optimize keyless read row (#2094)

2026-03-06 10:29:33 -06:00 · 2021-09-23 10:57:31 -07:00
parent b3c1b82b99
commit 5cc0852692
11 changed files with 141 additions and 80 deletions
--- a/go/libraries/doltcore/sqle/enginetest/dolt_engine_test.go
+++ b/go/libraries/doltcore/sqle/enginetest/dolt_engine_test.go
@@ -24,7 +24,8 @@ import (
 )

 func init() {
-	sqle.MinRowsPerPartition = 2
+	sqle.MinRowsPerPartition = 8
+	sqle.MaxRowsPerPartition = 1024
 }

 func TestQueries(t *testing.T) {
--- a/go/libraries/doltcore/sqle/rows.go
+++ b/go/libraries/doltcore/sqle/rows.go
@@ -26,6 +26,38 @@ import (
 	"github.com/dolthub/dolt/go/store/types"
 )

+var _ sql.RowIter = (*keylessRowIter)(nil)
+
+type keylessRowIter struct {
+	keyedIter *DoltMapIter
+
+	cardIdx     int
+	nonCardCols int
+
+	lastRead sql.Row
+	lastCard uint64
+}
+
+func (k *keylessRowIter) Next() (sql.Row, error) {
+	if k.lastCard == 0 {
+		r, err := k.keyedIter.Next()
+
+		if err != nil {
+			return nil, err
+		}
+
+		k.lastCard = r[k.cardIdx].(uint64)
+		k.lastRead = r[:k.nonCardCols]
+	}
+
+	k.lastCard--
+	return k.lastRead, nil
+}
+
+func (k keylessRowIter) Close(ctx *sql.Context) error {
+	return k.keyedIter.Close(ctx)
+}
+
 // An iterator over the rows of a table.
 type doltTableRowIter struct {
 	sql.RowIter
@@ -43,48 +75,71 @@ func newRowIterator(ctx *sql.Context, tbl *doltdb.Table, projCols []string, part

 	if schema.IsKeyless(sch) {
 		// would be more optimal to project columns into keyless tables also
-		return newKeylessRowIterator(ctx, tbl, partition)
+		return newKeylessRowIterator(ctx, tbl, projCols, partition)
 	} else {
 		return newKeyedRowIter(ctx, tbl, projCols, partition)
 	}
 }

-func newKeylessRowIterator(ctx *sql.Context, tbl *doltdb.Table, partition *doltTablePartition) (*doltTableRowIter, error) {
-	var iter table.SqlTableReader
-	var err error
-	if partition.end == NoUpperBound {
-		iter, err = table.NewBufferedTableReader(ctx, tbl)
-	} else {
-		iter, err = table.NewBufferedTableReaderForPartition(ctx, tbl, partition.start, partition.end)
-	}
-
+func newKeylessRowIterator(ctx *sql.Context, tbl *doltdb.Table, projectedCols []string, partition *doltTablePartition) (sql.RowIter, error) {
+	mapIter, err := iterForPartition(ctx, partition)
 	if err != nil {
 		return nil, err
 	}

-	return &doltTableRowIter{
-		ctx:    ctx,
-		reader: iter,
+	cols, tagToSqlColIdx, err := getTagToResColIdx(ctx, tbl, projectedCols)
+	if err != nil {
+		return nil, err
+	}
+
+	idxOfCardinality := len(cols)
+	tagToSqlColIdx[schema.KeylessRowCardinalityTag] = idxOfCardinality
+
+	colsCopy := make([]schema.Column, len(cols), len(cols)+1)
+	copy(colsCopy, cols)
+	colsCopy = append(colsCopy, schema.NewColumn("__cardinality__", schema.KeylessRowCardinalityTag, types.UintKind, false))
+
+	conv := NewKVToSqlRowConverter(tbl.Format(), tagToSqlColIdx, colsCopy, len(colsCopy))
+	keyedItr, err := NewDoltMapIter(ctx, mapIter.NextTuple, nil, conv), nil
+	if err != nil {
+		return nil, err
+	}
+
+	return &keylessRowIter{
+		keyedIter:   keyedItr,
+		cardIdx:     idxOfCardinality,
+		nonCardCols: len(cols),
 	}, nil
 }

