Smart Process API (SPA)

The semantic process API (SPA) is a vocabulary and engine that enables managing semantic descriptions of process/data-models and its instances. SPA becomes the globally accepted academic standard and reference implementation. -- InES

Coming from the original vision about the SPA, the view slightly shifted towards more generality. While the main focus for the first development phases are processes, the goal broadend to also support other domains. Thereby, three Use Cases are the current implementation drivers; namely Semantic Process integration over format and tool borders, Analytics like Process Mining and Process Optimization by combining usage data with process models.

Semantic Process Integration

Different tools may have different means to express processes. For example, they might differ in formats and interfaces. From a user perspective this is a rather unsatisfying situation, because processes may span over several systems implying additional manual orchastration effort. The SPA taggles this problem by consuming different interfaces and formats. Thereby, a mapping between the concepts of two formats is provided and, thus, the formats are integrated (cf. ASCII art below).

+------------+         +------------+          +------------+
|            |         |            |          |            |
|  Format A  +--------->  Format B  |+--------->  Format C  |
|            | Mapping |            |  Mapping |            |
+------------+         +------------+          +------------+

Due to the transitive nature of the mappings, data in Format A is instantly available in terms of Format B .

Domain

Different project s combine one or more vocabularies or schema s to express their problem domain. Data from different sources must adhere to the schema constraints in order to be drawn from such a domain. This pool of data is further divided into bucket s. For some projects, such a distinction is not useful. In the process domain, the schemas could be BPMN or YAWL . Then, stored data must be expressed in terms of BPMN or YAWL . The following ASCII Art depicts the domain and its relationships.

+------------+         +------------+
|            |         |            |
| Repository +--------->   Schema   |
|            | contains|            |
+------+-----+         +-----^------+
       |                     |
       |                     |
       |                     |
+------v-----+               |
|            |   links       |
|   Project  +---------------+
|            |
+------+-----+
       |
       |contains
       |
+------v-----+
|            |
| Data pool  |
|            |
+------+-----+
       |
       |consists of
       |
+------v-----+
|            |
|Data bucket |
|            |
+------------+

Core

The core module implements the domain and provides means for alteration. The package structure is depicted below.

core
├── application
├── domain
│   └── model
├── io
│   └── file
│       └── xes
├── persistence
└── storage
        └── jena

Package domain.model is the bounded context and contains the codified domain model. SPA as the sole application service is contained in the application package. Further, a SpaBuilder is available which implements a DSL for configuring SPA instances. For example, SpaBuilder.local().memory().shared() creates an instance of the SPA which stores its data in a runtime shared memory location. SPA itself provides CRUD functionalities for the domain.

Package io contains the Importer and Exporter interfaces for importing data from and to Jena's Model inteface and, thus, RDF. Further, classes ImporterSupport and ExporterSupport are containers responsible for managing groups of Importers and Exporters. All file-based implementations are stored in subpackage file.

Package persistence contains all implementations responsible for mapping Java instances to RDF and vice-versa. All Repository implementations rely on the Transformation class which implements a DSL for RDF <-> Java transformations.

A leaky abstraction over Jena is the content of the storage package. When a Store supports transactions, each connection is automatically run in a transaction conext.

Usage

You can either download the releases or use our artifactory instance located at the University of Mannheim. A minimal gradle build file is given below:

apply plugin: 'java'

repositories {
    jcenter()
    maven {
      url "http://eris-vm109.uni-mannheim.de:8081/artifactory/spa/"
    }
}

dependencies {
    compile(group: 'de.unima', name: 'spa-core', version: '0.0.11')
}

Currently SPA supports 3 different triple stores; namely Jena TDB, jena in-memory and Virtuoso. Each SPA instance is configured using the SPABuilder. The code below depicts how one can obtain instances.

// All instances have access to the same data
final SPA sharedMemorySpa = SPABuilder.local().sharedMemory().build();

// Connect to a remote virtuoso server
final SPA virtuosoRemoteSpa = SPABuilder.remote().virtuoso()
                               .url("http://virtuososerver:8890/sparql")
                               .username("yourUserName")
                               .password("yourPassword")
                               .build();

// In-Memory instance which does not share data with other instances
final SPA inMemorySpa = SPABuilder.local().uniqueMemory().build();

// Instance using Jena TDB
final SPA folderSpa = SPABuilder.local().folder(Files.createTempDirectory("TDB")).build();

To be continued