// Events is an endpoint that returns the latest event log entries, with the following
// optional URL parameters:
//
// type=STRING returns events with this type (e.g. "create_table")
// targetID=INT returns events for that have this targetID
func (s *adminServer) Events(
ctx context.Context, req *serverpb.EventsRequest,
) (*serverpb.EventsResponse, error) {
args := sql.SessionArgs{User: s.getUser(req)}
session := s.NewSessionForRPC(ctx, args)
defer session.Finish(s.server.sqlExecutor)
// Execute the query.
q := makeSQLQuery()
q.Append("SELECT timestamp, eventType, targetID, reportingID, info, uniqueID ")
q.Append("FROM system.eventlog ")
q.Append("WHERE true ") // This simplifies the WHERE clause logic below.
if len(req.Type) > 0 {
q.Append("AND eventType = $ ", parser.NewDString(req.Type))
}
if req.TargetId > 0 {
q.Append("AND targetID = $ ", parser.NewDInt(parser.DInt(req.TargetId)))
}
q.Append("ORDER BY timestamp DESC ")
q.Append("LIMIT $", parser.NewDInt(parser.DInt(apiEventLimit)))
if len(q.Errors()) > 0 {
return nil, s.serverErrors(q.Errors())
}
r := s.server.sqlExecutor.ExecuteStatements(session, q.String(), q.QueryArguments())
defer r.Close()
if err := s.checkQueryResults(r.ResultList, 1); err != nil {
return nil, s.serverError(err)
}
// Marshal response.
var resp serverpb.EventsResponse
scanner := makeResultScanner(r.ResultList[0].Columns)
for i, nRows := 0, r.ResultList[0].Rows.Len(); i < nRows; i++ {
row := r.ResultList[0].Rows.At(i)
var event serverpb.EventsResponse_Event
var ts time.Time
if err := scanner.ScanIndex(row, 0, &ts); err != nil {
return nil, err
}
event.Timestamp = serverpb.EventsResponse_Event_Timestamp{Sec: ts.Unix(), Nsec: uint32(ts.Nanosecond())}
if err := scanner.ScanIndex(row, 1, &event.EventType); err != nil {
return nil, err
}
if err := scanner.ScanIndex(row, 2, &event.TargetID); err != nil {
return nil, err
}
if err := scanner.ScanIndex(row, 3, &event.ReportingID); err != nil {
return nil, err
}
if err := scanner.ScanIndex(row, 4, &event.Info); err != nil {
return nil, err
}
if err := scanner.ScanIndex(row, 5, &event.UniqueID); err != nil {
return nil, err
}
resp.Events = append(resp.Events, event)
}
return &resp, nil
}
// golangFillQueryArguments populates the placeholder map with
// types and values from an array of Go values.
// TODO: This does not support arguments of the SQL 'Date' type, as there is not
// an equivalent type in Go's standard library. It's not currently needed by any
// of our internal tables.
func golangFillQueryArguments(pinfo *parser.PlaceholderInfo, args []interface{}) {
pinfo.Clear()
for i, arg := range args {
k := fmt.Sprint(i + 1)
if arg == nil {
pinfo.SetValue(k, parser.DNull)
continue
}
// A type switch to handle a few explicit types with special semantics:
// - Datums are passed along as is.
// - Time datatypes get special representation in the database.
var d parser.Datum
switch t := arg.(type) {
case parser.Datum:
d = t
case time.Time:
d = parser.MakeDTimestamp(t, time.Microsecond)
case time.Duration:
d = &parser.DInterval{Duration: duration.Duration{Nanos: t.Nanoseconds()}}
case *inf.Dec:
dd := &parser.DDecimal{}
dd.Set(t)
d = dd
}
if d == nil {
// Handle all types which have an underlying type that can be stored in the
// database.
// Note: if this reflection becomes a performance concern in the future,
// commonly used types could be added explicitly into the type switch above
// for a performance gain.
val := reflect.ValueOf(arg)
switch val.Kind() {
case reflect.Bool:
d = parser.MakeDBool(parser.DBool(val.Bool()))
case reflect.Int, reflect.Int8, reflect.Int16, reflect.Int32, reflect.Int64:
d = parser.NewDInt(parser.DInt(val.Int()))
case reflect.Uint, reflect.Uint8, reflect.Uint16, reflect.Uint32, reflect.Uint64:
d = parser.NewDInt(parser.DInt(val.Uint()))
case reflect.Float32, reflect.Float64:
d = parser.NewDFloat(parser.DFloat(val.Float()))
case reflect.String:
d = parser.NewDString(val.String())
case reflect.Slice:
// Handle byte slices.
if val.Type().Elem().Kind() == reflect.Uint8 {
d = parser.NewDBytes(parser.DBytes(val.Bytes()))
}
}
if d == nil {
panic(fmt.Sprintf("unexpected type %T", arg))
}
}
pinfo.SetValue(k, d)
}
}
// populateExplain invokes explain() with a makeRow method
// which populates a valuesNode.
func (e *explainer) populateExplain(v *valuesNode, plan planNode) error {
e.makeRow = func(level int, name, field, description string, plan planNode) {
if e.err != nil {
return
}
row := parser.DTuple{
parser.NewDInt(parser.DInt(level)),
parser.NewDString(name),
parser.NewDString(field),
parser.NewDString(description),
}
if e.showMetadata {
if plan != nil {
row = append(row, parser.NewDString(formatColumns(plan.Columns(), e.showTypes)))
row = append(row, parser.NewDString(plan.Ordering().AsString(plan.Columns())))
} else {
row = append(row, emptyString, emptyString)
}
}
if _, err := v.rows.AddRow(row); err != nil {
e.err = err
}
}
e.err = nil
e.explain(plan)
return e.err
}
// queryZone retrieves the specific ZoneConfig associated with the supplied ID,
// if it exists.
func (s *adminServer) queryZone(
session *sql.Session, id sqlbase.ID,
) (config.ZoneConfig, bool, error) {
const query = `SELECT config FROM system.zones WHERE id = $1`
params := parser.NewPlaceholderInfo()
params.SetValue(`1`, parser.NewDInt(parser.DInt(id)))
r := s.server.sqlExecutor.ExecuteStatements(session, query, params)
defer r.Close()
if err := s.checkQueryResults(r.ResultList, 1); err != nil {
return config.ZoneConfig{}, false, err
}
result := r.ResultList[0]
if result.Rows.Len() == 0 {
return config.ZoneConfig{}, false, nil
}
var zoneBytes []byte
scanner := resultScanner{}
err := scanner.ScanIndex(result.Rows.At(0), 0, &zoneBytes)
if err != nil {
return config.ZoneConfig{}, false, err
}
var zone config.ZoneConfig
if err := zone.Unmarshal(zoneBytes); err != nil {
return config.ZoneConfig{}, false, err
}
return zone, true, nil
}
// queryNamespaceID queries for the ID of the namespace with the given name and
// parent ID.
func (s *adminServer) queryNamespaceID(
session *sql.Session, parentID sqlbase.ID, name string,
) (sqlbase.ID, error) {
const query = `SELECT id FROM system.namespace WHERE parentID = $1 AND name = $2`
params := parser.NewPlaceholderInfo()
params.SetValue(`1`, parser.NewDInt(parser.DInt(parentID)))
params.SetValue(`2`, parser.NewDString(name))
r := s.server.sqlExecutor.ExecuteStatements(session, query, params)
defer r.Close()
if err := s.checkQueryResults(r.ResultList, 1); err != nil {
return 0, err
}
result := r.ResultList[0]
if result.Rows.Len() == 0 {
return 0, errors.Errorf("namespace %s with ParentID %d not found", name, parentID)
}
var id int64
scanner := resultScanner{}
err := scanner.ScanIndex(result.Rows.At(0), 0, &id)
if err != nil {
return 0, err
}
return sqlbase.ID(id), nil
}
// DecodeTableValue decodes a value encoded by EncodeTableValue.
func DecodeTableValue(a *DatumAlloc, valType parser.Type, b []byte) (parser.Datum, []byte, error) {
_, dataOffset, _, typ, err := encoding.DecodeValueTag(b)
if err != nil {
return nil, b, err
}
if typ == encoding.Null {
return parser.DNull, b[dataOffset:], nil
}
switch valType {
case parser.TypeBool:
var x bool
b, x, err = encoding.DecodeBoolValue(b)
// No need to chunk allocate DBool as MakeDBool returns either
// parser.DBoolTrue or parser.DBoolFalse.