 func newKeyedRowIter(ctx context.Context, tbl *doltdb.Table, projectedCols []string, partition *doltTablePartition) (sql.RowIter, error) {
-	var err error
-	var mapIter types.MapTupleIterator
+	mapIter, err := iterForPartition(ctx, partition)
+	if err != nil {
+		return nil, err
+	}
+
+	cols, tagToSqlColIdx, err := getTagToResColIdx(ctx, tbl, projectedCols)
+	if err != nil {
+		return nil, err
+	}
+
+	conv := NewKVToSqlRowConverter(tbl.Format(), tagToSqlColIdx, cols, len(cols))
+	return NewDoltMapIter(ctx, mapIter.NextTuple, nil, conv), nil
+}
+
+func iterForPartition(ctx context.Context, partition *doltTablePartition) (types.MapTupleIterator, error) {
 	rowData := partition.rowData
 	if partition.end == NoUpperBound {
-		mapIter, err = rowData.RangeIterator(ctx, 0, rowData.Len())
+		return rowData.RangeIterator(ctx, 0, rowData.Len())
 	} else {
-		mapIter, err = partition.IteratorForPartition(ctx, rowData)
-	}
-
-	if err != nil {
-		return nil, err
+		return partition.IteratorForPartition(ctx, rowData)
 	}
+}

+func getTagToResColIdx(ctx context.Context, tbl *doltdb.Table, projectedCols []string) ([]schema.Column, map[uint64]int, error) {
 	sch, err := tbl.GetSchema(ctx)
 	if err != nil {
-		return nil, err
+		return nil, nil, err
 	}

 	cols := sch.GetAllCols().GetColumns()
@@ -96,9 +151,7 @@ func newKeyedRowIter(ctx context.Context, tbl *doltdb.Table, projectedCols []str
 			tagToSqlColIdx[col.Tag] = i
 		}
 	}
-
-	conv := NewKVToSqlRowConverter(tbl.Format(), tagToSqlColIdx, cols, len(cols))
-	return NewDoltMapIter(ctx, mapIter.NextTuple, nil, conv), nil
+	return cols, tagToSqlColIdx, nil
 }

 // Next returns the next row in this row iterator, or an io.EOF error if there aren't any more.
--- a/go/libraries/doltcore/sqle/tables.go
+++ b/go/libraries/doltcore/sqle/tables.go
@@ -47,6 +47,7 @@ const (
 	partitionMultiplier = 2.0
 )

+var MaxRowsPerPartition uint64 = 32 * 1024
 var MinRowsPerPartition uint64 = 1024

 func init() {
@@ -358,15 +359,20 @@ func (t *DoltTable) Partitions(ctx *sql.Context) (sql.PartitionIter, error) {
 		return newDoltTablePartitionIter(rowData, doltTablePartition{0, 0, rowData}), nil
 	}

-	maxPartitions := uint64(partitionMultiplier * runtime.NumCPU())
-	numPartitions := (numElements / MinRowsPerPartition) + 1
+	itemsPerPartition := MaxRowsPerPartition
+	numPartitions := (numElements / itemsPerPartition) + 1
+	if numPartitions < uint64(partitionMultiplier*runtime.NumCPU()) {
+		itemsPerPartition = numElements / uint64(partitionMultiplier*runtime.NumCPU())

-	if numPartitions > maxPartitions {
-		numPartitions = maxPartitions
+		if itemsPerPartition == 0 {
+			itemsPerPartition = numElements
+			numPartitions = 1
+		} else {
+			numPartitions = (numElements / itemsPerPartition) + 1
+		}
 	}

 	partitions := make([]doltTablePartition, numPartitions)
-	itemsPerPartition := numElements / numPartitions
 	for i := uint64(0); i < numPartitions-1; i++ {
 		partitions[i] = doltTablePartition{i * itemsPerPartition, (i + 1) * itemsPerPartition, rowData}
 	}
--- a/integration-tests/bats/auto_increment.bats
+++ b/integration-tests/bats/auto_increment.bats
@@ -23,13 +23,13 @@ teardown() {

    run dolt sql -q "INSERT INTO test (c0) VALUES (44);"
    [ "$status" -eq 0 ]
-    run dolt sql -q "SELECT * FROM test WHERE c0 = 44;" -r csv
+    run dolt sql -q "SELECT * FROM test WHERE c0 = 44 ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "4,44" ]] || false

