Package sootup.spark

Class Spark

java.lang.Object
sootup.spark.Spark
public class Spark extends Object
Main access point to SPARK's functionality.

Exposes the call graph (CHA up-front, supplied by the client, or built on-the-fly when SparkOptions#isOnFlyCallGraph() is set) and a PointsToAnalysis over the resulting PAG. Lower-level components such as the solver are package-private; clients should not bypass this facade. 

   Spark spark = Spark.builder()
       .view(view)
       .entryPoints(List.of(mainSig))
       .sparkOptions(SparkOptions.builder().onFlyCallGraph(true).build())
       .build();
   PointsToAnalysis pta = spark.getPointsToAnalysis();
   CallGraph cg = spark.getCallGraph();

To reuse a call graph the client already computed instead of having SPARK run CHA, pass it via callGraph(...). This is ignored (with a warning logged) when onFlyCallGraph is on, since OTF mode grows its own call graph incrementally from the entry points instead of consuming one.

Spark has no <clinit>-handling knob of its own; this is also the way to control it in the non-OTF path. Pre-build the graph with ClassHierarchyAnalysisAlgorithm's seedEntryPointClinits/sootup.callgraph.scope.VirtualCallResolver constructor and pass it via callGraph(...) instead of letting Spark run CHA with its defaults.

reflectionModel(...) (default ReflectionModel.none()) makes reflective calls explicit in every body Spark reads. dynamicInvokeResolver(...) (default DynamicInvokeResolver.bootstrapMethodHandles()) decides invokedynamic targets; values captured by a lambda flow into its body. When passing a pre-built callGraph, build it with the same model and resolver instances, else reflective/invokedynamic edges and PAG disagree.

  • Method Details

    • solve

      public void solve()
      Builds the pointer assignment graph and (in OTF mode) the call graph. Invoked automatically by the accessors; calling it explicitly is optional and idempotent.

    • getCallGraph

      public CallGraph getCallGraph()
      Returns the call graph: the one passed to the builder, one built via class-hierarchy analysis at construction time, or one grown on-the-fly during solve() when SparkOptions#isOnFlyCallGraph() is set.

    • getPointsToAnalysis

      public PointsToAnalysis getPointsToAnalysis()
      Returns a points-to analysis over the solved PAG. The result is cached.

    • getPag

      public PAG getPag()
      Returns the underlying pointer assignment graph — intended for inspection and debugging.

    • getSparkOptions

      public SparkOptions getSparkOptions()