return parser.MakeDBool(parser.DBool(x)), b, err
case parser.TypeInt:
var i int64
b, i, err = encoding.DecodeIntValue(b)
return a.NewDInt(parser.DInt(i)), b, err
case parser.TypeFloat:
var f float64
b, f, err = encoding.DecodeFloatValue(b)
return a.NewDFloat(parser.DFloat(f)), b, err
case parser.TypeDecimal:
var d *inf.Dec
b, d, err = encoding.DecodeDecimalValue(b)
dd := a.NewDDecimal(parser.DDecimal{})
dd.Set(d)
return dd, b, err
case parser.TypeString:
var data []byte
b, data, err = encoding.DecodeBytesValue(b)
return a.NewDString(parser.DString(data)), b, err
case parser.TypeBytes:
var data []byte
b, data, err = encoding.DecodeBytesValue(b)
return a.NewDBytes(parser.DBytes(data)), b, err
case parser.TypeDate:
var i int64
b, i, err = encoding.DecodeIntValue(b)
return a.NewDDate(parser.DDate(i)), b, err
case parser.TypeTimestamp:
var t time.Time
b, t, err = encoding.DecodeTimeValue(b)
return a.NewDTimestamp(parser.DTimestamp{Time: t}), b, err
case parser.TypeTimestampTZ:
var t time.Time
b, t, err = encoding.DecodeTimeValue(b)
return a.NewDTimestampTZ(parser.DTimestampTZ{Time: t}), b, err
case parser.TypeInterval:
var d duration.Duration
b, d, err = encoding.DecodeDurationValue(b)
return a.NewDInterval(parser.DInterval{Duration: d}), b, err
default:
return nil, nil, errors.Errorf("TODO(pmattis): decoded index value: %s", valType)
}
}
func benchmarkWriteArray(b *testing.B, format formatCode) {
a := parser.NewDArray(parser.TypeInt)
for i := 0; i < 3; i++ {
if err := a.Append(parser.NewDInt(parser.DInt(1234))); err != nil {
b.Fatal(err)
}
}
benchmarkWriteType(b, a, format)
}
func (o *ordinalityNode) Next() (bool, error) {
hasNext, err := o.source.Next()
if !hasNext || err != nil {
return hasNext, err
}
copy(o.row, o.source.Values())
// o.row was allocated one spot larger than o.source.Values().
// Store the ordinality value there.
o.row[len(o.row)-1] = parser.NewDInt(parser.DInt(o.curCnt))
o.curCnt++
return true, nil
}
// colIDArrayToDatum returns an int[] containing the ColumnIDs, or NULL if there
// are no ColumnIDs.
func colIDArrayToDatum(arr []sqlbase.ColumnID) (parser.Datum, error) {
if len(arr) == 0 {
return parser.DNull, nil
}
d := parser.NewDArray(parser.TypeInt)
for _, val := range arr {
if err := d.Append(parser.NewDInt(parser.DInt(val))); err != nil {
return nil, err
}
}
return d, nil
}
func TestRowContainer(t *testing.T) {
defer leaktest.AfterTest(t)()
for _, numCols := range []int{1, 2, 3, 5, 10, 15} {
for _, numRows := range []int{5, 10, 100} {
for _, numPops := range []int{0, 1, 2, numRows / 3, numRows / 2} {
resCol := make(ResultColumns, numCols)
for i := range resCol {
resCol[i] = ResultColumn{Typ: parser.TypeInt}
}
m := mon.MakeUnlimitedMonitor(context.Background(), "test", nil, nil, math.MaxInt64)
rc := NewRowContainer(m.MakeBoundAccount(context.Background()), resCol, 0)
row := make(parser.DTuple, numCols)
for i := 0; i < numRows; i++ {
for j := range row {
row[j] = parser.NewDInt(parser.DInt(i*numCols + j))
}
if err := rc.AddRow(row); err != nil {
t.Fatal(err)
}
}
for i := 0; i < numPops; i++ {
rc.PopFirst()
}
// Given that we just deleted numPops rows, we have numRows -
// numPops rows remaining.
if rc.Len() != numRows-numPops {
t.Fatalf("invalid length, expected %d got %d", numRows-numPops, rc.Len())
}
// what was previously rc.At(i + numPops) is now rc.At(i).
for i := 0; i < rc.Len(); i++ {
row := rc.At(i)
for j := range row {
dint, ok := row[j].(*parser.DInt)
if !ok || int(*dint) != (i+numPops)*numCols+j {
t.Fatalf("invalid value %+v on row %d, col %d", row[j], i+numPops, j)
}
}
}
}
}
}
}
// RandDatum generates a random Datum of the given type.
// If null is true, the datum can be DNull.
func RandDatum(rng *rand.Rand, typ ColumnType_Kind, null bool) parser.Datum {
if null && rng.Intn(10) == 0 {
return parser.DNull
}
switch typ {
case ColumnType_BOOL:
return parser.MakeDBool(rng.Intn(2) == 1)
case ColumnType_INT:
return parser.NewDInt(parser.DInt(rng.Int63()))
case ColumnType_FLOAT:
return parser.NewDFloat(parser.DFloat(rng.NormFloat64()))
case ColumnType_DECIMAL:
d := &parser.DDecimal{}
d.Dec.SetScale(inf.Scale(rng.Intn(40) - 20))
d.Dec.SetUnscaled(rng.Int63())
return d
case ColumnType_DATE:
return parser.NewDDate(parser.DDate(rng.Intn(10000)))
case ColumnType_TIMESTAMP:
return &parser.DTimestamp{Time: time.Unix(rng.Int63n(1000000), rng.Int63n(1000000))}
case ColumnType_INTERVAL:
return &parser.DInterval{Duration: duration.Duration{Months: rng.Int63n(1000),
Days: rng.Int63n(1000),
Nanos: rng.Int63n(1000000),
}}
case ColumnType_STRING:
// Generate a random ASCII string.
p := make([]byte, rng.Intn(10))
for i := range p {
p[i] = byte(1 + rng.Intn(127))
}
return parser.NewDString(string(p))
case ColumnType_BYTES:
p := make([]byte, rng.Intn(10))
_, _ = rng.Read(p)
return parser.NewDBytes(parser.DBytes(p))
case ColumnType_TIMESTAMPTZ:
return &parser.DTimestampTZ{Time: time.Unix(rng.Int63n(1000000), rng.Int63n(1000000))}
case ColumnType_INT_ARRAY:
// TODO(cuongdo): we don't support for persistence of arrays yet
return parser.DNull
default:
panic(fmt.Sprintf("invalid type %s", typ))
}
}
// MakePrimaryIndexKey creates a key prefix that corresponds to a table row
// (in the primary index); it is intended for tests.
//
// The value types must match the primary key columns (or a prefix of them);
// supported types are: - Datum
// - bool (converts to DBool)
// - int (converts to DInt)
// - string (converts to DString)
func MakePrimaryIndexKey(desc *TableDescriptor, vals ...interface{}) (roachpb.Key, error) {
index := &desc.PrimaryIndex
if len(vals) > len(index.ColumnIDs) {
return nil, errors.Errorf("got %d values, PK has %d columns", len(vals), len(index.ColumnIDs))
}
datums := make([]parser.Datum, len(vals))
for i, v := range vals {
switch v := v.(type) {
case bool:
datums[i] = parser.MakeDBool(parser.DBool(v))
case int:
datums[i] = parser.NewDInt(parser.DInt(v))
case string:
datums[i] = parser.NewDString(v)
case parser.Datum:
datums[i] = v
default:
return nil, errors.Errorf("unexpected value type %T", v)
}
// Check that the value type matches.
colID := index.ColumnIDs[i]
for _, c := range desc.Columns {
if c.ID == colID {
if t := DatumTypeToColumnKind(datums[i].ResolvedType()); t != c.Type.Kind {
return nil, errors.Errorf("column %d of type %s, got value of type %s", i, c.Type.Kind, t)
}
break
}
}
}
// Create the ColumnID to index in datums slice map needed by
// MakeIndexKeyPrefix.
