graphql-engine/dc-agents/reference

Data Connector Agent Reference Implementation

This directory contains a barebones implementation of the Data Connector agent specification which fetches its data from static JSON files. It can be used as a reference implementation for testing, and as a reference for developers working on backend services.

Requirements

• NodeJS 16

Build & Run

> npm install
> npm start

Docker Build & Run

> docker build . -t dc-reference-agent:latest
> docker run -it --rm -p 8100:8100 dc-reference-agent:latest

Dataset

The dataset exposed by the reference agent is sourced from https://github.com/lerocha/chinook-database/

More specifically, the ChinookData.xml.gz file is a GZipped version of https://raw.githubusercontent.com/lerocha/chinook-database/ce27c48d9f375f81b7b68bacdfddf3c4458acc49/ChinookDatabase/DataSources/_Xml/ChinookData.xml

The schema-tables.json is manually derived from the schema of the data as can be seen from the CREATE TABLE etc DML statements in the various per-database-vendor SQL scripts that can be found in /ChinookDatabase/DataSources in that repo.

Configuration

The reference agent supports some configuration properties that can be set via the value property of configuration on a source in Hasura metadata. The configuration is passed to the agent on each request via the X-Hasura-DataConnector-Config header.

The configuration that the reference agent can take supports two properties:

• tables: This is a list of table names that should be exposed by the agent. If omitted all Chinook dataset tables are exposed. If specified, it filters all available table names by the specified list.
• schema: If specified, this places all the tables within a schema of the specified name. For example, if schema is set to my_schema, all table names will be namespaced under my_schema, for example ["my_schema","Album"]. If not specified, then tables are not namespaced, for example ["Album"].

Here's an example configuration that only exposes the Artist and Album tables, and namespaces them under my_schema:

{
  "tables": ["Artist", "Album"],
  "schema": "my_schema"
}

Here's an example configuration that exposes all tables, un-namespaced:

{}