Incremental load in Query Service

The incremental load design pattern is a solution for managing data. The pattern only processes information in the dataset that has been created or modified since the last load execution.

Incremental load uses various features provided by Adobe Experience Platform Query Service such as anonymous block and snapshots. This design pattern increases processing efficiency as any data already processed from the source is skipped. It can be used with both streaming and batch data processing.

This document provides a series of instructions to build a design pattern for incremental processing. These steps can be used as a template to create your own incremental data load queries.

Getting started

The SQL examples throughout this document require you to have an understanding of the anonymous block and snapshot capabilities. It is recommended that you read the sample anonymous block queries documentation and also the snapshot clause documentation.

For guidance on any terminology used within this guide, refer to the SQL syntax guide.

Incrementally load data

The steps below demonstrate how to create and incrementally load data using snapshots and the anonymous block feature. The design pattern can be used as a template for your own sequence of queries.

  1. Create a checkpoint_log table to track the most recent snapshot used to process data successfully. The tracking table (checkpoint_log in this example) must first be initialized to null in order to incrementally process a dataset.

    code language-sql
    DROP TABLE IF EXISTS checkpoint_log;
    CREATE TABLE  checkpoint_log AS
    SELECT
       cast(NULL AS string) process_name,
       cast(NULL AS string) process_status,
       cast(NULL AS string) last_snapshot_id,
       cast(NULL AS TIMESTAMP) process_timestamp
       WHERE false;
    
  2. Populate the checkpoint_log table with one empty record for the dataset that needs incremental processing. DIM_TABLE_ABC is the dataset to be processed in the example below. On the first occasion of processing DIM_TABLE_ABC, the last_snapshot_id is initialized as null. This allows you to process the entire dataset on the first occasion and incrementally thereafter.

    code language-sql
    INSERT INTO
       checkpoint_log
       SELECT
          'DIM_TABLE_ABC' process_name,
          'SUCCESSFUL' process_status,
          cast(NULL AS string) last_snapshot_id,
          CURRENT_TIMESTAMP process_timestamp;
    
  3. Next, initialize DIM_TABLE_ABC_Incremental to contain processed output from DIM_TABLE_ABC. The anonymous block in the required execution section of the SQL example below, as described in steps one to four, is executed sequentially to process data incrementally.

    1. Set the from_snapshot_id which indicates where the processing starts from. The from_snapshot_id in the example is queried from the checkpoint_log table for use with DIM_TABLE_ABC. On the initial run, the snapshot ID will be null meaning that the entire dataset will be processed.
    2. Set the to_snapshot_id as the current snapshot ID of the source table (DIM_TABLE_ABC). In the example, this is queried from the metadata table of the source table.
    3. Use the CREATE keyword to create DIM_TABLE_ABC_Incremenal as the destination table. The destination table persists the processed data from the source dataset (DIM_TABLE_ABC). This allows the processed data from the source table between from_snapshot_id and to_snapshot_id, to be incrementally appended to the destination table.
    4. Update the checkpoint_log table with the to_snapshot_id for the source data that DIM_TABLE_ABC processed successfully.
    5. If any of the sequentially executed queries of the anonymous block fail, the optional exception section is executed. This returns an error and ends the process.
    note note
    NOTE
    The history_meta('source table name') is a convenient method used to gain access to available snapshot in a dataset.
    code language-sql
    $$ BEGIN
        SET @from_snapshot_id = SELECT coalesce(last_snapshot_id, 'HEAD') FROM checkpoint_log a JOIN
                                (SELECT MAX(process_timestamp)process_timestamp FROM checkpoint_log
                                    WHERE process_name = 'DIM_TABLE_ABC' AND process_status = 'SUCCESSFUL' )b
                                    ON a.process_timestamp=b.process_timestamp;
        SET @to_snapshot_id = SELECT snapshot_id FROM (SELECT history_meta('DIM_TABLE_ABC')) WHERE  is_current = true;
        SET @last_updated_timestamp= SELECT CURRENT_TIMESTAMP;
        CREATE TABLE DIM_TABLE_ABC_Incremental AS
         SELECT  *  FROM DIM_TABLE_ABC SNAPSHOT BETWEEN @from_snapshot_id AND @to_snapshot_id ;
    
