A JSON Descriptor is a JSON file containing information describing a specific entity, e.g. a node, a graphlet, or (part of) an application.

JSON Schemas

For each entity type a JSON schema <type>.schema.json describes the data format. The schema provides a human readable documentation about each key as well enables validating that JSON files adhere to the schema.

app.schema.json : describing an application
app-config.schema.json : describing configuration values for an application
graphlet.schema.json : describing a graphlet
node.schema.json : describing a node
node-params.schema.json : describing parameter sets of a node : sibling file to .node.json with the same stem name
node-wcets.schema.json : describing WCET sets of a node : sibling file to .node.json with the same stem name
required-sensors.schema.json : describing required sensors : referenced by applications
sensor-mappings.schema.json : describing sensor mappings for various layouts : lookup-ed by applications, optional

Validation

JSON schemas can only express a subset of constraints for the data in a JSON file and are also limited to validate a single file. Therefore a validation script perform additional constraint checking as well as checking cross-file constraints.

‍Note: in most cases the order of keys in JSON is strictly enforced to ensure a deterministic layout.

Visualization and Graphical Editing

Graphlets and applications are exclusively described in JSON and can be visualized and edited using a graphical UI.

Node Descriptor

C++ API to .node.json

See development time workflow.

The information in a node descriptor can be extracted from the C++ API of an implemented node using the command line tool nodedescriptor. Since the JSON information is derived in this case, the graphical tool doesn't offer editing the node information.

.node.json to C++ Stubs

See design time workflow.

If no implementation of a node exists yet, it can be designed from scratch in the graphical UI. This enables designing nodes, graphlets, and entire applications to e.g. analyze the scheduling without having to implement any code. The command line tool nodestub can generate C++ stub headers and sources from a node descriptor file to speed up the start of the implementation.