Find the rows of a correlated subquery through an index for an expression of the outer columns

[tamnd]

Summary

A correlated subquery finds its rows through an index of a column when its condition is an = of the column and a value of the outer query. Before, the value had to be a column of the outer query. Now it can be an expression of the columns of the outer query, as t2.thousand = t1.tenthous + 1. In the regression test create_index, one such query read 100 million rows and took 281 seconds.

Changes

• outer_expr replaces outer_var in rupg-exec. It accepts a column of an outer query, a constant, a parameter, a cast, and a function that is not volatile and does not return a set, of such expressions. The expression must read at least one column of an outer query, so a plain query still scans its rows once.
• volatile in rupg-exec is now pub(crate).

Known differences

None. The result of each query is the same, and the scan only reads fewer rows.

Tests

connection/outer_lookups.test is the output of PostgreSQL 19 for the same script in a new database. fmt, clippy, the workspace tests and the checks of xtask pass on server1.