colIDToRowIndex := make(map[ColumnID]int)
for i := range vals {
colIDToRowIndex[index.ColumnIDs[i]] = i
}
keyPrefix := MakeIndexKeyPrefix(desc, index.ID)
key, _, err := EncodeIndexKey(desc, index, colIDToRowIndex, datums, keyPrefix)
if err != nil {
return nil, err
}
return roachpb.Key(key), nil
}
func TestUnorderedSync(t *testing.T) {
defer leaktest.AfterTest(t)()
columnTypeInt := &sqlbase.ColumnType{Kind: sqlbase.ColumnType_INT}
mrc := &MultiplexedRowChannel{}
mrc.Init(5)
for i := 1; i <= 5; i++ {
go func(i int) {
for j := 1; j <= 100; j++ {
a := sqlbase.DatumToEncDatum(*columnTypeInt, parser.NewDInt(parser.DInt(i)))
b := sqlbase.DatumToEncDatum(*columnTypeInt, parser.NewDInt(parser.DInt(j)))
row := sqlbase.EncDatumRow{a, b}
mrc.PushRow(row)
}
mrc.Close(nil)
}(i)
}
var retRows sqlbase.EncDatumRows
for {
row, err := mrc.NextRow()
if err != nil {
t.Fatal(err)
}
if row == nil {
break
}
retRows = append(retRows, row)
}
// Verify all elements.
for i := 1; i <= 5; i++ {
j := 1
for _, row := range retRows {
if int(*(row[0].Datum.(*parser.DInt))) == i {
if int(*(row[1].Datum.(*parser.DInt))) != j {
t.Errorf("Expected [%d %d], got %s", i, j, row)
}
j++
}
}
if j != 101 {
t.Errorf("Missing [%d %d]", i, j)
}
}
// Test case when one source closes with an error.
mrc = &MultiplexedRowChannel{}
mrc.Init(5)
for i := 1; i <= 5; i++ {
go func(i int) {
for j := 1; j <= 100; j++ {
a := sqlbase.DatumToEncDatum(*columnTypeInt, parser.NewDInt(parser.DInt(i)))
b := sqlbase.DatumToEncDatum(*columnTypeInt, parser.NewDInt(parser.DInt(j)))
row := sqlbase.EncDatumRow{a, b}
mrc.PushRow(row)
}
var err error
if i == 3 {
err = fmt.Errorf("Test error")
}
mrc.Close(err)
}(i)
}
for {
row, err := mrc.NextRow()
if err != nil {
if err.Error() != "Test error" {
t.Error(err)
}
break
}
if row == nil {
t.Error("Did not receive expected error")
}
}
}
// TestAdminAPIZoneDetails verifies the zone configuration information returned
// for both DatabaseDetailsResponse AND TableDetailsResponse.
func TestAdminAPIZoneDetails(t *testing.T) {
defer leaktest.AfterTest(t)()
s, _, _ := serverutils.StartServer(t, base.TestServerArgs{})
defer s.Stopper().Stop()
ts := s.(*TestServer)
// Create database and table.
ac := log.AmbientContext{Tracer: tracing.NewTracer()}
ctx, span := ac.AnnotateCtxWithSpan(context.Background(), "test")
defer span.Finish()
session := sql.NewSession(
ctx, sql.SessionArgs{User: security.RootUser}, ts.sqlExecutor, nil, &sql.MemoryMetrics{})
session.StartUnlimitedMonitor()
setupQueries := []string{
"CREATE DATABASE test",
"CREATE TABLE test.tbl (val STRING)",
}
for _, q := range setupQueries {
res := ts.sqlExecutor.ExecuteStatements(session, q, nil)
defer res.Close()
if res.ResultList[0].Err != nil {
t.Fatalf("error executing '%s': %s", q, res.ResultList[0].Err)
}
}
// Function to verify the zone for table "test.tbl" as returned by the Admin
// API.
verifyTblZone := func(
expectedZone config.ZoneConfig, expectedLevel serverpb.ZoneConfigurationLevel,
) {
var resp serverpb.TableDetailsResponse
if err := getAdminJSONProto(s, "databases/test/tables/tbl", &resp); err != nil {
t.Fatal(err)
}
if a, e := &resp.ZoneConfig, &expectedZone; !proto.Equal(a, e) {
t.Errorf("actual table zone config %v did not match expected value %v", a, e)
}
if a, e := resp.ZoneConfigLevel, expectedLevel; a != e {
t.Errorf("actual table ZoneConfigurationLevel %s did not match expected value %s", a, e)
}
if t.Failed() {
t.FailNow()
}
}
// Function to verify the zone for database "test" as returned by the Admin
// API.
verifyDbZone := func(
expectedZone config.ZoneConfig, expectedLevel serverpb.ZoneConfigurationLevel,
) {
var resp serverpb.DatabaseDetailsResponse
if err := getAdminJSONProto(s, "databases/test", &resp); err != nil {
t.Fatal(err)
}
if a, e := &resp.ZoneConfig, &expectedZone; !proto.Equal(a, e) {
t.Errorf("actual db zone config %v did not match expected value %v", a, e)
}
if a, e := resp.ZoneConfigLevel, expectedLevel; a != e {
t.Errorf("actual db ZoneConfigurationLevel %s did not match expected value %s", a, e)
}
if t.Failed() {
t.FailNow()
}
}
// Function to store a zone config for a given object ID.
setZone := func(zoneCfg config.ZoneConfig, id sqlbase.ID) {
zoneBytes, err := zoneCfg.Marshal()
if err != nil {
t.Fatal(err)
}
const query = `INSERT INTO system.zones VALUES($1, $2)`
params := parser.NewPlaceholderInfo()
params.SetValue(`1`, parser.NewDInt(parser.DInt(id)))
params.SetValue(`2`, parser.NewDBytes(parser.DBytes(zoneBytes)))
res := ts.sqlExecutor.ExecuteStatements(session, query, params)
defer res.Close()
if res.ResultList[0].Err != nil {
t.Fatalf("error executing '%s': %s", query, res.ResultList[0].Err)
}
}
// Verify zone matches cluster default.
verifyDbZone(config.DefaultZoneConfig(), serverpb.ZoneConfigurationLevel_CLUSTER)
verifyTblZone(config.DefaultZoneConfig(), serverpb.ZoneConfigurationLevel_CLUSTER)
// Get ID path for table. This will be an array of three IDs, containing the ID of the root namespace,
// the database, and the table (in that order).
idPath, err := ts.admin.queryDescriptorIDPath(session, []string{"test", "tbl"})
if err != nil {
t.Fatal(err)
}
// Apply zone configuration to database and check again.
dbZone := config.ZoneConfig{
RangeMinBytes: 456,
}
setZone(dbZone, idPath[1])
verifyDbZone(dbZone, serverpb.ZoneConfigurationLevel_DATABASE)
verifyTblZone(dbZone, serverpb.ZoneConfigurationLevel_DATABASE)
// Apply zone configuration to table and check again.
