Spring State Machine Explorer

Static analysis tool to extract and visualize Spring State Machines with deep code traceability.

Core Modules

• state_machine_exporter: Java AST analyzer. Scans source code, resolves local Maven/Gradle dependencies, and extracts logic (Lambdas, Actions, Guards) into JSON/PUML.
• state_machine_exporter_html: Generates an interactive HTML portal with SVG diagrams and rich tooltips.

Usage

Analysis is source-first: the exporter reads project source and build metadata (pom.xml, settings.gradle) statically. It does not run Maven or Gradle against the project being analyzed.

1. Extract Metadata (JSON)

./gradlew :state_machine_exporter:run --args="-i ./my-project -o ./out -f json"

Common options:

Flag	Description
-i, --input	Project directory to scan (defaults to .)
-j, --json	Re-render from an existing JSON export instead of scanning source
-o, --output	Output directory (default ./out)
-f, --format	Comma-separated formats: json, dot, puml, scxml
-p, --profiles	Spring profiles for property placeholder resolution
--event, --state	Enum formatting: fn (default), fqn, sn
--debug	Verbose analysis logging and unresolved chain diagnostics
--no-diamonds	Render choice/junction states as regular nodes (default: diamonds)
--include-generated-sources	Also scan target/ / build/ generated .java trees (default: on)
--no-inline-accessors	Disable accessor indexing/inlining (enabled by default)

2. Generate Interactive Portal (HTML)

./gradlew :state_machine_exporter_html:run --args="-i ./my-project -o ./out_html"

HTML-specific options:

Flag	Description
-f, --flows	Custom flows.json path
-m, --machine	Export only machines whose name contains this filter
--open	Open the generated portal in the default browser
--no-metadata-pane	Hide the left metadata pane (entry points, flows)
--no-diamonds	Render choice pseudostates as normal nodes
--debug	Same diagnostic mode as the core exporter
--include-generated-sources	Same generated-source scanning as the core exporter
--no-inline-accessors	Disable accessor indexing/inlining (enabled by default)
-p, --profiles	Spring profiles for property resolution

Business Flows

Define sequence of events in src/main/resources/flows.json:

[
  {
    "name": "Order Success",
    "description": "Happy path for order placement",
    "steps": ["PAY", "CHECK_AVAILABILITY->PENDING", "SHIP"]
  }
]

JSON Structure

• metadata.entryPoints: REST, WebFlux, and JMS entry points.
• metadata.callChains: Trace from API call to machine trigger (sendEvent).
• transitions.actions[].internalLogic: Extracted source code/lambdas.
• metadata.properties: Dictionary of all detected Spring profiles.

Static analysis: accessor inlining

During scan, the exporter builds an accessor index of trivial getters, setters, record components, and constant-returning methods. At analysis time, AccessorResolver uses this index first (O(1) lookup), then falls back to constructor/field tracing and method-body evaluation.

This improves resolution of DTO/event chains such as request.getPayload().getType() without running Maven or Gradle on the target project.

# Enabled by default; disable for debugging or A/B comparison
./gradlew :state_machine_exporter:run --args="-i ./my-project -o ./out --no-inline-accessors"

Pipeline components live under analysis/pipeline/ (AccessorResolver, ResolutionBudget, MethodInvocationUnwrapper, FieldInitializerFinder, LibraryUnwrapRegistry).