PAT: a knowledge-based program analysis tool

A Knowledge-Based Approach to Program Understanding

“…Most of the automated program understanding approaches produce program documentation which is, more or less, in the form of structured natural language text [1,2,4,6,7,8,10,12,13]. Such informal documentation gives expressive and intuitive descriptions of the code.…”

“…Heuristic-based object-oriented analysis: The knowledge-based Program Analysis Tool (PAT) designed by Harandi and Ning [6,7,54] uses an objectoriented framework to represent programming concepts and a heuristic-based recognition mechanism to derive abstract functional concepts from the source code.…”

“…As a result, different automation approaches have been developed. Some of these approaches are: graph-parsing [1,2]; top-down analysis using the program's goals as input [3,4,5]; heuristic-based object-oriented recognition [6,7]; transformation of a program into a semantically equivalent but more abstract form with the help of plans and transformation rules [8,9]; and decomposition of a program into smaller more tractable parts using proper decomposition [10] or program slicing [11]. Even though these approaches demonstrate the feasibility and usefulness of the automation of program understanding, they lack some important features.…”

Improving program comprehension by combining code understanding with comment understanding

“…For this reasons, heuristic approaches acquired new importance [27]. One such heuristic approach was that of Harandi and Ning [9]. In their heuristic concept recognition approach, lower level events such as statement, condition, loop, search, sort, etc., were combined within a knowledge base to form higher level events.…”

A knowledge-based approach to the analysis of loops

“…These knowledge-based approaches are all implemented, to varying degrees, in automatic analysis systems. Some of these approaches are: graph-parsing [38], [50]; topdown analysis using the program's goals as input [231, [241; top-down analysis using a functional representation of programs that relates the program code and goals to a proof of correctness [6], [33]; heuristic-based object-oriented recognition [15], [16]; transformation of a program into a semantically equivalent but more abstract form with the help of plans and transformation rules [27], [29], 1461; and decomposition of a program into smaller more tractable parts using control flow analysis [17] or program slicing 1181. Even though these approaches demonstrate the feasibility and usefulness of the automation of program understanding, they lack some important features.…”