tblZone := config.ZoneConfig{
RangeMinBytes: 789,
}
setZone(tblZone, idPath[2])
verifyDbZone(dbZone, serverpb.ZoneConfigurationLevel_DATABASE)
verifyTblZone(tblZone, serverpb.ZoneConfigurationLevel_TABLE)
}
func TestOrderedSync(t *testing.T) {
defer leaktest.AfterTest(t)()
columnTypeInt := &sqlbase.ColumnType{Kind: sqlbase.ColumnType_INT}
v := [6]sqlbase.EncDatum{}
for i := range v {
v[i] = sqlbase.DatumToEncDatum(*columnTypeInt, parser.NewDInt(parser.DInt(i)))
}
asc := encoding.Ascending
desc := encoding.Descending
testCases := []struct {
sources []sqlbase.EncDatumRows
ordering sqlbase.ColumnOrdering
expected sqlbase.EncDatumRows
}{
{
sources: []sqlbase.EncDatumRows{
{
{v[0], v[1], v[4]},
{v[0], v[1], v[2]},
{v[0], v[2], v[3]},
{v[1], v[1], v[3]},
},
{
{v[1], v[0], v[4]},
},
{
{v[0], v[0], v[0]},
{v[4], v[4], v[4]},
},
},
ordering: sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: asc},
{ColIdx: 1, Direction: asc},
},
expected: sqlbase.EncDatumRows{
{v[0], v[0], v[0]},
{v[0], v[1], v[4]},
{v[0], v[1], v[2]},
{v[0], v[2], v[3]},
{v[1], v[0], v[4]},
{v[1], v[1], v[3]},
{v[4], v[4], v[4]},
},
},
{
sources: []sqlbase.EncDatumRows{
{},
{
{v[1], v[0], v[4]},
},
{
{v[3], v[4], v[1]},
{v[4], v[4], v[4]},
{v[3], v[2], v[0]},
},
{
{v[4], v[4], v[5]},
{v[3], v[3], v[0]},
{v[0], v[0], v[0]},
},
},
ordering: sqlbase.ColumnOrdering{
{ColIdx: 1, Direction: desc},
{ColIdx: 0, Direction: asc},
{ColIdx: 2, Direction: asc},
},
expected: sqlbase.EncDatumRows{
{v[3], v[4], v[1]},
{v[4], v[4], v[4]},
{v[4], v[4], v[5]},
{v[3], v[3], v[0]},
{v[3], v[2], v[0]},
{v[0], v[0], v[0]},
{v[1], v[0], v[4]},
},
},
}
for testIdx, c := range testCases {
var sources []RowSource
for _, srcRows := range c.sources {
rowBuf := &RowBuffer{rows: srcRows}
sources = append(sources, rowBuf)
}
src, err := makeOrderedSync(c.ordering, sources)
if err != nil {
t.Fatal(err)
}
var retRows sqlbase.EncDatumRows
for {
row, err := src.NextRow()
if err != nil {
t.Fatal(err)
}
if row == nil {
break
}
retRows = append(retRows, row)
}
expStr := c.expected.String()
retStr := retRows.String()
if expStr != retStr {
t.Errorf("invalid results for case %d; expected:\n %s\ngot:\n %s",
testIdx, expStr, retStr)
}
}
}
func TestMergeJoiner(t *testing.T) {
defer leaktest.AfterTest(t)()
v := [6]sqlbase.EncDatum{}
for i := range v {
v[i] = sqlbase.DatumToEncDatum(sqlbase.ColumnType_INT, parser.NewDInt(parser.DInt(i)))
}
null := sqlbase.EncDatum{Datum: parser.DNull}
testCases := []struct {
spec MergeJoinerSpec
inputs []sqlbase.EncDatumRows
expected sqlbase.EncDatumRows
}{
{
spec: MergeJoinerSpec{
LeftOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: encoding.Ascending},
}),
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: encoding.Ascending},
}),
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_INNER,
OutputColumns: []uint32{0, 3, 4},
// Implicit @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[1], v[4]},
{v[2], v[4]},
{v[3], v[1]},
{v[4], v[5]},
{v[5], v[5]},
},
{
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
},
},
expected: sqlbase.EncDatumRows{
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
},
},
{
spec: MergeJoinerSpec{
LeftOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: encoding.Ascending},
}),
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: encoding.Ascending},
}),
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_INNER,
OutputColumns: []uint32{0, 1, 3},
// Implicit @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[0], v[1]},
},
{
{v[0], v[4]},
{v[0], v[1]},
{v[0], v[0]},
{v[0], v[5]},
{v[0], v[4]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], v[0], v[4]},
{v[0], v[0], v[1]},
{v[0], v[0], v[0]},
{v[0], v[0], v[5]},
{v[0], v[0], v[4]},
{v[0], v[1], v[4]},
{v[0], v[1], v[1]},
{v[0], v[1], v[0]},
{v[0], v[1], v[5]},
{v[0], v[1], v[4]},
},
},
{
spec: MergeJoinerSpec{
LeftOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: encoding.Ascending},
}),
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: encoding.Ascending},
}),
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_INNER,
OutputColumns: []uint32{0, 1, 3},
Expr: Expression{Expr: "@4 >= 4"},
// Implicit AND @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[0], v[1]},
{v[1], v[0]},
{v[1], v[1]},
},
{
{v[0], v[4]},
{v[0], v[1]},
{v[0], v[0]},
{v[0], v[5]},
{v[0], v[4]},
{v[1], v[4]},
{v[1], v[1]},
{v[1], v[0]},
{v[1], v[5]},
{v[1], v[4]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], v[0], v[4]},
{v[0], v[0], v[5]},
{v[0], v[0], v[4]},
{v[0], v[1], v[4]},
{v[0], v[1], v[5]},
{v[0], v[1], v[4]},
{v[1], v[0], v[4]},
{v[1], v[0], v[5]},
{v[1], v[0], v[4]},
{v[1], v[1], v[4]},
{v[1], v[1], v[5]},
{v[1], v[1], v[4]},
},
},
{
spec: MergeJoinerSpec{
LeftOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: encoding.Ascending},
}),
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: encoding.Ascending},
}),
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_LEFT_OUTER,
OutputColumns: []uint32{0, 3, 4},
// Implicit @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[1], v[4]},
{v[2], v[4]},
{v[3], v[1]},
{v[4], v[5]},
{v[5], v[5]},
},
{
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], null, null},
{v[1], v[0], v[4]},
{v[2], null, null},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
{v[5], null, null},
},
},
{
spec: MergeJoinerSpec{
LeftOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: encoding.Ascending},
}),
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: encoding.Ascending},
}),
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_RIGHT_OUTER,
OutputColumns: []uint32{3, 1, 2},
// Implicit @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
},
{
{v[0], v[0]},
{v[1], v[4]},
{v[2], v[4]},
{v[3], v[1]},
{v[4], v[5]},
{v[5], v[5]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], null, null},
{v[1], v[0], v[4]},
{v[2], null, null},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
{v[5], null, null},
},
},
{
spec: MergeJoinerSpec{
LeftOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: encoding.Ascending},
}),
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: encoding.Ascending},
}),
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_FULL_OUTER,
OutputColumns: []uint32{0, 3, 4},
// Implicit @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[1], v[4]},
{v[2], v[4]},
{v[3], v[1]},
{v[4], v[5]},
},
{
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
{v[5], v[5], v[1]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], null, null},
{v[1], v[0], v[4]},
{v[2], null, null},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
{null, v[5], v[1]},
},
},
}
for _, c := range testCases {
ms := c.spec
inputs := []RowSource{&RowBuffer{rows: c.inputs[0]}, &RowBuffer{rows: c.inputs[1]}}
out := &RowBuffer{}
flowCtx := FlowCtx{Context: context.Background(), evalCtx: &parser.EvalContext{}}
m, err := newMergeJoiner(&flowCtx, &ms, inputs, out)
if err != nil {
t.Fatal(err)
}
m.Run(nil)
var retRows sqlbase.EncDatumRows
for {
row, err := out.NextRow()
if err != nil {
t.Fatal(err)
}
if row == nil {
break
}
retRows = append(retRows, row)
}
expStr := c.expected.String()
retStr := retRows.String()
if expStr != retStr {
t.Errorf("invalid results; expected:\n %s\ngot:\n %s",
expStr, retStr)
}
}
}
func TestSorter(t *testing.T) {
defer leaktest.AfterTest(t)()
v := [6]sqlbase.EncDatum{}
for i := range v {
v[i].SetDatum(sqlbase.ColumnType_INT, parser.NewDInt(parser.DInt(i)))
}
asc := encoding.Ascending
desc := encoding.Descending
testCases := []struct {
spec SorterSpec
input sqlbase.EncDatumRows
expected sqlbase.EncDatumRows
}{
{
// No specified input ordering and unspecified limit.
spec: SorterSpec{
OutputOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: asc},
{ColIdx: 1, Direction: desc},
{ColIdx: 2, Direction: asc},
}),
},
input: sqlbase.EncDatumRows{
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[4], v[4], v[4]},
{v[3], v[2], v[0]},
{v[4], v[4], v[5]},
{v[3], v[3], v[0]},
{v[0], v[0], v[0]},
},
expected: sqlbase.EncDatumRows{
{v[0], v[0], v[0]},
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[3], v[3], v[0]},
{v[3], v[2], v[0]},
{v[4], v[4], v[4]},
{v[4], v[4], v[5]},
},
}, {
// No specified input ordering but specified limit.
spec: SorterSpec{
Limit: 4,
OutputOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: asc},
{ColIdx: 1, Direction: asc},
{ColIdx: 2, Direction: asc},
}),
},
input: sqlbase.EncDatumRows{
{v[3], v[3], v[0]},
{v[3], v[4], v[1]},
{v[1], v[0], v[4]},
{v[0], v[0], v[0]},
{v[4], v[4], v[4]},
{v[4], v[4], v[5]},
{v[3], v[2], v[0]},
},
expected: sqlbase.EncDatumRows{
{v[0], v[0], v[0]},
{v[1], v[0], v[4]},
{v[3], v[2], v[0]},
{v[3], v[3], v[0]},
},
}, {
// Specified match ordering length but no specified limit.
spec: SorterSpec{
OrderingMatchLen: 2,
OutputOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 0, Direction: asc},
{ColIdx: 1, Direction: asc},
{ColIdx: 2, Direction: asc},
}),
},
input: sqlbase.EncDatumRows{
{v[0], v[1], v[2]},
{v[0], v[1], v[0]},
{v[1], v[0], v[5]},
{v[1], v[1], v[5]},
{v[1], v[1], v[4]},
{v[3], v[4], v[3]},
{v[3], v[4], v[2]},
{v[3], v[5], v[1]},
{v[4], v[4], v[5]},
{v[4], v[4], v[4]},
},
expected: sqlbase.EncDatumRows{
{v[0], v[1], v[0]},
{v[0], v[1], v[2]},
{v[1], v[0], v[5]},
{v[1], v[1], v[4]},
{v[1], v[1], v[5]},
{v[3], v[4], v[2]},
{v[3], v[4], v[3]},
{v[3], v[5], v[1]},
{v[4], v[4], v[4]},
{v[4], v[4], v[5]},
},
}, {
// Specified input ordering but no specified limit.
spec: SorterSpec{
OrderingMatchLen: 2,
OutputOrdering: convertToSpecOrdering(
sqlbase.ColumnOrdering{
{ColIdx: 1, Direction: asc},
{ColIdx: 2, Direction: asc},
{ColIdx: 3, Direction: asc},
}),
},
input: sqlbase.EncDatumRows{
{v[1], v[1], v[2], v[5]},
{v[0], v[1], v[2], v[4]},
{v[0], v[1], v[2], v[3]},
{v[1], v[1], v[2], v[2]},
{v[1], v[2], v[2], v[5]},
{v[0], v[2], v[2], v[4]},
{v[0], v[2], v[2], v[3]},
{v[1], v[2], v[2], v[2]},
},
expected: sqlbase.EncDatumRows{
{v[1], v[1], v[2], v[2]},
{v[0], v[1], v[2], v[3]},
{v[0], v[1], v[2], v[4]},
{v[1], v[1], v[2], v[5]},
{v[1], v[2], v[2], v[2]},
{v[0], v[2], v[2], v[3]},
{v[0], v[2], v[2], v[4]},
{v[1], v[2], v[2], v[5]},
},
},
}
for _, c := range testCases {
ss := c.spec
in := &RowBuffer{rows: c.input}
out := &RowBuffer{}
flowCtx := FlowCtx{Context: context.Background()}
s := newSorter(&flowCtx, &ss, in, out)
s.Run(nil)
var retRows sqlbase.EncDatumRows
for {
row, err := out.NextRow()
if err != nil {
t.Fatal(err)
}
if row == nil {
break
}
retRows = append(retRows, row)
}
expStr := c.expected.String()
retStr := retRows.String()
if expStr != retStr {
t.Errorf("invalid results; expected:\n %s\ngot:\n %s",
expStr, retStr)
}
}
}
// RowModuloFn creates a GenValueFn that returns the row number modulo a given
// value as a DInt
func RowModuloFn(modulo int) GenValueFn {
return func(row int) parser.Datum {
return parser.NewDInt(parser.DInt(row % modulo))
}
}
func TestDistinct(t *testing.T) {
defer leaktest.AfterTest(t)()
v := [15]sqlbase.EncDatum{}
for i := range v {
v[i] = sqlbase.DatumToEncDatum(sqlbase.ColumnType{Kind: sqlbase.ColumnType_INT},
parser.NewDInt(parser.DInt(i)))
}
testCases := []struct {
spec DistinctSpec
input sqlbase.EncDatumRows
expected sqlbase.EncDatumRows
}{
{
spec: DistinctSpec{},
input: sqlbase.EncDatumRows{
{v[2], v[3]},
{v[5], v[6]},
{v[2], v[3]},
{v[5], v[6]},
{v[2], v[6]},
{v[3], v[5]},
{v[2], v[9]},
},
expected: sqlbase.EncDatumRows{
{v[2], v[3]},
{v[5], v[6]},
{v[2], v[6]},
{v[3], v[5]},
{v[2], v[9]},
},
}, {
spec: DistinctSpec{
OrderedColumns: []uint32{1},
},
input: sqlbase.EncDatumRows{
{v[2], v[3]},
{v[2], v[3]},
{v[2], v[6]},
{v[2], v[9]},
{v[3], v[5]},
{v[5], v[6]},
{v[5], v[6]},
},
expected: sqlbase.EncDatumRows{
{v[2], v[3]},
{v[2], v[6]},
{v[2], v[9]},
{v[3], v[5]},
{v[5], v[6]},
},
},
}
for _, c := range testCases {
ds := c.spec
in := NewRowBuffer(nil, c.input)
out := &RowBuffer{}
flowCtx := FlowCtx{
Context: context.Background(),
}
d, err := newDistinct(&flowCtx, &ds, in, out)
if err != nil {
t.Fatal(err)
}
d.Run(nil)
if out.Err != nil {
t.Fatal(out.Err)
}
if !out.Closed {
t.Fatalf("output RowReceiver not closed")
}
if result := out.Rows.String(); result != c.expected.String() {
t.Errorf("invalid results: %s, expected %s'", result, c.expected.String())
}
}
}
func TestHashJoiner(t *testing.T) {
defer leaktest.AfterTest(t)()
v := [10]sqlbase.EncDatum{}
for i := range v {
v[i] = sqlbase.DatumToEncDatum(sqlbase.ColumnType_INT, parser.NewDInt(parser.DInt(i)))
}
null := sqlbase.EncDatum{Datum: parser.DNull}
testCases := []struct {
spec HashJoinerSpec
inputs []sqlbase.EncDatumRows
expected sqlbase.EncDatumRows
}{
{
spec: HashJoinerSpec{
LeftEqColumns: []uint32{0},
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightEqColumns: []uint32{0},
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_INNER,
OutputColumns: []uint32{0, 3, 4},
// Implicit @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[1], v[4]},
{v[2], v[4]},
{v[3], v[1]},
{v[4], v[5]},
{v[5], v[5]},
},
{
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
},
},
expected: sqlbase.EncDatumRows{
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
},
},
{
spec: HashJoinerSpec{
LeftEqColumns: []uint32{0},
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightEqColumns: []uint32{0},
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_INNER,
OutputColumns: []uint32{0, 1, 3},
// Implicit @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[0], v[1]},
},
{
{v[0], v[4]},
{v[0], v[1]},
{v[0], v[0]},
{v[0], v[5]},
{v[0], v[4]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], v[0], v[4]},
{v[0], v[0], v[1]},
{v[0], v[0], v[0]},
{v[0], v[0], v[5]},
{v[0], v[0], v[4]},
{v[0], v[1], v[4]},
{v[0], v[1], v[1]},
{v[0], v[1], v[0]},
{v[0], v[1], v[5]},
{v[0], v[1], v[4]},
},
},
{
spec: HashJoinerSpec{
LeftEqColumns: []uint32{0},
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightEqColumns: []uint32{0},
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_INNER,
OutputColumns: []uint32{0, 1, 3},
Expr: Expression{Expr: "@4 >= 4"},
// Implicit AND @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[0], v[1]},
{v[1], v[0]},
{v[1], v[1]},
},
{
{v[0], v[4]},
{v[0], v[1]},
{v[0], v[0]},
{v[0], v[5]},
{v[0], v[4]},
{v[1], v[4]},
{v[1], v[1]},
{v[1], v[0]},
{v[1], v[5]},
{v[1], v[4]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], v[0], v[4]},
{v[0], v[0], v[5]},
{v[0], v[0], v[4]},
{v[0], v[1], v[4]},
{v[0], v[1], v[5]},
{v[0], v[1], v[4]},
{v[1], v[0], v[4]},
{v[1], v[0], v[5]},
{v[1], v[0], v[4]},
{v[1], v[1], v[4]},
{v[1], v[1], v[5]},
{v[1], v[1], v[4]},
},
},
{
spec: HashJoinerSpec{
LeftEqColumns: []uint32{0},
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightEqColumns: []uint32{0},
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_LEFT_OUTER,
OutputColumns: []uint32{0, 3, 4},
// Implicit @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[1], v[4]},
{v[2], v[4]},
{v[3], v[1]},
{v[4], v[5]},
{v[5], v[5]},
},
{
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], null, null},
{v[1], v[0], v[4]},
{v[2], null, null},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
{v[5], null, null},
},
},
{
spec: HashJoinerSpec{
LeftEqColumns: []uint32{0},
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightEqColumns: []uint32{0},
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_RIGHT_OUTER,
OutputColumns: []uint32{3, 1, 2},
// Implicit @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
},
{
{v[0], v[0]},
{v[1], v[4]},
{v[2], v[4]},
{v[3], v[1]},
{v[4], v[5]},
{v[5], v[5]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], null, null},
{v[1], v[0], v[4]},
{v[2], null, null},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
{v[5], null, null},
},
},
{
spec: HashJoinerSpec{
LeftEqColumns: []uint32{0},
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightEqColumns: []uint32{0},
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_FULL_OUTER,
OutputColumns: []uint32{0, 3, 4},
// Implicit @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[1], v[4]},
{v[2], v[4]},
{v[3], v[1]},
{v[4], v[5]},
},
{
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
{v[5], v[5], v[1]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], null, null},
{v[1], v[0], v[4]},
{v[2], null, null},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
{null, v[5], v[1]},
},
},
{
spec: HashJoinerSpec{
LeftEqColumns: []uint32{0},
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightEqColumns: []uint32{0},
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_INNER,
OutputColumns: []uint32{0, 3, 4},
// Implicit @1 = @3 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[2], v[4]},
{v[3], v[1]},
{v[4], v[5]},
{v[5], v[5]},
},
{
{v[1], v[0], v[4]},
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
},
},
expected: sqlbase.EncDatumRows{
{v[3], v[4], v[1]},
{v[4], v[4], v[5]},
},
},
// Tests for behavior when input contains NULLs.
