Extraction Strategy: Dependencies and FQNs

Goal: Creation of DEPENDS_ON relations between the various concept nodes

Strategy:

1. LCE: While traversing the AST register all referenced identifiers as dependencies (with source and target FQN) in a central data structure
2. LCE: After finishing the traversal aggregate all detected dependencies and store them as concepts
   • Note: as FQNs can’t always be resolved fully during the traversal, a resolution of them has to precede the dependency aggregation
3. jQA Plugin: Create dependencies using the stored global FQNs
4. jQA Plugin: Fill in all transitive dependencies via Cypher queries

Central Components:

• LCEDependency: concept that represents a direct dependency observed in the code
  • for transitive dependencies see DependencyResolver
• DependencyResolutionProcessor: processor that resolves FQN references and collects registered dependencies within the AST
  • executed on the root node of the AST of each source file
  • registers Local Contexts that serve as indexes for registering declarations and dependencies within the module as well as schedule FQN resolutions at the end of the AST traversal
  • provides static utility methods for the construction of FQNs and the interaction with the Local Contexts from above
• IdentifierDependencyProcessor: processor that registers dependencies for all referenced identifiers within the code that are not part of one of the other processed declaration constructs
  • MemberExpressionDependencyProcessor handles dependency registration specifically for member expressions (e.g. myObj.b)
• ScopeProcessor/DeclarationScopeProcessor: processors that register (un)named FQN scopes at all relevant points of the AST traversal to aid FQN construction
• DependencyResolver (Java): creates DEPENDS_ON relations for the extracted direct dependencies and fills in all transitive dependencies via Cypher queries
  • in a final step all DEPENDS_ON relations are aggregated