Syntactic methods

Often, syntactic methods are used when formal methods are not an option. In non-mission-critical systems, formal methods may prove to be too expensive for the benefit they provide. The costs of modelling, personnel, execution, and development may often outweight the benefits gained by preventing possible failures.

Syntactic methods are often a simpler, and more importantly, cheaper alternative to formal methods. This approach revolves around the use of a abstract dependency graph which is created from the system in question. A abstract dependency graph is a directed graph, a graph of vertices connected by one-way edges. Most often, the vertices and edges of the graph represent the inputs and outputs of functions in, or components of, the system.

By looking at the created abstract dependency graph, the developer can detect syntactic anomalies (or Preece anomalies) in the system. While anomalies are not always defects, they often provide clues to finding defects in a system. Therefore, the anomalies in a system help point the developer in the right direction in finding defects.

There are four main types of anomalies:

• Redundancies - A chunk of the graph is redundant if its terminals can be reached if the chunk is removed from the graph
• Conflicts - A system contains conflicts if the same inputs can infer differenct outputs
• Circularities - A loop in the graph indicates a circularity in the system
• Deficiencies - A chunk is deficient if a subset of inputs leads to no terminals

While anomalies often point to defects, they can just as easily, reflect normal intended functionality in the system. It is up to the developer to look into anomalies in order to determine whether they are clues to problems or simply false alarms.

By creating a visual directed graph of a system, there are several obvious visual flags that indicate the above anomalies:

• a sub-graph with no input is probably missing something important
• while looking at the transitive closure of a system (all nodes downstream from a node), a node in its own transitive closure indicates a circularity
• while looking at the transitive closure of a system, subsumption between pairs of rows indicates redundancy
• conflicts are somewhat more difficult as they become more semantic than syntactic

When formal methods prove too costly, a system can be checked solely on its syntax. This is not as thorough, as it only looks at a system on a surface level. However, it does give a developer many clues as to where a system's defects may lie.