    run dolt sql -q "INSERT INTO test (c0) VALUES (55),(66);"
    [ "$status" -eq 0 ]
-    run dolt sql -q "SELECT * FROM test WHERE c0 > 50;" -r csv
+    run dolt sql -q "SELECT * FROM test WHERE c0 > 50 ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "5,55" ]] || false
    [[ "$output" =~ "6,66" ]] || false
@@ -45,7 +45,7 @@ CREATE TABLE ai (
 INSERT INTO ai VALUES (NULL,1),(NULL,2),(NULL,3);
 SQL
    [ "$status" -eq 0 ]
-    run dolt sql -q "SELECT * FROM ai;" -r csv
+    run dolt sql -q "SELECT * FROM ai ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "1,1" ]] || false
    [[ "$output" =~ "2,2" ]] || false
@@ -55,7 +55,7 @@ SQL
@test "auto_increment: insert into empty auto_increment table" {
    run dolt sql -q "INSERT INTO test (c0) VALUES (1);"
    [ "$status" -eq 0 ]
-    run dolt sql -q "SELECT * FROM test WHERE c0 = 1;" -r csv
+    run dolt sql -q "SELECT * FROM test WHERE c0 = 1 ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "1,1" ]] || false
 }
@@ -67,7 +67,7 @@ SQL
    [ "$status" -eq 0 ]
    run dolt sql -q "INSERT INTO test VALUES (2,2);"
    [ "$status" -eq 0 ]
-    run dolt sql -q "SELECT * FROM test;" -r csv
+    run dolt sql -q "SELECT * FROM test ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "1,1" ]] || false
    [[ "$output" =~ "2,2" ]] || false
@@ -80,13 +80,14 @@ SQL

    run dolt sql -q "INSERT INTO test (pk,c0) VALUES (NULL,2);"
    [ "$status" -eq 0 ]
-    run dolt sql -q "SELECT * FROM test WHERE c0 > 1;" -r csv
+    run dolt sql -q "SELECT * FROM test WHERE c0 > 1 ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "2,2" ]] || false

    run dolt sql -q "INSERT INTO test VALUES (NULL,3), (10,10), (NULL,11);"
    [ "$status" -eq 0 ]
-    run dolt sql -q "SELECT * FROM test WHERE c0 > 2;" -r csv
+    run dolt sql -q "SELECT * FROM test WHERE c0 > 2 ORDER BY pk;" -r csv
+    echo $output
    [ "$status" -eq 0 ]
    [[ "$output" =~ "3,3" ]] || false
    [[ "$output" =~ "10,10" ]] || false
@@ -98,13 +99,13 @@ SQL

    run dolt sql -q "INSERT INTO test (pk,c0) VALUES (0,2);"
    [ "$status" -eq 0 ]
-    run dolt sql -q "SELECT * FROM test WHERE c0 > 1;" -r csv
+    run dolt sql -q "SELECT * FROM test WHERE c0 > 1 ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "2,2" ]] || false

    run dolt sql -q "INSERT INTO test VALUES (0,3), (10,10), (0,11);"
    [ "$status" -eq 0 ]
-    run dolt sql -q "SELECT * FROM test WHERE c0 > 2;" -r csv
+    run dolt sql -q "SELECT * FROM test WHERE c0 > 2 ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "3,3" ]] || false
    [[ "$output" =~ "10,10" ]] || false
@@ -121,7 +122,7 @@ INSERT INTO test (c0) VALUES (21);
 SQL
    [ "$status" -eq 0 ]

-    run dolt sql -q "SELECT * FROM test;" -r csv
+    run dolt sql -q "SELECT * FROM test ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "1,1" ]] || false
    [[ "$output" =~ "2,2" ]] || false
@@ -143,7 +144,7 @@ INSERT INTO test2 SELECT (pk + 20), c0 FROM test2;
 SQL
    [ "$status" -eq 0 ]

-    run dolt sql -q "select * from test2 order by pk" -r csv
+    run dolt sql -q "select * from test2 ORDER BY pk" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "1,1" ]] || false
    [[ "$output" =~ "2,2" ]] || false
@@ -170,7 +171,7 @@ SQL
    dolt sql -q "INSERT INTO auto_float (pk, c0) VALUES (3.9,4);"
    dolt sql -q "INSERT INTO auto_float (c0) VALUES (5);"

-    run dolt sql -q "SELECT * FROM auto_float;" -r csv
+    run dolt sql -q "SELECT * FROM auto_float ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "1,1" ]] || false
    [[ "$output" =~ "2.1,2" ]] || false
@@ -193,7 +194,7 @@ SQL
        echo "$TYPE"
        run dolt sql -q "INSERT INTO auto_$TYPE (c0) VALUES (2);"
        [ "$status" -eq 0 ]
-        run dolt sql -q "SELECT * FROM auto_$TYPE WHERE c0 > 1;" -r csv
+        run dolt sql -q "SELECT * FROM auto_$TYPE WHERE c0 > 1 ORDER BY pk;" -r csv
        [ "$status" -eq 0 ]
        [[ "$output" =~ "2,2" ]] || false
    done
@@ -211,7 +212,7 @@ SQL
        echo "$TYPE"
        run dolt sql -q "INSERT INTO auto2_$TYPE (c0) VALUES (2);"
        [ "$status" -eq 0 ]
-        run dolt sql -q "SELECT * FROM auto2_$TYPE WHERE c0 > 1;" -r csv
+        run dolt sql -q "SELECT * FROM auto2_$TYPE WHERE c0 > 1 ORDER BY pk;" -r csv
        [ "$status" -eq 0 ]
        [[ "$output" =~ "2,2" ]] || false
    done
@@ -248,7 +249,7 @@ SQL
    dolt merge other

    dolt sql -q "INSERT INTO test VALUES (NULL,22);"
-    run dolt sql -q "SELECT pk FROM test WHERE c0 = 22;" -r csv
+    run dolt sql -q "SELECT pk FROM test WHERE c0 = 22 ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "22" ]] || false
 }
@@ -271,7 +272,7 @@ SQL
    dolt checkout master
    dolt merge other
    dolt sql -q "INSERT INTO test VALUES (NULL,22);"
-    run dolt sql -q "SELECT pk FROM test WHERE c0 = 22;" -r csv
+    run dolt sql -q "SELECT pk FROM test WHERE c0 = 22 ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "22" ]] || false
 }
@@ -280,7 +281,7 @@ SQL
    run dolt sql -q "ALTER TABLE test AUTO_INCREMENT = 10;"
    [ "$status" -eq 0 ]
    dolt sql -q "INSERT INTO test VALUES (NULL,10);"
-    run dolt sql -q "SELECT * FROM test;" -r csv
+    run dolt sql -q "SELECT * FROM test ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "10,10" ]] || false

@@ -288,7 +289,7 @@ SQL
 ALTER TABLE test AUTO_INCREMENT = 20;
 INSERT INTO test VALUES (NULL,20),(30,30),(NULL,31);
 SQL
-    run dolt sql -q "SELECT * FROM test;" -r csv
+    run dolt sql -q "SELECT * FROM test ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "10,10" ]] || false
    [[ "$output" =~ "20,20" ]] || false
@@ -310,7 +311,7 @@ ALTER TABLE index_test ADD INDEX (c0);
 INSERT INTO index_test (c0) VALUES (4),(5),(6);
 SQL

-    run dolt sql -q "select * from index_test" -r csv
+    run dolt sql -q "select * from index_test ORDER BY pk" -r csv
    [ "$status" -eq 0 ]
    [[ "${lines[0]}" =~ "pk,c0" ]] || false
    [[ "${lines[1]}" =~ "1,1" ]] || false
@@ -329,7 +330,7 @@ SQL

    dolt sql -q "INSERT INTO test (c0) SELECT pk FROM other;"

-    run dolt sql -q "SELECT * FROM test;" -r csv
+    run dolt sql -q "SELECT * FROM test ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "${lines[0]}" =~ "pk,c0" ]] || false
    [[ "${lines[1]}" =~ "1,1" ]] || false
@@ -351,7 +352,7 @@ TRUNCATE t;
 INSERT INTO t (c0) VALUES (1),(2),(3);
 SQL

-    run dolt sql -q "SELECT * FROM t;" -r csv
+    run dolt sql -q "SELECT * FROM t ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "${lines[0]}" =~ "pk,c0" ]] || false
    [[ "${lines[1]}" =~ "1,1" ]] || false
@@ -370,7 +371,7 @@ INSERT INTO t (c0) VALUES (1),(2),(3);
 INSERT INTO t (c0) VALUES (4),(5),(6);
 SQL

-    run dolt sql -q "SELECT * FROM t;" -r csv
+    run dolt sql -q "SELECT * FROM t ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "${lines[0]}" =~ "pk,c0" ]] || false
    [[ "${lines[1]}" =~ "1,1" ]] || false
@@ -393,7 +394,7 @@ INSERT INTO t VALUES (4, 4);
 INSERT INTO t (c0) VALUES (5),(6),(7);
 SQL

-    run dolt sql -q "SELECT * FROM t;" -r csv
+    run dolt sql -q "SELECT * FROM t ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "${lines[0]}" =~ "pk,c0" ]] || false
    [[ "${lines[1]}" =~ "1,1" ]] || false
@@ -419,7 +420,7 @@ INSERT into t VALUES (3, 3);
 INSERT INTO t (c0) VALUES (8);
 SQL

-    run dolt sql -q "SELECT * FROM t;" -r csv
+    run dolt sql -q "SELECT * FROM t ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "${lines[0]}" =~ "pk,c0" ]] || false
    [[ "${lines[1]}" =~ "1,1" ]] || false
@@ -451,7 +452,7 @@ SELECT DOLT_MERGE('test');
 INSERT INTO t VALUES (NULL,5),(6,6),(NULL,7);
 SQL

-    run dolt sql -q "SELECT * FROM t;" -r csv
+    run dolt sql -q "SELECT * FROM t ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "${lines[0]}" =~ "pk,c0" ]] || false
    [[ "${lines[1]}" =~ "1,1" ]] || false
@@ -482,7 +483,7 @@ SELECT DOLT_MERGE('test');
 INSERT INTO t VALUES (10,10),(NULL,11);
 SQL

-    run dolt sql -q "SELECT * FROM t;" -r csv
+    run dolt sql -q "SELECT * FROM t ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "${lines[0]}" =~ "pk,c0" ]] || false
    [[ "${lines[1]}" =~ "1,1" ]] || false
@@ -515,7 +516,7 @@ SELECT DOLT_MERGE('test');
 INSERT INTO t VALUES (3,3),(NULL,6);
 SQL

-    run dolt sql -q "SELECT * FROM t;" -r csv
+    run dolt sql -q "SELECT * FROM t ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "${lines[0]}" =~ "pk,c0" ]] || false
    [[ "${lines[1]}" =~ "1,1" ]] || false
@@ -548,7 +549,7 @@ SELECT DOLT_MERGE('test');
 INSERT INTO t VALUES (NULL,6);
 SQL

-    run dolt sql -q "SELECT * FROM t;" -r csv
+    run dolt sql -q "SELECT * FROM t ORDER BY pk;" -r csv
    [ "$status" -eq 0 ]
    [[ "${lines[0]}" =~ "pk,c0" ]] || false
    [[ "${lines[1]}" =~ "1,1" ]] || false
@@ -564,7 +565,7 @@ SQL
    [ "$status" -eq 0 ]

    dolt sql -q "insert into test2 values (0, 'john', 0)"
-    run dolt sql -q "SELECT * from test2" -r csv
+    run dolt sql -q "SELECT * from test2 ORDER BY pk" -r csv
    [ "$status" -eq 0 ]
    [[ "$output" =~ "0,john,0" ]] || false
 }
@@ -574,7 +575,7 @@ SQL
    dolt sql -q "ALTER TABLE t CHANGE COLUMN pk pk int NOT NULL AUTO_INCREMENT PRIMARY KEY;"

    dolt sql -q 'insert into t values (NULL), (NULL), (NULL)'
-    run dolt sql -q "SELECT * FROM t" -r csv
+    run dolt sql -q "SELECT * FROM t ORDER BY pk" -r csv
    [[ "${lines[0]}" =~ "pk" ]] || false
    [[ "${lines[1]}" =~ "1" ]] || false
    [[ "${lines[2]}" =~ "2" ]] || false
--- a/integration-tests/bats/helper/index-on-writes-common.bash
+++ b/integration-tests/bats/helper/index-on-writes-common.bash
@@ -100,12 +100,12 @@ test_mutation() {
    expected="$3"
    uses_pk="$4"
    dolt sql -q "$dml"
-    run dolt sql -q "select * from $table" -r csv
+    run dolt sql -q "select * from $table ORDER BY pk1" -r csv
    [ "$status" -eq "0" ]
    [ "$output" == "$expected" ] || (echo $output && exit 1)
    dolt reset --hard
    dolt sql --batch -q "$dml ; $dml"
-    run dolt sql -q "select * from $table" -r csv
+    run dolt sql -q "select * from $table ORDER BY pk1" -r csv
    [ "$status" -eq "0" ]
    [ "$output" == "$expected" ] || (echo $output && exit 1)
    run dolt sql -q "explain $dml"
--- a/integration-tests/bats/import-create-tables.bats
+++ b/integration-tests/bats/import-create-tables.bats
@@ -392,7 +392,7 @@ SQL
    run dolt ls
    [ "$status" -eq 0 ]
    [[ "$output" =~ "test" ]] || false
-    run dolt sql -q "select * from test"
+    run dolt sql -q "select * from test ORDER BY pk"
    [ "$status" -eq 0 ]
    [ "${#lines[@]}" -eq 11 ]
    [ "${lines[3]}" = '| a  | ""        | 1         |' ]
@@ -418,7 +418,7 @@ SQL
    [ "$status" -eq 0 ]

    # schema argument subsets the data and adds empty column
-    run dolt sql -r csv -q "select * from subset"
+    run dolt sql -r csv -q "select * from subset ORDER BY pk"
    [ "$status" -eq 0 ]
    [ "${lines[0]}" = "pk,c1,c3,noData" ]
    [ "${lines[1]}" = "0,1,3," ]
@@ -440,7 +440,7 @@ SQL
    run dolt ls
    [ "$status" -eq 0 ]
    [[ "$output" =~ "empty_strings_null_values" ]] || false
-    run dolt sql -q "select * from empty_strings_null_values"
+    run dolt sql -q "select * from empty_strings_null_values ORDER BY pk"
    [ "$status" -eq 0 ]
    [ "${#lines[@]}" -eq 11 ]
    [ "${lines[3]}" = '| a  | ""        | 1         |' ]
@@ -467,7 +467,7 @@ SQL
    run dolt ls
    [ "$status" -eq 0 ]
    [[ "$output" =~ "test" ]] || false
-    run dolt sql -q "select * from test"
+    run dolt sql -q "select * from test ORDER BY pk"
    [ "$status" -eq 0 ]
    [ "${#lines[@]}" -eq 11 ]
    [ "${lines[3]}" = '| a  | ""        | 1         |' ]
--- a/integration-tests/bats/index.bats
+++ b/integration-tests/bats/index.bats
@@ -2025,13 +2025,13 @@ CREATE TABLE test2 (
 INSERT INTO test VALUES (0, NULL), (1, NULL), (2, NULL);
 INSERT INTO test2 VALUES (0, NULL, NULL), (1, NULL, NULL), (2, 1, NULL), (3, 1, NULL), (4, NULL, 1), (5, NULL, 1);
 SQL
-    run dolt sql -q "SELECT * FROM test" -r=json
+    run dolt sql -q "SELECT * FROM test order by pk" -r=json
    [ "$status" -eq "0" ]
    [[ "$output" =~ '{"rows": [{"pk":0},{"pk":1},{"pk":2}]}' ]] || false
-    run dolt sql -q "SELECT * FROM test WHERE v1 IS NULL" -r=json
+    run dolt sql -q "SELECT * FROM test WHERE v1 IS NULL order by pk" -r=json
    [ "$status" -eq "0" ]
    [[ "$output" =~ '{"rows": [{"pk":0},{"pk":1},{"pk":2}]}' ]] || false
-    run dolt sql -q "SELECT * FROM test2" -r=json
+    run dolt sql -q "SELECT * FROM test2 order by pk" -r=json
    [ "$status" -eq "0" ]
    [[ "$output" =~ '{"rows": [{"pk":0},{"pk":1},{"pk":2,"v1":1},{"pk":3,"v1":1},{"pk":4,"v2":1},{"pk":5,"v2":1}]}' ]] || false
 }
--- a/integration-tests/bats/merge.bats
+++ b/integration-tests/bats/merge.bats
@@ -336,7 +336,7 @@ SQL
    dolt checkout master
    run dolt merge other
    [ "$status" -eq 0 ]
-    run dolt sql -q "SELECT * FROM quiz;" -r csv
+    run dolt sql -q "SELECT * FROM quiz ORDER BY pk;" -r csv
    [[ "${lines[0]}" =~ "pk" ]] || false
    [[ "${lines[1]}" =~ "10" ]] || false
    [[ "${lines[2]}" =~ "11" ]] || false
--- a/integration-tests/bats/query-catalog.bats
+++ b/integration-tests/bats/query-catalog.bats
@@ -100,7 +100,7 @@ teardown() {