{
spec: HashJoinerSpec{
LeftEqColumns: []uint32{0, 1},
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightEqColumns: []uint32{0, 1},
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_INNER,
OutputColumns: []uint32{0, 1, 2, 3, 4},
// Implicit @1,@2 = @3,@4 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[1], null},
{null, v[2]},
{null, null},
},
{
{v[0], v[0], v[4]},
{v[1], null, v[5]},
{null, v[2], v[6]},
{null, null, v[7]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], v[0], v[0], v[0], v[4]},
},
},
{
spec: HashJoinerSpec{
LeftEqColumns: []uint32{0, 1},
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightEqColumns: []uint32{0, 1},
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_LEFT_OUTER,
OutputColumns: []uint32{0, 1, 2, 3, 4},
// Implicit @1,@2 = @3,@4 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[1], null},
{null, v[2]},
{null, null},
},
{
{v[0], v[0], v[4]},
{v[1], null, v[5]},
{null, v[2], v[6]},
{null, null, v[7]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], v[0], v[0], v[0], v[4]},
{v[1], null, null, null, null},
{null, v[2], null, null, null},
{null, null, null, null, null},
},
},
{
spec: HashJoinerSpec{
LeftEqColumns: []uint32{0, 1},
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightEqColumns: []uint32{0, 1},
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_RIGHT_OUTER,
OutputColumns: []uint32{0, 1, 2, 3, 4},
// Implicit @1,@2 = @3,@4 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[1], null},
{null, v[2]},
{null, null},
},
{
{v[0], v[0], v[4]},
{v[1], null, v[5]},
{null, v[2], v[6]},
{null, null, v[7]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], v[0], v[0], v[0], v[4]},
{null, null, v[1], null, v[5]},
{null, null, null, v[2], v[6]},
{null, null, null, null, v[7]},
},
},
{
spec: HashJoinerSpec{
LeftEqColumns: []uint32{0, 1},
LeftTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
RightEqColumns: []uint32{0, 1},
RightTypes: []sqlbase.ColumnType_Kind{
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
sqlbase.ColumnType_INT,
},
Type: JoinType_FULL_OUTER,
OutputColumns: []uint32{0, 1, 2, 3, 4},
// Implicit @1,@2 = @3,@4 constraint.
},
inputs: []sqlbase.EncDatumRows{
{
{v[0], v[0]},
{v[1], null},
{null, v[2]},
{null, null},
},
{
{v[0], v[0], v[4]},
{v[1], null, v[5]},
{null, v[2], v[6]},
{null, null, v[7]},
},
},
expected: sqlbase.EncDatumRows{
{v[0], v[0], v[0], v[0], v[4]},
{null, null, v[1], null, v[5]},
{null, null, null, v[2], v[6]},
{null, null, null, null, v[7]},
{v[1], null, null, null, null},
{null, v[2], null, null, null},
{null, null, null, null, null},
},
},
}
for _, c := range testCases {
hs := c.spec
inputs := []RowSource{&RowBuffer{rows: c.inputs[0]}, &RowBuffer{rows: c.inputs[1]}}
out := &RowBuffer{}
flowCtx := FlowCtx{Context: context.Background(), evalCtx: &parser.EvalContext{}}
h, err := newHashJoiner(&flowCtx, &hs, inputs, out)
if err != nil {
t.Fatal(err)
}
h.Run(nil)
if out.err != nil {
t.Fatal(out.err)
}
if !out.closed {
t.Fatalf("output RowReceiver not closed")
}
var expected []string
for _, row := range c.expected {
expected = append(expected, row.String())
}
sort.Strings(expected)
expStr := strings.Join(expected, "")
var rets []string
for {
row, err := out.NextRow()
if err != nil {
t.Fatal(err)
}
if row == nil {
break
}
rets = append(rets, row.String())
}
sort.Strings(rets)
retStr := strings.Join(rets, "")
if expStr != retStr {
t.Errorf("invalid results; expected:\n %s\ngot:\n %s",
expStr, retStr)
}
}
}
// TODO(irfansharif): Add tests to verify the following aggregation functions:
// AVG
// BOOL_AND
// BOOL_OR
// CONCAT_AGG
// STDDEV
// VARIANCE
// TODO(irfansharif): Replicate sql/testdata and TestLogic for distsql, this kind of manual
// case-by-case testing is error prone making it very easy to miss edge cases.
func TestAggregator(t *testing.T) {
defer leaktest.AfterTest(t)()
v := [15]sqlbase.EncDatum{}
for i := range v {
v[i].SetDatum(sqlbase.ColumnType_INT, parser.NewDInt(parser.DInt(i)))
}
testCases := []struct {
spec AggregatorSpec
input sqlbase.EncDatumRows
expected sqlbase.EncDatumRows
}{
{
// SELECT $1, COUNT($0), GROUP BY $1.
spec: AggregatorSpec{
Types: []sqlbase.ColumnType_Kind{sqlbase.ColumnType_INT, sqlbase.ColumnType_INT},
GroupCols: []uint32{1},
Exprs: []AggregatorSpec_Expr{
{
Func: AggregatorSpec_IDENT,
ColIdx: 1,
},
{
Func: AggregatorSpec_COUNT,
ColIdx: 0,
},
},
},
input: sqlbase.EncDatumRows{
{v[1], v[2]},
{v[3], v[4]},
{v[6], v[2]},
{v[7], v[2]},
{v[8], v[4]},
},
expected: sqlbase.EncDatumRows{
{v[4], v[2]},
{v[2], v[3]},
},
}, {
// SELECT $1, SUM($0), GROUP BY $1.
spec: AggregatorSpec{
Types: []sqlbase.ColumnType_Kind{sqlbase.ColumnType_INT, sqlbase.ColumnType_INT},
GroupCols: []uint32{1},
Exprs: []AggregatorSpec_Expr{
{
Func: AggregatorSpec_IDENT,
ColIdx: 1,
},
{
Func: AggregatorSpec_SUM,
ColIdx: 0,
},
},
},
input: sqlbase.EncDatumRows{
{v[1], v[2]},
{v[3], v[4]},
{v[6], v[2]},
{v[7], v[2]},
{v[8], v[4]},
},
expected: sqlbase.EncDatumRows{
{v[2], v[14]},
{v[4], v[11]},
},
}, {
// SELECT COUNT($0), SUM($0), GROUP BY [] (empty group key).
