How to Select Tables and Review Joins in Generated SQL

Load the tables SQL Mocker needs

Select Write SQL with AI, select your Database System, and then choose Extract Schema. Enter the tables you already know are relevant. SQL Mocker uses these as the starting point for metadata extraction and may include a one-step bridge or link table where supported foreign-key metadata identifies one.

Choose the initial Exact table names

Begin with the information the question must return. Enter the tables you know supply its business entities, measures, filters, or grouping fields. Use Schema Discovery to identify an unfamiliar supporting table, then add it to the extraction before investigating how the loaded tables connect.

You do not need to know the full join path or enter join syntax in advance. SQL Mocker uses relationships imported from the database metadata to determine how the reviewed tables connect. Avoid adding unrelated tables merely because their names match the general topic.

Review the relationship context SQL Mocker will use

What SQL Mocker determines and what you should review

Use Schema Discovery first, then Investigate Relationships

If you are unsure which tables supply the required fields or whether an intermediate table is needed, describe the intended result in Schema Discovery. It can suggest likely tables, columns, joins, and bridge tables to investigate. After confirming the database system and likely table names, generate the metadata SQL and load the returned structure into the workspace.

If those tables remain disconnected, open Relationships Overview and select Investigate Relationships. Run the generated investigation SQL in your database, return the results, and review the structural evidence for each candidate. You can also select one or more candidates for further data validation before confirming or rejecting them. Confirmed suggestions are stored as discovered workspace relationships, not as database-defined connections.

See Discover Missing Table Relationships for a walkthrough of the investigation and confirmation workflow.

Example: joining orders to customer regions

The query below follows two reviewed relationships: orders to customers, then customers to regions. Each join has a clear purpose in the requested output.

This example is illustrative. SQL Mocker generates the appropriate syntax for the database system selected in the workspace.

SELECT
    r.region_name,
    SUM(o.total_amount) AS total_order_value
FROM orders AS o
JOIN customers AS c
  ON c.customer_id = o.customer_id
JOIN regions AS r
  ON r.region_id = c.region_id
GROUP BY r.region_name
ORDER BY total_order_value DESC;

Review the generated SQL before use

Relationship metadata identifies how tables connect, but it does not prescribe a fixed join type. SQL Mocker chooses the join type from the user's question, so the final query should be reviewed before it is copied to a database tool.

• Confirm that the generated query uses the tables expected for the question.
• Check that any bridge or link table follows the path reviewed in Relationships Overview.
• Confirm that INNER, LEFT, or another join type preserves or excludes unmatched rows as intended.
• Refine the question when the reporting requirement or row-preservation intent needs to be more explicit.