Skip to content

Apache DataFusion SQL Query Engine

Add smj SQL benchmark (#22803)

Bruce Ritchiea0e05f65a229
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q01.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q01.benchmark
new file mode 100644
index 0000000000..f1d44a6fb3
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q01.benchmark
@@ -0,0 +1,21 @@
+name Q01
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q1: INNER 1M x 1M | 1:1
+WITH t1_sorted AS (
+ SELECT value as key FROM range(1000000) ORDER BY value
+ ),
+ t2_sorted AS (
+SELECT value as key FROM range(1000000) ORDER BY value
+ )
+SELECT t1_sorted.key as k1, t2_sorted.key as k2
+FROM t1_sorted JOIN t2_sorted ON t1_sorted.key = t2_sorted.key
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q02.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q02.benchmark
new file mode 100644
index 0000000000..cd30f53256
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q02.benchmark
@@ -0,0 +1,25 @@
+name Q02
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q2: INNER 1M x 10M | 1:10
+WITH t1_sorted AS (
+ SELECT value % 100000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key, value as data
+FROM range(10000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, t1_sorted.data as d1, t2_sorted.data as d2
+FROM t1_sorted JOIN t2_sorted ON t1_sorted.key = t2_sorted.key
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q03.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q03.benchmark
new file mode 100644
index 0000000000..02bab2a685
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q03.benchmark
@@ -0,0 +1,25 @@
+name Q03
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q3: INNER 1M x 1M | 1:100
+WITH t1_sorted AS (
+ SELECT value % 10000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 10000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, t1_sorted.data as d1, t2_sorted.data as d2
+FROM t1_sorted JOIN t2_sorted ON t1_sorted.key = t2_sorted.key
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q04.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q04.benchmark
new file mode 100644
index 0000000000..2442906a3f
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q04.benchmark
@@ -0,0 +1,26 @@
+name Q04
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q4: INNER 1M x 10M | 1:10 | 1%
+WITH t1_sorted AS (
+ SELECT value % 100000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key, value as data
+FROM range(10000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, t1_sorted.data as d1, t2_sorted.data as d2
+FROM t1_sorted JOIN t2_sorted ON t1_sorted.key = t2_sorted.key
+WHERE t2_sorted.data % 100 = 0
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q05.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q05.benchmark
new file mode 100644
index 0000000000..1735d0e0e6
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q05.benchmark
@@ -0,0 +1,26 @@
+name Q05
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q5: INNER 1M x 1M | 1:100 | 10%
+WITH t1_sorted AS (
+ SELECT value % 10000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 10000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, t1_sorted.data as d1, t2_sorted.data as d2
+FROM t1_sorted JOIN t2_sorted ON t1_sorted.key = t2_sorted.key
+WHERE t1_sorted.data <> t2_sorted.data AND t2_sorted.data % 10 = 0
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q06.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q06.benchmark
new file mode 100644
index 0000000000..8c18ee164f
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q06.benchmark
@@ -0,0 +1,25 @@
+name Q06
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q6: LEFT 1M x 10M | 1:10
+WITH t1_sorted AS (
+ SELECT value % 105000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key, value as data
+FROM range(10000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, t1_sorted.data as d1, t2_sorted.data as d2
+FROM t1_sorted LEFT JOIN t2_sorted ON t1_sorted.key = t2_sorted.key
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q07.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q07.benchmark
new file mode 100644
index 0000000000..20619b0948
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q07.benchmark
@@ -0,0 +1,26 @@
+name Q07
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q7: LEFT 1M x 10M | 1:10 | 50%
+WITH t1_sorted AS (
+ SELECT value % 100000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key, value as data
+FROM range(10000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, t1_sorted.data as d1, t2_sorted.data as d2
+FROM t1_sorted LEFT JOIN t2_sorted ON t1_sorted.key = t2_sorted.key
+WHERE t2_sorted.data IS NULL OR t2_sorted.data % 2 = 0
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q08.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q08.benchmark
new file mode 100644
index 0000000000..7597f2012a
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q08.benchmark
@@ -0,0 +1,26 @@
+name Q08
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q8: FULL 1M x 1M | 1:10
+WITH t1_sorted AS (
+ SELECT value % 100000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 125000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key as k1, t1_sorted.data as d1,
+ t2_sorted.key as k2, t2_sorted.data as d2
+FROM t1_sorted FULL JOIN t2_sorted ON t1_sorted.key = t2_sorted.key
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q09.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q09.benchmark
new file mode 100644
index 0000000000..ca0565c6a6
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q09.benchmark
@@ -0,0 +1,29 @@
+name Q09
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q9: FULL 1M x 10M | 1:10 | 10%
+WITH t1_sorted AS (
+ SELECT value % 100000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key, value as data
+FROM range(10000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key as k1, t1_sorted.data as d1,
+ t2_sorted.key as k2, t2_sorted.data as d2
+FROM t1_sorted FULL JOIN t2_sorted ON t1_sorted.key = t2_sorted.key
+WHERE (t1_sorted.data IS NULL OR t2_sorted.data IS NULL
+ OR t1_sorted.data <> t2_sorted.data)
+ AND (t1_sorted.data IS NULL OR t1_sorted.data % 10 = 0)
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q10.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q10.benchmark
new file mode 100644
index 0000000000..cbba0610c5
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q10.benchmark
@@ -0,0 +1,29 @@
+name Q10
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q10: LEFT SEMI 1M x 10M | 1:10
+WITH t1_sorted AS (
+ SELECT value % 100000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key
+FROM range(10000000)
+ORDER BY key
+ )
+SELECT t1_sorted.key, t1_sorted.data
+FROM t1_sorted
+WHERE EXISTS (
+ SELECT 1 FROM t2_sorted
+ WHERE t2_sorted.key = t1_sorted.key
+)
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q11.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q11.benchmark
new file mode 100644
index 0000000000..2431b2646e
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q11.benchmark
@@ -0,0 +1,31 @@
+name Q11
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q11: LEFT SEMI 1M x 10M | 1:10 | 1%
+WITH t1_sorted AS (
+ SELECT value % 100000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key, value as data
+FROM range(10000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, t1_sorted.data
+FROM t1_sorted
+WHERE EXISTS (
+ SELECT 1 FROM t2_sorted
+ WHERE t2_sorted.key = t1_sorted.key
+ AND t2_sorted.data <> t1_sorted.data
+ AND t2_sorted.data % 100 = 0
+)
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q12.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q12.benchmark
new file mode 100644
index 0000000000..79e0a8e8a5
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q12.benchmark
@@ -0,0 +1,31 @@
+name Q12
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q12: LEFT SEMI 1M x 10M | 1:10 | 50%
+WITH t1_sorted AS (
+ SELECT value % 100000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key, value as data
+FROM range(10000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, t1_sorted.data
+FROM t1_sorted
+WHERE EXISTS (
+ SELECT 1 FROM t2_sorted
+ WHERE t2_sorted.key = t1_sorted.key
+ AND t2_sorted.data <> t1_sorted.data
+ AND t2_sorted.data % 2 = 0
+)
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q13.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q13.benchmark
new file mode 100644
index 0000000000..7e13e68743
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q13.benchmark