spec: AggregatorSpec{
Types: []sqlbase.ColumnType_Kind{sqlbase.ColumnType_INT, sqlbase.ColumnType_INT},
Exprs: []AggregatorSpec_Expr{
{
Func: AggregatorSpec_COUNT,
ColIdx: 0,
},
{
Func: AggregatorSpec_SUM,
ColIdx: 0,
},
},
},
input: sqlbase.EncDatumRows{
{v[1], v[2]},
{v[1], v[4]},
{v[3], v[2]},
{v[4], v[2]},
{v[5], v[4]},
},
expected: sqlbase.EncDatumRows{
{v[5], v[14]},
},
},
{
// SELECT SUM DISTINCT ($0), GROUP BY [] (empty group key).
spec: AggregatorSpec{
Types: []sqlbase.ColumnType_Kind{sqlbase.ColumnType_INT, sqlbase.ColumnType_INT},
Exprs: []AggregatorSpec_Expr{
{
Func: AggregatorSpec_SUM,
Distinct: true,
ColIdx: 0,
},
},
},
input: sqlbase.EncDatumRows{
{v[2]},
{v[4]},
{v[2]},
{v[2]},
{v[4]},
},
expected: sqlbase.EncDatumRows{
{v[6]},
},
},
{
// SELECT $0, GROUP BY [] (empty group key).
spec: AggregatorSpec{
Types: []sqlbase.ColumnType_Kind{sqlbase.ColumnType_INT, sqlbase.ColumnType_INT},
Exprs: []AggregatorSpec_Expr{
{
Func: AggregatorSpec_IDENT,
ColIdx: 0,
},
},
},
input: sqlbase.EncDatumRows{
{v[1]},
{v[1]},
{v[1]},
},
expected: sqlbase.EncDatumRows{
{v[1]},
},
}, {
// SELECT MAX($0), MIN($1), COUNT($1), COUNT DISTINCT ($1), GROUP BY [] (empty group key).
spec: AggregatorSpec{
Types: []sqlbase.ColumnType_Kind{sqlbase.ColumnType_INT, sqlbase.ColumnType_INT},
Exprs: []AggregatorSpec_Expr{
{
Func: AggregatorSpec_MAX,
ColIdx: 0,
},
{
Func: AggregatorSpec_MIN,
ColIdx: 1,
},
{
Func: AggregatorSpec_COUNT,
ColIdx: 1,
},
{
Func: AggregatorSpec_COUNT,
Distinct: true,
ColIdx: 1,
},
},
},
input: sqlbase.EncDatumRows{
{v[2], v[2]},
{v[1], v[4]},
{v[3], v[2]},
{v[4], v[2]},
{v[5], v[4]},
},
expected: sqlbase.EncDatumRows{
{v[5], v[2], v[5], v[2]},
},
},
}
for _, c := range testCases {
ags := c.spec
in := &RowBuffer{rows: c.input}
out := &RowBuffer{}
flowCtx := FlowCtx{
Context: context.Background(),
evalCtx: &parser.EvalContext{},
}
ag, err := newAggregator(&flowCtx, &ags, in, out)
if err != nil {
t.Fatal(err)
}
ag.Run(nil)
var expected []string
for _, row := range c.expected {
expected = append(expected, row.String())
}
sort.Strings(expected)
expStr := strings.Join(expected, "")
var rets []string
for {
row, err := out.NextRow()
if err != nil {
t.Fatal(err)
}
if row == nil {
break
}
rets = append(rets, row.String())
}
sort.Strings(rets)
retStr := strings.Join(rets, "")
if expStr != retStr {
t.Errorf("invalid results; expected:\n %s\ngot:\n %s",
expStr, retStr)
}
}
}
// DecodeTableKey decodes a table key/value.
func DecodeTableKey(
a *DatumAlloc, valType parser.Type, key []byte, dir encoding.Direction,
) (parser.Datum, []byte, error) {
if (dir != encoding.Ascending) && (dir != encoding.Descending) {
return nil, nil, errors.Errorf("invalid direction: %d", dir)
}
var isNull bool
if key, isNull = encoding.DecodeIfNull(key); isNull {
return parser.DNull, key, nil
}
var rkey []byte
var err error
switch valType {
case parser.TypeBool:
var i int64
if dir == encoding.Ascending {
rkey, i, err = encoding.DecodeVarintAscending(key)
} else {
rkey, i, err = encoding.DecodeVarintDescending(key)
}
// No need to chunk allocate DBool as MakeDBool returns either
// parser.DBoolTrue or parser.DBoolFalse.
return parser.MakeDBool(parser.DBool(i != 0)), rkey, err
case parser.TypeInt:
var i int64
if dir == encoding.Ascending {
rkey, i, err = encoding.DecodeVarintAscending(key)
} else {
rkey, i, err = encoding.DecodeVarintDescending(key)
}
return a.NewDInt(parser.DInt(i)), rkey, err
case parser.TypeFloat:
var f float64
if dir == encoding.Ascending {
rkey, f, err = encoding.DecodeFloatAscending(key)
} else {
rkey, f, err = encoding.DecodeFloatDescending(key)
}
return a.NewDFloat(parser.DFloat(f)), rkey, err
case parser.TypeDecimal:
var d *inf.Dec
if dir == encoding.Ascending {
rkey, d, err = encoding.DecodeDecimalAscending(key, nil)
} else {
rkey, d, err = encoding.DecodeDecimalDescending(key, nil)
}
dd := a.NewDDecimal(parser.DDecimal{})
dd.Set(d)
return dd, rkey, err
case parser.TypeString:
var r string
if dir == encoding.Ascending {
rkey, r, err = encoding.DecodeUnsafeStringAscending(key, nil)
} else {
rkey, r, err = encoding.DecodeUnsafeStringDescending(key, nil)
}
return a.NewDString(parser.DString(r)), rkey, err
case parser.TypeBytes:
var r []byte
if dir == encoding.Ascending {
rkey, r, err = encoding.DecodeBytesAscending(key, nil)
} else {
rkey, r, err = encoding.DecodeBytesDescending(key, nil)
}
return a.NewDBytes(parser.DBytes(r)), rkey, err
case parser.TypeDate:
var t int64
if dir == encoding.Ascending {
rkey, t, err = encoding.DecodeVarintAscending(key)
} else {
rkey, t, err = encoding.DecodeVarintDescending(key)
}
return a.NewDDate(parser.DDate(t)), rkey, err
case parser.TypeTimestamp:
var t time.Time
if dir == encoding.Ascending {
rkey, t, err = encoding.DecodeTimeAscending(key)
} else {
rkey, t, err = encoding.DecodeTimeDescending(key)
}
return a.NewDTimestamp(parser.DTimestamp{Time: t}), rkey, err
case parser.TypeTimestampTZ:
var t time.Time
if dir == encoding.Ascending {
rkey, t, err = encoding.DecodeTimeAscending(key)
} else {
rkey, t, err = encoding.DecodeTimeDescending(key)
}
return a.NewDTimestampTZ(parser.DTimestampTZ{Time: t}), rkey, err
case parser.TypeInterval:
var d duration.Duration
if dir == encoding.Ascending {
rkey, d, err = encoding.DecodeDurationAscending(key)
} else {
rkey, d, err = encoding.DecodeDurationDescending(key)
}
return a.NewDInterval(parser.DInterval{Duration: d}), rkey, err
default:
return nil, nil, errors.Errorf("TODO(pmattis): decoded index key: %s", valType)
}
}
func TestTableReader(t *testing.T) {
defer leaktest.AfterTest(t)()
s, sqlDB, kvDB := serverutils.StartServer(t, base.TestServerArgs{})
defer s.Stopper().Stop()
// Create a table where each row is:
//
// | a | b | sum | s |
// |-----------------------------------------------------------------|
// | rowId/10 | rowId%10 | rowId/10 + rowId%10 | IntToEnglish(rowId) |
aFn := func(row int) parser.Datum {
return parser.NewDInt(parser.DInt(row / 10))
}
bFn := func(row int) parser.Datum {
return parser.NewDInt(parser.DInt(row % 10))
}
sumFn := func(row int) parser.Datum {
return parser.NewDInt(parser.DInt(row/10 + row%10))
}
sqlutils.CreateTable(t, sqlDB, "t",
"a INT, b INT, sum INT, s STRING, PRIMARY KEY (a,b), INDEX bs (b,s)",
99,
sqlutils.ToRowFn(aFn, bFn, sumFn, sqlutils.RowEnglishFn))
td := sqlbase.GetTableDescriptor(kvDB, "test", "t")
makeIndexSpan := func(start, end int) TableReaderSpan {
var span roachpb.Span
prefix := roachpb.Key(sqlbase.MakeIndexKeyPrefix(td, td.Indexes[0].ID))
span.Key = append(prefix, encoding.EncodeVarintAscending(nil, int64(start))...)
