Data Flows
Use a data flow to prepare data from Gnok tables. Each step defines an operation such as cleaning text, joining sources, or calculating a measure. The canvas shows how steps feed one another and which step produces the output. Data Flows appears in Studio when your workspace has it enabled.
Follow Build a regional revenue flow to compare two years, inspect a join, and publish a dataset that a dashboard can use.
Build a flow on the canvas
- Open Data Flows and choose New flow. Give it a name and select the connection that can access your source tables.
- Add a table source and select its catalog, schema, and table. Inspect the loaded fields and choose Preview step to check representative rows.
- Add an operation from the palette and connect it to its input. Drag from the palette or use its buttons. Use a second input for joins and appends. Select a card to configure it in the inspector.
- Select the step that should provide the output. Use Validate output to resolve the output schema and run configured checks. Choose Save draft to save explicitly while you work.
| Task | Operations to use |
|---|---|
| Clean and reshape | Select or rename fields, trim or change text case, deduplicate, pivot, or unpivot. |
| Filter and summarize | Apply typed filters, group rows, and add aggregated measures. |
| Calculate values | Combine fields, constants, conversions, conditional expressions, and functions offered by the connection. |
| Combine inputs | Join on one or more key pairs, look up matching attributes, or append rows using explicit field mappings. |
| Compare or rank rows | Use ordered calculations for rank, previous values, or running and rolling totals. Specify an order that resolves ties when position matters. |
| Test assumptions | Add Check rows for required values, unique keys, allowed values, ranges, or reconciled totals. |
Inspect the results before publishing
- Preview step shows up to 200 result rows. A preview of a join or aggregation can read more input rows than it displays.
- Profile columns summarizes sampled values, missing or distinct values, and numeric bounds. Use Compare row counts to compare the selected step with its immediate inputs.
- Inspect join counts and exceptions when a join introduces missing matches or unexpectedly multiplies rows. Check the key on both sides before trusting downstream totals.
- Select the output step and run Validate output after schema changes. Use the result and generated SQL to confirm that the expected field names and types reach the output.
Example of comparing periods by region
Use a daily sales table with fields for date, region, revenue, profit, and orders. Add one source branch for the current fiscal year and a second for the previous fiscal year. Filter each branch to its year, then group each by region and sum all three measures. Join the two summaries on region and select both years’ measures in the output.
Decide how to handle a region that exists in only one period. Use Inspect exceptions to find it and use a join type that expresses the intended result. Compare the final revenue, profit, and order counts with independent totals for each year. This keeps the report's assumptions visible when someone else opens the flow.
Continue with Publish, Share, and Use Flow Outputs to create a version and reuse it in a dashboard or destination table.