Design and Implementation of a Lightweight Workflow Engine

The marketing automation platform faces long delivery cycles, duplicated development, and high maintenance costs, prompting the need to separate business logic from control flow.

Workflow (Workflow) is an industrial‑grade solution standardized by the Workflow Management Coalition (WfMC) that abstracts business rules between process steps, allowing computers to handle them automatically.

Key characteristics of a workflow engine include visual process building, business orchestration reuse, and separation of business logic from control flow.

Workflow types are categorized as sequential, state‑machine, and rule‑driven, each with distinct execution models; state‑machine workflows are often combined with conditions and rules for complex scenarios.

The article compares workflow engines with state machines, highlighting that state machines are event‑driven and suitable for simple, low‑complexity problems, while workflow engines better handle complex business process automation.

Popular open‑source workflow engines are surveyed, noting issues such as extensive configuration tables and the need for custom adaptations.

To address these challenges, a lightweight self‑developed workflow engine is introduced, focusing on three core modules: process template creation, process instance publishing, and task execution.

Core service interfaces are defined as follows:

public interface FlowEngine {
    FlowInstance startInstance(String processDefKey, Map<String, Object> args);
    void execInstance(Long instanceId, Map<String, Object> args) throws FlowAuthorityException;
    ProcessService process();
    InstanceService instance();
    TaskService task();
}

public interface ProcessService {
    void create(String definition);
    void deploy(String fileName);
    FlowProcess getProcessByDefKey(String processDefKey);
}

public interface InstanceService {
    FlowInstance createInstance(FlowProcess process, Map<String, Object> args);
    void exec(Long instanceId);
    FlowInstance getById(Long instanceId);
}

public interface TaskService {
    FlowTask createTask(TaskModel taskModel, Execution execution);
    FlowTask complete(Long taskId, Map<String, Object> args);
    FlowTask getActiveTask(Long instanceId);
    FlowHistTask getLastDoneTask(Long instanceId);
}

The engine initialization loads configuration, parses process definitions, and caches nodes; runtime creates and binds process instances, persists node states, and executes tasks.

Further considerations include parsing performance, version compatibility, plugin‑based node orchestration, and execution monitoring with alerting.

In conclusion, the workflow engine abstracts process templates to instantiate real activities, offering visualization, orchestration, and business‑control separation, and the presented design provides a practical, extensible solution for complex, frequently changing business workflows.