Skip to main content

Schema auto-discovery on ingestion jobs

When you ingest data into Imply Polaris, Polaris can automatically detect the schema of the input fields of your source data. Unless you specify otherwise, Polaris ingests the input fields into dimensions with the same names and data types. This is known as schema auto-discovery for ingestion jobs. A long-running streaming ingestion job automatically detects new input fields and ingests them into the table as events get consumed or published from the event stream.


You can use schema auto-discovery when the following conditions hold:

  • The source of ingested data comes from an event stream.
  • The target of the ingestion job has a flexible schema mode. If you do not create a table before ingestion, Polaris automatically creates a new table in flexible mode. For more details, see Create a table.

The target table may either be detail or aggregate.

Required mappings

While not required for schema auto-discovery for jobs, you still need to define the input fields and mappings in the following situations:

  • When you specify the input and mapping for the primary timestamp, __time.
  • When you change the name or data type of the input field.
  • When you apply a transformation to the input field.
  • When you apply an aggregation function to an input field. The output of the aggregation defined in the ingestion job maps to a measure in an aggregate table.

A mapping can only refer to fields declared in the ingestion job's input schema. It cannot refer to other mappings or to auto-discovered fields that are not declared.

Deny list

There may be columns in your input data that you do not plan to query. You can exclude input fields from ingestion by adding them to the deny list. You may also want to add the date column that gets mapped to __time to avoid ingesting that input field twice.

Access the deny list on the Map source to table stage of ingestion:

Ingestion deny list

You cannot update a job to update its deny list. To make changes to the deny list, cancel the ingestion job and submit a new one with the updated list.


This example shows how schema auto-discovery works for streaming ingestion jobs. You should already be familiar with how to create tables and ingestion jobs.

This example assumes you have a working streaming source, such as from Confluent Cloud or Amazon Kinesis. For details on creating streaming connections, see Create a connection.

  1. Create a table in Flexible schema mode. Fill in the table name and select a table type. In this example, the table name is Example table, and the table type is set to Aggregate.

  2. In the table view, click Load data > Insert data.

  3. In the Select source step, select your streaming source. Click Next.

  4. In the Parse data stage, select the input format of your data. Click Continue.

  5. In the Map source to table stage, Polaris displays the data types of auto-discovered dimensions as Auto. Auto-discovered columns

    To add a new column, select Add column from the menu bar. Select Dimension or Measure, then define the column's name and input expression. Add new column

    If the input expression references an input field that hasn't been defined, Polaris displays a dialog for you to add the input field. Declare input

    Auto-discovered dimensions are undeclared and may change data type as more data is ingested. To enforce a strict schema for a certain column, select the toggle to Declare column in table schema. Declared columns display the declared data type in the column header. For information about declaring columns in a flexible table, see Flexible table. Declare column

  6. Click Start ingestion to begin the ingestion job.

Learn more

See the following topics for more information: