Publish a table result

How-to guideApplies to: Unified engine deployments

Turn the inline orders summary into a durable Iceberg table. Verify the commit, query the output, and repeat the run without adding duplicate summary rows.

Before you start

You need a writable Iceberg catalog with namespace and table creation. Use an unused namespace; this lesson uses learn_workloads_01. Replace it everywhere if it already exists. In Workbench, select your writable catalog. Run the setup as one script:

CREATE NAMESPACE learn_workloads_01;
CREATE TABLE learn_workloads_01.category_sales (
  category STRING,
  sales BIGINT
) USING iceberg;

Expect an empty table. Record the catalog name. Do not assume that a catalog named default exists or supports writes.

1. Open the publication workflow

Replace the workload’s catalog with your writable catalog. Keep namespace learn_workloads_01 and apply it to every task. The publish task waits for summarize and reads its category_totals result.

{
  "name": "Learn - publish category sales",
  "tasks": [
    {
      "name": "summarize",
      "sql": "WITH orders(id, category, region, amount) AS (\n  VALUES\n    (1, 'Books', 'North', 20),\n    (2, 'Games', 'South', 30),\n    (3, 'Games', 'North', 15),\n    (4, 'Books', 'North', 40),\n    (5, 'Garden', 'South', 25),\n    (6, 'Garden', 'North', 10)\n)\nSELECT category, SUM(amount) AS sales\nFROM orders\nGROUP BY category;",
      "settings": {
        "current_catalog": "default",
        "namespace": [
          "learn_workloads_01"
        ],
        "time_zone": "UTC",
        "case_sensitive": false,
        "ansi_enabled": true
      }
    },
    {
      "name": "publish",
      "sql": "INSERT OVERWRITE TABLE learn_workloads_01.category_sales\nSELECT category, sales FROM category_totals;",
      "settings": {
        "current_catalog": "default",
        "namespace": [
          "learn_workloads_01"
        ],
        "time_zone": "UTC",
        "case_sensitive": false,
        "ansi_enabled": true
      }
    }
  ],
  "dependencies": [
    {
      "upstream": "summarize",
      "downstream": "publish",
      "input_name": "category_totals"
    }
  ],
  "options": {
    "execution_target": "DuckDb"
  },
  "policy": "Manual"
}

2. Run and inspect the commit

Validate, save with a Manual schedule, and start a run. Confirm both tasks succeed. Open the publication task and inspect its commit details. A successful summary query alone does not prove that the table write committed.

3. Verify the table

In Workbench, use the same catalog and run:

SELECT category, sales
FROM learn_workloads_01.category_sales
ORDER BY category;

category	sales
Books	60
Games	45
Garden	35

Start the workload again. Expect the same three rows. INSERT OVERWRITE replaces the table contents; INSERT INTO would append another set of rows.

Choose repeat behavior deliberately

Overwrite suits this complete summary because each run produces the full desired table. A windowed publication needs a write strategy that preserves other windows, such as a supported keyed MERGE. Replacing the whole table with one day’s output would lose other days. A workload is not one transaction across all tasks. If an earlier task commits and a later task fails, the earlier write can remain visible. Selective reruns do not undo it or freeze the latest contents of an external table.

Clean up

Keep the schedule Manual, or pause it if you enabled recurring execution. Wait for active runs to finish and remove the lesson workload when no longer needed. Drop only the objects created for this lesson:

DROP TABLE learn_workloads_01.category_sales;
DROP NAMESPACE learn_workloads_01;

Use the same catalog. See Results and table commits before repeating an uncertain write.

Transformation Engine

Learn Laminar

Workloads

SQL reference

HTTP API

Recipes

Operations

Publish a table result

Before you start

1. Open the publication workflow

2. Run and inspect the commit

3. Verify the table

Choose repeat behavior deliberately

Clean up

​Before you start

​1. Open the publication workflow

​2. Run and inspect the commit

​3. Verify the table

​Choose repeat behavior deliberately

​Clean up

Before you start

1. Open the publication workflow

2. Run and inspect the commit

3. Verify the table

Choose repeat behavior deliberately

Clean up