@@ -0,0 +1,31 @@
+name Q13
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q13: LEFT SEMI 1M x 10M | 1:10 | 90%
+WITH t1_sorted AS (
+ SELECT value % 100000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key, value as data
+FROM range(10000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, t1_sorted.data
+FROM t1_sorted
+WHERE EXISTS (
+ SELECT 1 FROM t2_sorted
+ WHERE t2_sorted.key = t1_sorted.key
+ AND t2_sorted.data <> t1_sorted.data
+ AND t2_sorted.data % 10 <> 0
+)
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q14.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q14.benchmark
new file mode 100644
index 0000000000..a56a0d5863
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q14.benchmark
@@ -0,0 +1,29 @@
+name Q14
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q14: LEFT ANTI 1M x 10M | 1:10
+WITH t1_sorted AS (
+ SELECT value % 105000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key
+FROM range(10000000)
+ORDER BY key
+ )
+SELECT t1_sorted.key, t1_sorted.data
+FROM t1_sorted
+WHERE NOT EXISTS (
+ SELECT 1 FROM t2_sorted
+ WHERE t2_sorted.key = t1_sorted.key
+)
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q15.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q15.benchmark
new file mode 100644
index 0000000000..bd64d74422
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q15.benchmark
@@ -0,0 +1,29 @@
+name Q15
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q15: LEFT ANTI 1M x 10M | 1:10 | partial match
+WITH t1_sorted AS (
+ SELECT value % 120000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key
+FROM range(10000000)
+ORDER BY key
+ )
+SELECT t1_sorted.key, t1_sorted.data
+FROM t1_sorted
+WHERE NOT EXISTS (
+ SELECT 1 FROM t2_sorted
+ WHERE t2_sorted.key = t1_sorted.key
+)
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q16.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q16.benchmark
new file mode 100644
index 0000000000..282d6374eb
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q16.benchmark
@@ -0,0 +1,29 @@
+name Q16
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q16: LEFT ANTI 1M x 1M | 1:1 | stress
+WITH t1_sorted AS (
+ SELECT value % 110000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key
+FROM range(1000000)
+ORDER BY key
+ )
+SELECT t1_sorted.key, t1_sorted.data
+FROM t1_sorted
+WHERE NOT EXISTS (
+ SELECT 1 FROM t2_sorted
+ WHERE t2_sorted.key = t1_sorted.key
+)
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q17.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q17.benchmark
new file mode 100644
index 0000000000..7f1c9a0ae2
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q17.benchmark
@@ -0,0 +1,26 @@
+name Q17
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q17: INNER 1M x 50M | 1:50 | 5%
+WITH t1_sorted AS (
+ SELECT value % 100000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key, value as data
+FROM range(50000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, t1_sorted.data as d1, t2_sorted.data as d2
+FROM t1_sorted JOIN t2_sorted ON t1_sorted.key = t2_sorted.key
+WHERE t2_sorted.data <> t1_sorted.data AND t2_sorted.data % 20 = 0
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q18.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q18.benchmark
new file mode 100644
index 0000000000..fac7edd19b
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q18.benchmark
@@ -0,0 +1,31 @@
+name Q18
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q18: LEFT SEMI 1M x 50M | 1:50 | 2%
+WITH t1_sorted AS (
+ SELECT value % 100000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key, value as data
+FROM range(50000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, t1_sorted.data
+FROM t1_sorted
+WHERE EXISTS (
+ SELECT 1 FROM t2_sorted
+ WHERE t2_sorted.key = t1_sorted.key
+ AND t2_sorted.data <> t1_sorted.data
+ AND t2_sorted.data % 50 = 0
+)
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q19.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q19.benchmark
new file mode 100644
index 0000000000..a867bb6bff
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q19.benchmark
@@ -0,0 +1,29 @@
+name Q19
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q19: LEFT ANTI 1M x 50M | 1:50 | partial match
+WITH t1_sorted AS (
+ SELECT value % 150000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key
+FROM range(50000000)
+ORDER BY key
+ )
+SELECT t1_sorted.key, t1_sorted.data
+FROM t1_sorted
+WHERE NOT EXISTS (
+ SELECT 1 FROM t2_sorted
+ WHERE t2_sorted.key = t1_sorted.key
+)
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q20.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q20.benchmark
new file mode 100644
index 0000000000..317c6290f7
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q20.benchmark
@@ -0,0 +1,26 @@
+name Q20
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q20: INNER 1M x 10M | 1:100 + GROUP BY
+WITH t1_sorted AS (
+ SELECT value % 10000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 10000 as key, value as data
+FROM range(10000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, count(*) as cnt
+FROM t1_sorted JOIN t2_sorted ON t1_sorted.key = t2_sorted.key
+GROUP BY t1_sorted.key
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q21.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q21.benchmark
new file mode 100644
index 0000000000..3fe460ea00
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q21.benchmark
@@ -0,0 +1,25 @@
+name Q21
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q21: INNER 10M x 10M | unique keys (1:1) | 50% join filter
+WITH t1_sorted AS (
+ SELECT value as key, value as data
+FROM range(10000000) ORDER BY value
+ ),
+ t2_sorted AS (
+SELECT value as key, value as data
+FROM range(10000000) ORDER BY value
+ )
+SELECT t1_sorted.key, t1_sorted.data as d1, t2_sorted.data as d2
+FROM t1_sorted JOIN t2_sorted
+ ON t1_sorted.key = t2_sorted.key
+ AND t1_sorted.data + t2_sorted.data < 10000000
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q22.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q22.benchmark
new file mode 100644
index 0000000000..fe0063de57
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q22.benchmark
@@ -0,0 +1,25 @@
+name Q22
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q22: LEFT 10M x 10M | unique keys (1:1) | 50% join filter
+WITH t1_sorted AS (
+ SELECT value as key, value as data
+FROM range(10000000) ORDER BY value
+ ),
+ t2_sorted AS (
+SELECT value as key, value as data
+FROM range(10000000) ORDER BY value
+ )
+SELECT t1_sorted.key, t1_sorted.data as d1, t2_sorted.data as d2
+FROM t1_sorted LEFT JOIN t2_sorted
+ ON t1_sorted.key = t2_sorted.key
+ AND t1_sorted.data + t2_sorted.data < 10000000
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q23.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q23.benchmark
new file mode 100644
index 0000000000..592effd993
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q23.benchmark
@@ -0,0 +1,26 @@
+name Q23
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q23: FULL 10M x 10M | unique keys (1:1) | 50% join filter
+WITH t1_sorted AS (
+ SELECT value as key, value as data
+FROM range(10000000) ORDER BY value
+ ),
+ t2_sorted AS (
+SELECT value as key, value as data
+FROM range(10000000) ORDER BY value
+ )
+SELECT t1_sorted.key as k1, t1_sorted.data as d1,
+ t2_sorted.key as k2, t2_sorted.data as d2
+FROM t1_sorted FULL JOIN t2_sorted
+ ON t1_sorted.key = t2_sorted.key
+ AND t1_sorted.data + t2_sorted.data < 10000000
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;
diff --git a/benchmarks/sql_benchmarks/smj/benchmarks/q24.benchmark b/benchmarks/sql_benchmarks/smj/benchmarks/q24.benchmark
new file mode 100644
index 0000000000..893eb1fb78
--- /dev/null
+++ b/benchmarks/sql_benchmarks/smj/benchmarks/q24.benchmark
@@ -0,0 +1,32 @@
+name Q24
+group smj
+
+init
+set datafusion.optimizer.prefer_hash_join=false;
+
+expect_plan SortMergeJoinExec
+
+run
+-- Q24: LEFT MARK 1M x 10M | 1:10 | 1%
+WITH t1_sorted AS (
+ SELECT value % 100000 as key, value as data
+FROM range(1000000)
+ORDER BY key, data
+ ),
+ t2_sorted AS (
+SELECT value % 100000 as key, value as data
+FROM range(10000000)
+ORDER BY key, data
+ )
+SELECT t1_sorted.key, t1_sorted.data
+FROM t1_sorted
+WHERE t1_sorted.data < 0
+ OR EXISTS (
+ SELECT 1 FROM t2_sorted
+ WHERE t2_sorted.key = t1_sorted.key
+ AND t2_sorted.data <> t1_sorted.data
+ AND t2_sorted.data % 100 = 0
+)
+
+cleanup
+reset datafusion.optimizer.prefer_hash_join;