    INSERT INTO
       checkpoint_log
       SELECT
           'DIM_TABLE_ABC' process_name,
           'SUCCESSFUL' process_status,
          cast( @to_snapshot_id AS string) last_snapshot_id,
          cast( @last_updated_timestamp AS TIMESTAMP) process_timestamp;
    
    EXCEPTION
      WHEN OTHER THEN
        SELECT 'ERROR';
    END
    $$;
    
  4. Use the incremental data load logic in the anonymous block example below to allow any new data from the source dataset (since the most recent timestamp), to be processed and appended to the destination table at a regular cadence. In the example, data changes to DIM_TABLE_ABC will be processed and appended to DIM_TABLE_ABC_incremental.

    note note
    NOTE
    _ID is the Primary Key in both DIM_TABLE_ABC_Incremental and SELECT history_meta('DIM_TABLE_ABC').
    code language-sql
    $$ BEGIN
        SET @from_snapshot_id = SELECT coalesce(last_snapshot_id, 'HEAD') FROM checkpoint_log a join
                                (SELECT MAX(process_timestamp)process_timestamp FROM checkpoint_log
                                    WHERE process_name = 'DIM_TABLE_ABC' AND process_status = 'SUCCESSFUL' )b
                                    ON a.process_timestamp=b.process_timestamp;
        SET @to_snapshot_id = SELECT snapshot_id FROM (SELECT history_meta('DIM_TABLE_ABC')) WHERE  is_current = true;
        SET @last_updated_timestamp= SELECT CURRENT_TIMESTAMP;
        INSERT INTO DIM_TABLE_ABC_Incremental
         SELECT  *  FROM DIM_TABLE_ABC SNAPSHOT BETWEEN @from_snapshot_id AND @to_snapshot_id WHERE NOT EXISTS (SELECT _id FROM DIM_TABLE_ABC_Incremental a WHERE _id=a._id);
    
    INSERT INTO
       checkpoint_log
       SELECT
           'DIM_TABLE_ABC' process_name,
           'SUCCESSFUL' process_status,
          cast( @to_snapshot_id AS string) last_snapshot_id,
          cast( @last_updated_timestamp AS TIMESTAMP) process_timestamp;
    
    EXCEPTION
      WHEN OTHER THEN
        SELECT 'ERROR';
    END
    $$;
    

This logic can be applied to any table to perform incremental loads.

Expired Snapshots

IMPORTANT
Snapshot metadata expires after two days. An expired snapshot invalidates the logic of the script provided above.

To resolve the issue of an expired snapshot ID, insert the following command at the beginning of the anonymous block. The following line of code overrides the @from_snapshot_id with the earliest available snapshot_id from metadata.

SET resolve_fallback_snapshot_on_failure=true;

The entire code block looks as follows:

$$ BEGIN
    SET resolve_fallback_snapshot_on_failure=true;
    SET @from_snapshot_id = SELECT coalesce(last_snapshot_id, 'HEAD') FROM checkpoint_log a JOIN
                            (SELECT MAX(process_timestamp)process_timestamp FROM checkpoint_log
                                WHERE process_name = 'DIM_TABLE_ABC' AND process_status = 'SUCCESSFUL' )b
                                on a.process_timestamp=b.process_timestamp;
    SET @to_snapshot_id = SELECT snapshot_id FROM (SELECT history_meta('DIM_TABLE_ABC')) WHERE  is_current = true;
    SET @last_updated_timestamp= SELECT CURRENT_TIMESTAMP;
    INSERT INTO DIM_TABLE_ABC_Incremental
     SELECT  *  FROM DIM_TABLE_ABC SNAPSHOT BETWEEN @from_snapshot_id AND @to_snapshot_id WHERE NOT EXISTS (SELECT _id FROM DIM_TABLE_ABC_Incremental a WHERE _id=a._id);

Insert Into
   checkpoint_log
   SELECT
       'DIM_TABLE_ABC' process_name,
       'SUCCESSFUL' process_status,
      cast( @to_snapshot_id AS string) last_snapshot_id,
      cast( @last_updated_timestamp AS TIMESTAMP) process_timestamp;
EXCEPTION
  WHEN OTHER THEN
    SELECT 'ERROR';
END
$$;

Next steps

By reading this document, you should have a better understanding of how to use anonymous block and snapshot features to perform incremental loads and can apply this logic to your own specific queries. For general guidance on query execution, please read the guide on query execution in Query Service.

recommendation-more-help
ccf2b369-4031-483f-af63-a93b5ae5e3fb