Bundled Examples

SQLSage ships a demonstration catalog and twelve realistic queries. They cover the correctness and performance problems the tool is designed to catch, and none of them touch your data:

sqlsage list
sqlsage analyze --corpus q05 --format text

ID	Title	The problem it shows
q01-nonsargable-date	Monthly revenue via date_trunc on the filtered column	An index the planner cannot use.
q02-leading-wildcard-or	Customer search with leading wildcard and OR across columns	Non-sargable LIKE and OR.
q03-correlated-scalar-subquery	Per-customer order counts via correlated scalar subqueries	Correlated work.
q04-deep-offset-pagination	Deep OFFSET pagination over the orders feed	The cost of skipping rows.
q05-not-in-nullable	Customers with no events, via NOT IN on a nullable column	A query that returns the wrong answer.
q06-fanout-double-count	Revenue per customer joined through order_items	Join fan-out double counting.
q07-left-join-demoted	Customers and their complete orders, filtered in WHERE	Demotion of an outer join.
q08-distinct-hides-fanout	DISTINCT used to undo a join fan-out	Fan-out hidden behind DISTINCT.
q09-cast-on-column	Daily order counts with a cast applied to the indexed column	A cast that defeats the index.
q10-having-instead-of-where	Aggregate with a row filter placed in HAVING	Row filter misplaced in HAVING.
q11-top-n-per-group	Most recent order per customer via a correlated max()	Per-group top-N.
q12-jsonb-and-unbounded-sort	Event funnel filtered on a jsonb attribute	jsonb filtering and a full sort.

Start with the wrong answer

q05 is the demo query — a nullable NOT IN that can silently return zero rows instead of 196,000:

sqlsage analyze --corpus q05 --format text

WRONG RESULTS
This query returns wrong answers today. Fix that before you spend a minute on
its speed.

It is the best first example because it shows the product's strongest claim: SQLSage tells you a query is wrong before it tells you it is slow.

Run them all

for q in $(sqlsage list | awk '{print $1}'); do
  echo "== $q =="
  sqlsage analyze --corpus "$q" --format markdown
done

Each example is also a regression case: the analysis runs in CI on every commit, and the --check IR assertions pin the project's real regression history.