Evidence Datasource Connector Template

Usage

Configuring package.json

1. Update the package name
2. Update evidence.datasources
   1. This contains an array of arbitrary data source names that users will use to select this source in their connection.yaml files
   2. If your connector supports multiple sources, or you have several aliases (e.g. psql, postgres, postgresql), you can provide a nested array, this will show only the first item in the UI
      • In this example, postgres, psql, postgresql, redshift, and timescaledb will all select this connector
        However, only postgres, redshift, and timescaledb will be presented as options in the UI.

        {
            "evidence": {
                "datasources": [
                    [ "postgres", "psql", "postgresql" ],  // Shows only `postgres` in the UI
                    "redshift",
                    "timescaledb"
                ]
            }
        }

3. Optionally, specify an icon for your datasource.
   1. Icons can come from Simple Icons, or Tabler Icons.
   2. Evidence uses Steeze UI for our icons, so the casing must match
      the Steeze UI export

Specify your connector's options

index.js defines the type ConnectorOptions, and exports an options constant.

ConnectorOptions should be typed to the expected configuration for your datasource (e.g. hostname, port, etc)

options defines how your connector will be configured in the UI, we recommend reading the docs, and/or taking a look at the Evidence Postgres Connector. Technically, implementing this is optional, but it provides a much better user experience when your datasource is installed.

Choosing an interface

Evidence accepts 2 different interfaces when using datasources, one is much easier to write, but is much less flexible.

Note that lib.js has a stubbed databaseTypeToEvidenceType, which is helpful for building ColumnTypes more easily.

Simple Interface

For the simple interface, implement the getRunner function; which is a factory pattern for building a configured QueryRunner. Each query can either return an array of results, or an async generator function if implementing cursor logic (this enables much larger datasets)

Advanced Interface

For the advanced interface, implement the processSource function; which is an async generator function returning tables directly.

processSource receives a proxy of the source's filetree, so it must look for files itself, but returns results in the same available formats as getRunner. processSource should be used in instances where output tables do not map one to one with input files (e.g. if a list of tables is provided in connection.yaml that should all be SELECT *'d)

[Recommended] Write Unit Tests

This template comes with vitest pre-installed. If you've used jest, vitest implements a very similar API.

Tests have been stubbed in index.spec.js, and can be run with npm run test

Typescript has also been included with a basic configuration, and your types can be checked with npm run check

Testing your datasource

1. Install your connector in test-project using npm i ../datasource
2. Add your connector's package name to evidence.plugins.yaml
3. Create a source that uses your connector
   1. You can either create it manually in sources, or in the settings ui.
   2. If you populated your options object, testing it in the settings UI is recommended.
4. Use npm run sources in the test project to execute your datasource, and modify index.md to use your new source.
   1. It can also be helpful to referenecs the schema explorer to make sure that the types returned
      by your connector match what you are expecting.