was the most cost effective method of detecting errors during the verification process when compared to functional testing and structural testing And Standish STA85 reported that program understanding may represent as much as 90 of the cost of maintenance Hence improved comprehension efficiency resulting from the integration of graphical notations and source code could have a significant impact on the overall cost of software production Since the flowchart was introduced in the mid 50 s numerous notations for representing algorithms have been proposed and utilized
76. this item Most were as a result of several known bugs which have since been removed Some simply indicated improper use of the prototype and or a lack of expertise in Ada Again many expressed the desire for a User Manual 4 Rank the following items in order of importance in the prototype Note some of these items are available in the current version and others are under consideration as modifications enhancements Also feel free to comment on each in the space provided 1 least important 7 most important The overall rank of the items is indicated a 4 69 Integration of CSD generation editing capabilities with a CASE tool such as StP to facilitate development of process pspecs and or module PDL b 4 84 GRASP Ada User s Manual Gi 4 84 Error messages resulting from CSD generation 35 5 excellent good moderate very little 8 d 4 47 Integration of CSD editing generation with automatic generation of object diagrams to show software architectural design i e the object diagrams indicate the dependencies among a set of CSDs e 2 81 Spatial options line spacing amount of indentation etc f 4 22 Direct access to a compiler from the User Interface to facilitate use of the CSD during implementation g 5 19 Extension of the CSD editor and generator to handle other languages such as C and Pascal Rate your knowledge of Ada virtually none 1 76 indicates knowledge of Ada was between virtually none and ve
77. tion of the CSD may be desirable in an operational setting The ASIS cited above may offer a bridge for these incremental changes to the CSD 5 3 2 Displaying the CSD Future Considerations Layout Spacing The general layout of the CSD is highly structured by design However the user should have control over such attributes as horizontal and vertical spacing and the optional use of some diagramming symbols In the current Version 3 CSDgen prototype horizontal and vertical spacing are not user selectable They are a part of the CSD file generation and are defaulted to single spacing with 80 characters per line In order to change these e g from single to double spacing the CSD file would have to be regenerated In future versions of the prototype these options are expected to be handled by the new display routines and as such can be modified dynamically without regenerating the CSD file Vertical spacing options will include single double and triple spacing default is single Margins will be roughly controlled by the character line length selected either 80 or 132 characters per line default is 80 Indentation of the CSD constructs has been a constant three blank characters Support for variable margins and indentation is being investigated in conjunction with the new display routines In addition several display options involving CSD graphical constructs are under consideration For example the boxes drawn around procedure and task ent
78. ures Data Flow Diagram Editor Structure Chart Editor Integration Procedure ToolInfo The Main Menu Specification File Structure of the Specification File GRASP label GRASP image Format_version 1 Width 64 Height 64 Depth 1 Valid bits per item 16 OxFFFF 0xFFFF 0xFFFF 0xFFFF 0xFFFF 0xFFFF 0xFFFF 0xFFFF 0xC000 0x0000 0x0000 0x0003 0xC000 0x0000 0x0000 0x0003 0xC000 0x0FFF 0xFFFF 0xFC03 0xC000 0x0FFF 0xFFFF 0xFC03 0xC000 0x0C00 0x0000 0x0003 0xC000 0x0C00 0x0000 0x0003 0xC000 0x0C00 0x0000 0x0003 0xC000 0x0FFF OxFFFF 0xFC03 0xC000 OxOFFF OxFFFF 0xFC03 0xC000 0x0001 0x8000 0x0003 0xC000 0x0001 0x8000 0x0003 0xC000 0x0001 0x8000 0x0003 0xC000 0x0001 0x8000 0x0003 0xC3E0 0x0001 0x8000 0x0003 0xC7F0 0x0001 0x8060 0x0003 0xCC30 0x0001 0x80F0 0x0003 0xCC01 0xE001 0xFF80 0x0003 0xCC03 0xF001 0xFF98 0x0003 0xCC07 0x3801 0x8198 0x0003 0xCC07 0x1BE1 0x819F 0xE003 0xCC03 0x83F1 0x819F 0xE003 0xCC01 0xC339 0x8198 0x0003 0xCC30 0xE319 0x8198 0x0003 0xC7F0 0x7319 0x819F 0xE003 0xC3E6 0x3B19 0x819F 0xE003 0xC007 0x3B19 0x8198 0x0003 0xC003 0xF319 0x8198 0x0003 0xC001 0xE319 0x819F 0xE003 0xC000 0x0339 0x819F 0xE003 0xC000 0x03F1 0x8198 0x0003 0xC000 0x03E1 0x80F0 0x0003 0xC000 0x0001 0x8060 0x0003 0xC000 0x0001 0x8000 0x0003 0xC000 0x0001 0x8000 0x0003 0xC000 0x0001 0x8000 0x0003 0xC000 0x0001 0x8000 0x0003 0xC000 0x0001 0x8000 0x0003 0xC000 0x0001 0x8000 0x0003 0xC000 0x0001 0x8000 0x0003 0xC
79. vigating among CSDs and Object Diagrams A GRASP library is required to provide the overall organization of the generated diagrams The current file organization uses standard UNIX directory conventions as well as default naming conventions For example all Ada source files end in a or ada the corresponding CSD files end in a csd and the corresponding print files end in a csd ps In the present prototype library complexity has been keep to a minimum by relying on the UNIX directory organization In future versions a GRASP library entry will be generated for each procedure function package task task entry and label The library entry will contain minimally the following fields identifier note unique key should be composed of the identifier scoping scoping visibility type procedure function etc parameter list to aid in overload resolution source file file name line number note the page number can be computed from the line number CSD file file name line number OD file file name Referenced by list References to list Alternatives for generation and updating of the library entries include the following 1 During CSD generation the library entry is established and the entry is updated on subsequent CSD generations 23 2 During the processing via the ASIS of DIANA nets Alternatives for implementing the GRASP library include the following 1 Developing an Ada package or equivalent C module wh
80. vironments could be utilized to return the line number and descriptive message whenever an error is encountered during syntax checking and or compiling These could be sent to the CSD window opened 21 on the file or unit in question and the offending element could be highlighted awaiting corrective action from the user Direct Generation Using the Ada Semantic Interface Specification ASIS If tight coupling and integration with a commercial Ada development system such as Rational Apex or Verdix VADS is desired then the ASIS may provide for the direct generation of the CSD from the DIANA net or other underlying intermediate representation produced as a result of compilation This would require a layer of software which traverses the DIANA net and calls the appropriate CSD primitives as control nodes are encountered This approach would eliminate the possibility of directly editing the CSD since the DIANA interface does not support modifying the net only reading it In practice it may prove more efficient to allow the CSD generator to simply reparse the entire compilable unit being edited Incremental Changes to the CSD In the present prototype there is no capability for incrementally modifying the CSD When the CSD or source code is modified in the CSD window the CSD must be regenerated by reparsing the entire file While this has been sufficient for prototyping especially for small programs editing capabilities with incremental modifica

GRASP/Ada - Auburn University

Contents

Download Pdf Manuals

Related Search

Related Contents