check_references

List the child rows whose foreign key has no matching parent. Read-only analytics (stays available under --mcp-read-only).

When to use

• Before a join, to find out why it will drop rows.
• After an import, to check the two files still line up.
• As an agent's sanity check on a dataset it did not produce.

Input schema

Parameter	Type	Required?	Default	Description
path	string	yes*	(no default)	The parent file
open_tab	string	no	(none)	Open GUI tab holding the parent, or @active
table	string	no	first table	Table within a multi-table parent source
parent_column	string	yes	(no default)	The parent's key column
child_path	string	no	parent	The child file
child_open_tab	string	no	parent	Open tab holding the child
child_table	string	no	first table	Table within a multi-table child source
child_column	string	yes	(no default)	The child's foreign-key column
unlimited	bool	no	false	Lift the streaming row cap

* One of path or open_tab is required. Omitting the child source checks a self-reference, such as a manager_id pointing at id in one table.

Response

{
  "clean": false,
  "sentence": "2 child row(s) across 1 value(s) have no parent.",
  "parent_values": 3,
  "checked_rows": 4,
  "null_keys": 1,
  "orphan_rows": 2,
  "orphan_values": 1,
  "orphans": [{ "value": "9", "rows": 2 }]
}

orphans is most-rows-first and capped at 500 listed values; orphan_rows and orphan_values stay exact however long the list would have been.

Two conventions worth knowing

• A null or empty key is not an orphan. In every relational database a null foreign key means "no parent", not "a parent that vanished". Those rows are counted as null_keys and left out of checked_rows.
• Keys are compared as trimmed text, the same as suggest_join_keys and the relationship map. That is what makes 1 match 1 when one side came from a CSV and the other from a database.