@test "query-catalog: executed saved" {
    Q1="select pk, pk1, pk2 from one_pk,two_pk where one_pk.c1=two_pk.c1 order by 1"
-    Q2="select pk from one_pk"
+    Q2="select pk from one_pk order by pk"
    dolt sql -q "$Q1" -s name1
    dolt sql -q "$Q2" -s name2

@@ -138,7 +138,7 @@ EOF
    EXPECTED=$(cat <<'EOF'
 id,display_order,name,query,description
 name1,1,name1,"select pk, pk1, pk2 from one_pk,two_pk where one_pk.c1=two_pk.c1 order by 1",""
-name2,2,name2,select pk from one_pk,""
+name2,2,name2,select pk from one_pk order by pk,""
 EOF
 )

@@ -154,7 +154,7 @@ EOF
    EXPECTED=$(cat <<'EOF'
 id,display_order,name,query,description
 name1,1,name1,"select pk, pk1, pk2 from one_pk,two_pk where one_pk.c1=two_pk.c1 and pk < 3 order by 1 desc",""
-name2,2,name2,select pk from one_pk,""
+name2,2,name2,select pk from one_pk order by pk,""
 EOF
 )

--- a/integration-tests/bats/sql-server.bats
+++ b/integration-tests/bats/sql-server.bats
@@ -683,7 +683,7 @@ SQL
     SELECT DOLT_MERGE('feature-branch');
     "

-     server_query repo1 1 "SELECT * FROM test" "pk\n1\n2\n3\n1000"
+     server_query repo1 1 "SELECT * FROM test ORDER BY pk" "pk\n1\n2\n3\n1000"

     server_query repo1 1 "SELECT COUNT(*) FROM dolt_log" "COUNT(*)\n3"
 }
--- a/integration-tests/bats/sql.bats
+++ b/integration-tests/bats/sql.bats
@@ -324,23 +324,23 @@ SQL
 }

@test "sql: basic inner join" {
-    run dolt sql -q "select pk,pk1,pk2 from one_pk join two_pk on one_pk.c1=two_pk.c1"
+    run dolt sql -q "select pk,pk1,pk2 from one_pk join two_pk on one_pk.c1=two_pk.c1 order by pk"
    [ "$status" -eq 0 ]
    [ "${#lines[@]}" -eq 8 ]
    first_join_output=$output
-    run dolt sql -q "select pk,pk1,pk2 from two_pk join one_pk on one_pk.c1=two_pk.c1"
+    run dolt sql -q "select pk,pk1,pk2 from two_pk join one_pk on one_pk.c1=two_pk.c1 order by pk"
    [ "$status" -eq 0 ]
    [ "${#lines[@]}" -eq 8 ]
    [ "$output" = "$first_join_output" ]
-    run dolt sql -q "select pk,pk1,pk2 from one_pk join two_pk on one_pk.c1=two_pk.c1 where pk=1"
+    run dolt sql -q "select pk,pk1,pk2 from one_pk join two_pk on one_pk.c1=two_pk.c1 where pk=1 order by pk"
    [ "$status" -eq 0 ]
    [ "${#lines[@]}" -eq 5 ]
-    run dolt sql -q "select pk,pk1,pk2,one_pk.c1 as foo,two_pk.c1 as bar from one_pk join two_pk on one_pk.c1=two_pk.c1"
+    run dolt sql -q "select pk,pk1,pk2,one_pk.c1 as foo,two_pk.c1 as bar from one_pk join two_pk on one_pk.c1=two_pk.c1 order by pk"
    [ "$status" -eq 0 ]
    [[ "$output" =~ foo ]] || false
    [[ "$output" =~ bar ]] || false
    [ "${#lines[@]}" -eq 8 ]
-    run dolt sql -q "select pk,pk1,pk2,one_pk.c1 as foo,two_pk.c1 as bar from one_pk join two_pk on one_pk.c1=two_pk.c1  where one_pk.c1=10"
+    run dolt sql -q "select pk,pk1,pk2,one_pk.c1 as foo,two_pk.c1 as bar from one_pk join two_pk on one_pk.c1=two_pk.c1 where one_pk.c1=10 order by pk"
    [ "$status" -eq 0 ]
    [ "${#lines[@]}" -eq 5 ]
    [[ "$output" =~ "10" ]] || false