Skip to main content

Transformations

note

Before you begin: You must already know about connecting data in Scenes apps before continuing with this guide.

Transformations are a powerful way to manipulate data returned by the SceneQueryRunner object before scenes render a visualization. Using transformations, you can:

  • Rename fields
  • Join time series data
  • Perform mathematical operations across queries
  • Use the output of one transformation as the input for another transformation

With transformations you can query data once, manipulate it, and display it in a scene.

Learn more about Grafana transformations in the official Grafana documentation.

Transform query results in a scene​

Step 1. Create a scene​

Create a scene with a single Table panel and a Prometheus query. The example query returns the average duration of HTTP requests for Prometheus API endpoints. The resulting table consists of three columns: Time, handler, and Value.

const queryRunner = new SceneQueryRunner({
  $timeRange: new SceneTimeRange(),
  datasource: {
    type: 'prometheus',
    uid: '<PROVIDE_GRAFANA_DS_UID>',
  },
  queries: [
    {
      refId: 'A',
      expr: 'sort_desc(avg by(handler) (rate(prometheus_http_request_duration_seconds_sum {}[5m]) * 1e3))',
      format: 'table',
      instant: true,
    },
  ],
});

const scene = new EmbeddedScene({
  $data: queryRunner,
  body: new SceneFlexLayout({
    direction: 'column',
    children: [
      new SceneFlexItem({
        body: PanelBuilders.table().setTitle('Average duration of HTTP request').build(),
      }),
    ],
  }),
});

Step 2. Configure a data transformation​

The resulting table from the previous step will look similar to the one that follows:

TimehandlerValue
2023-05-09 14:00:00.000/metrics1.10
2023-05-09 14:00:00.000/api/v1/label/:name/values0.361
2023-05-09 14:00:00.000/api/v1/metadata0.113
2023-05-09 14:00:00.000/api/v1/query_range0.0847
2023-05-09 14:00:00.000/api/v1/query0.14
2023-05-09 14:00:00.000/api/v1/labels0.0194
2023-05-09 14:00:00.000/api/v1/series0
2023-05-09 14:00:00.000/api/v1/status/buildinfo0

Add the Organize fields transformation to hide the Time field:

Create a SceneDataTransformer object:

const transformedData = new SceneDataTransformer({
  $data: queryRunner,
  transformations: [
    {
      id: 'organize',
      options: {
        excludeByName: {
          Time: true,
        },
        indexByName: {},
        renameByName: {},
      },
    },
  ],
});
note

Objects used in transformations are the same transformation configuration objects you would see in a typical dashboard panel when you view the JSON tab in the panel inspect drawer. To access this tab, click Inspect > Panel JSON in the panel edit menu.

Use the newly created transformedData object in place of the previously used SceneQueryRunner:

const scene = new EmbeddedScene({
  $data: transformedData,
  body: new SceneFlexLayout({
    direction: 'column',
    children: [
      new SceneFlexItem({
        body: PanelBuilders.table().setTitle('Average duration of HTTP request').build(),
      }),
    ],
  }),
});

The resulting table will look similar to the one that follows:

handlerValue
/metrics1.10
/api/v1/label/:name/values0.361
/api/v1/metadata0.113
/api/v1/query_range0.0847
/api/v1/query0.14
/api/v1/labels0.0194
/api/v1/series0
/api/v1/status/buildinfo0

Step 3. Configure multiple transformations​

SceneDataTransformer allows you to configure multiple transformations. The transformations are executed in the same order as they are added to the transformations array.

Modify the transformedData object and add Rename by regex transformations to change the field names: handler to Handler and Value to Average duration:

const transformedData = new SceneDataTransformer({
  $data: queryRunner,
  transformations: [
    {
      id: 'organize',
      options: {
        excludeByName: {
          Time: true,
        },
        indexByName: {},
        renameByName: {},
      },
    },
    {
      id: 'renameByRegex',
      options: {
        regex: 'handler',
        renamePattern: 'Handler',
      },
    },
    {
      id: 'renameByRegex',
      options: {
        regex: 'Value',
        renamePattern: 'Average duration',
      },
    },
  ],
});

The resulting table will look similar to the one that follows:

HandlerAverage duration
/metrics1.10
/api/v1/label/:name/values0.361
/api/v1/metadata0.113
/api/v1/query_range0.0847
/api/v1/query0.14
/api/v1/labels0.0194
/api/v1/series0
/api/v1/status/buildinfo0

Add custom transformations​

In addition to all the transformations available in Grafana, scenes allow you to create custom transformations.

SceneDataTransformer accepts CustomTransformOperator as an item of the transformations array:

  transformations: Array<DataTransformerConfig | CustomTransformOperator>;

CustomTransformOperator is a function that returns the RxJS Operator, which transforms data:

type CustomTransformOperator = (context: DataTransformContext) => MonoTypeOperatorFunction<DataFrame[]>;
note

Read more about RxJS operators in the RxJS official documentation.

Step 1. Create a custom transformation​

Create a custom transformation that will apply to the handler field and prefix all values with a URL:

note

Custom transformations depend heavily on manipulating internal Grafana data objects called data frames. Learn more about data frames in the official Grafana documentation.

import { DataFrame } from '@grafana/data';
import { Observable } from 'rxjs';
import { map } from 'rxjs/operators';

const prefixHandlerTransformation: CustomTransformOperator = () => (source: Observable<DataFrame[]>) => {
  return source.pipe(
    map((data: DataFrame[]) => {
      return data.map((frame: DataFrame) => {
        return {
          ...frame,
          fields: frame.fields.map((field) => {
            if (field.name === 'handler') {
              return {
                ...field,
                values: field.values.map((v) => 'http://www.my-api.com' + v),
              };
            }
            return field;
          }),
        };
      });
    })
  );
};

Step 2. Use a custom transformation​

Add a custom transformation to the previously created transformedData object:

const transformedData = new SceneDataTransformer({
  $data: queryRunner,
  transformations: [
    prefixHandlerTransformation,
    {
      id: 'organize',
      options: {
        excludeByName: {
          Time: true,
        },
        indexByName: {},
        renameByName: {},
      },
    },
    {
      id: 'renameByRegex',
      options: {
        regex: 'handler',
        renamePattern: 'Handler',
      },
    },
    {
      id: 'renameByRegex',
      options: {
        regex: 'Value',
        renamePattern: 'Average duration',
      },
    },
  ],
});
note

The prefixHandlerTransformation custom transformation is added as the first one because it applies to the handler field that's renamed to Handler in the following transformations. You can modify the custom transformation implementation so that it doesn't have to be used prior to other transformations.

The resulting table will look similar to the one that follows:

HandlerAverage duration
http://www.my-api.com/metrics1.10
http://www.my-api.com/api/v1/label/:name/values0.361
http://www.my-api.com/api/v1/metadata0.113
http://www.my-api.com/api/v1/query_range0.0847
http://www.my-api.com/api/v1/query0.14
http://www.my-api.com/api/v1/labels0.0194
http://www.my-api.com/api/v1/series0
http://www.my-api.com/api/v1/status/buildinfo0

Combine and filter​

One powerful thing you can do with transformations (custom and built-in) is share query results between panels in interesting ways. This allows you to place most of your queries in a single query runner that lives at the top of the scene. Then you can use a SceneDataTransformer object on the VizPanel level to join and filter the resulting data in different ways. Some panels may need the result of two of the queries, and another may need the results of all of them.

It's easy to filter the resulting DataFrame array by which query they came from using the refId property on DataFrame.

Source code​

View the example source code