Message-ID: <gh-postgresql-interfaces-psqlodbc-171@github.com>
From: "davecramer (@davecramer)" <noreply+davecramer@github.com>
To: "postgresql-interfaces/psqlodbc" <noreply+postgresql-interfaces-psqlodbc@github.com>
Date: Sun, 12 Apr 2026 11:40:57 +0000
Subject: [postgresql-interfaces/psqlodbc] PR #171:  Fix SQLPrimaryKeys to exclude INCLUDE columns from results
List-Id: <gh-postgresql-interfaces-psqlodbc.github.com>
X-GitHub-Author-Id: 406518
X-GitHub-Author-Login: davecramer
X-GitHub-Issue: 171
X-GitHub-Repo: postgresql-interfaces/psqlodbc
X-GitHub-State: merged
X-GitHub-Type: pull_request
X-GitHub-Url: https://github.com/postgresql-interfaces/psqlodbc/pull/171
Content-Type: text/plain; charset=utf-8

  When a PRIMARY KEY is created with an INCLUDE clause (PostgreSQL 11+),
  SQLPrimaryKeys incorrectly returned both the key columns and the
  included columns. For example, given PRIMARY KEY (a, b) INCLUDE (c, d),
  all four columns were returned instead of just a and b.

  The root cause was that the pg_attribute join on the index relation
  iterated over all index attributes without distinguishing key columns
  from included columns. Fix by adding a filter on i.indnkeyatts when
  connected to PostgreSQL >= 11, which limits results to only the actual
  key attributes. Both query paths in PGAPI_PrimaryKeys are fixed.

  Add a regression test that creates a table with a composite primary key
  using INCLUDE and verifies only the key columns are returned.

fixes #170