span.EndKey = append(span.EndKey, prefix...)
span.EndKey = append(span.EndKey, encoding.EncodeVarintAscending(nil, int64(end))...)
return TableReaderSpan{Span: span}
}
testCases := []struct {
spec TableReaderSpec
expected string
}{
{
spec: TableReaderSpec{
Filter: Expression{Expr: "@3 < 5 AND @2 != 3"}, // sum < 5 && b != 3
OutputColumns: []uint32{0, 1},
},
expected: "[[0 1] [0 2] [0 4] [1 0] [1 1] [1 2] [2 0] [2 1] [2 2] [3 0] [3 1] [4 0]]",
},
{
spec: TableReaderSpec{
Filter: Expression{Expr: "@3 < 5 AND @2 != 3"},
OutputColumns: []uint32{3}, // s
HardLimit: 4,
},
expected: "[['one'] ['two'] ['four'] ['one-zero']]",
},
{
spec: TableReaderSpec{
IndexIdx: 1,
Reverse: true,
Spans: []TableReaderSpan{makeIndexSpan(4, 6)},
Filter: Expression{Expr: "@1 < 3"}, // sum < 8
OutputColumns: []uint32{0, 1},
SoftLimit: 1,
},
expected: "[[2 5] [1 5] [0 5] [2 4] [1 4] [0 4]]",
},
}
for _, c := range testCases {
ts := c.spec
ts.Table = *td
flowCtx := FlowCtx{
Context: context.Background(),
evalCtx: &parser.EvalContext{},
txnProto: &roachpb.Transaction{},
clientDB: kvDB,
}
out := &RowBuffer{}
tr, err := newTableReader(&flowCtx, &ts, out)
if err != nil {
t.Fatal(err)
}
tr.Run(nil)
if out.err != nil {
t.Fatal(out.err)
}
if !out.closed {
t.Fatalf("output RowReceiver not closed")
}
if result := out.rows.String(); result != c.expected {
t.Errorf("invalid results: %s, expected %s'", result, c.expected)
}
}
}
func dumpTable(w io.Writer, conn *sqlConn, origDBName, origTableName string) error {
const limit = 100
// Escape names since they can't be used in placeholders.
dbname := parser.Name(origDBName).String()
tablename := parser.Name(origTableName).String()
if err := conn.Exec(fmt.Sprintf("SET DATABASE = %s", dbname), nil); err != nil {
return err
}
// Fetch all table metadata in a transaction and its time to guarantee it
// doesn't change between the various SHOW statements.
if err := conn.Exec("BEGIN", nil); err != nil {
return err
}
vals, err := conn.QueryRow("SELECT cluster_logical_timestamp()", nil)
if err != nil {
return err
}
clusterTS := string(vals[0].([]byte))
// A previous version of the code did a SELECT on system.descriptor. This
// required the SELECT privilege to the descriptor table, which only root
// has. Allowing non-root to do this would let users see other users' table
// descriptors which is a problem in multi-tenancy.
// Fetch column types.
rows, err := conn.Query(fmt.Sprintf("SHOW COLUMNS FROM %s", tablename), nil)
if err != nil {
return err
}
vals = make([]driver.Value, 2)
coltypes := make(map[string]string)
for {
if err := rows.Next(vals); err == io.EOF {
break
} else if err != nil {
return err
}
nameI, typI := vals[0], vals[1]
name, ok := nameI.(string)
if !ok {
return fmt.Errorf("unexpected value: %T", nameI)
}
typ, ok := typI.(string)
if !ok {
return fmt.Errorf("unexpected value: %T", typI)
}
coltypes[name] = typ
}
if err := rows.Close(); err != nil {
return err
}
// index holds the names, in order, of the primary key columns.
var index []string
// Primary index is always the first index returned by SHOW INDEX.
rows, err = conn.Query(fmt.Sprintf("SHOW INDEX FROM %s", tablename), nil)
if err != nil {
return err
}
vals = make([]driver.Value, 5)
var primaryIndex string
// Find the primary index columns.
for {
if err := rows.Next(vals); err == io.EOF {
break
} else if err != nil {
return err
}
b, ok := vals[1].(string)
if !ok {
return fmt.Errorf("unexpected value: %T", vals[1])
}
if primaryIndex == "" {
primaryIndex = b
} else if primaryIndex != b {
break
}
b, ok = vals[4].(string)
if !ok {
return fmt.Errorf("unexpected value: %T", vals[4])
}
index = append(index, parser.Name(b).String())
}
if err := rows.Close(); err != nil {
return err
}
if len(index) == 0 {
return fmt.Errorf("no primary key index found")
}
indexes := strings.Join(index, ", ")
// Build the SELECT query.
var sbuf bytes.Buffer
fmt.Fprintf(&sbuf, "SELECT %s, * FROM %s@%s AS OF SYSTEM TIME %s", indexes, tablename, primaryIndex, clusterTS)
var wbuf bytes.Buffer
fmt.Fprintf(&wbuf, " WHERE ROW (%s) > ROW (", indexes)
for i := range index {
if i > 0 {
wbuf.WriteString(", ")
}
fmt.Fprintf(&wbuf, "$%d", i+1)
}
wbuf.WriteString(")")
// No WHERE clause first time, so add a place to inject it.
fmt.Fprintf(&sbuf, "%%s ORDER BY %s LIMIT %d", indexes, limit)
bs := sbuf.String()
vals, err = conn.QueryRow(fmt.Sprintf("SHOW CREATE TABLE %s", tablename), nil)
if err != nil {
return err
}
create := vals[1].(string)
if _, err := w.Write([]byte(create)); err != nil {
return err
}
if _, err := w.Write([]byte(";\n")); err != nil {
return err
}
if err := conn.Exec("COMMIT", nil); err != nil {
return err
}
// pk holds the last values of the fetched primary keys
var pk []driver.Value
q := fmt.Sprintf(bs, "")
for {
rows, err := conn.Query(q, pk)
if err != nil {
return err
}
cols := rows.Columns()
pkcols := cols[:len(index)]
cols = cols[len(index):]
inserts := make([][]string, 0, limit)
i := 0
for i < limit {
vals := make([]driver.Value, len(cols)+len(pkcols))
if err := rows.Next(vals); err == io.EOF {
break
} else if err != nil {
return err
}
if pk == nil {
q = fmt.Sprintf(bs, wbuf.String())
}
pk = vals[:len(index)]
vals = vals[len(index):]
ivals := make([]string, len(vals))
// Values need to be correctly encoded for INSERT statements in a text file.
for si, sv := range vals {
switch t := sv.(type) {
case nil:
ivals[si] = "NULL"
case bool:
ivals[si] = parser.MakeDBool(parser.DBool(t)).String()
case int64:
ivals[si] = parser.NewDInt(parser.DInt(t)).String()
case float64:
ivals[si] = parser.NewDFloat(parser.DFloat(t)).String()
case string:
ivals[si] = parser.NewDString(t).String()
case []byte:
switch ct := coltypes[cols[si]]; ct {
case "INTERVAL":
ivals[si] = fmt.Sprintf("'%s'", t)
case "BYTES":
ivals[si] = parser.NewDBytes(parser.DBytes(t)).String()
default:
// STRING and DECIMAL types can have optional length
// suffixes, so only examine the prefix of the type.
if strings.HasPrefix(coltypes[cols[si]], "STRING") {
ivals[si] = parser.NewDString(string(t)).String()
} else if strings.HasPrefix(coltypes[cols[si]], "DECIMAL") {
ivals[si] = string(t)
} else {
panic(errors.Errorf("unknown []byte type: %s, %v: %s", t, cols[si], coltypes[cols[si]]))
}
}
case time.Time:
var d parser.Datum
ct := coltypes[cols[si]]
switch ct {
case "DATE":
d = parser.NewDDateFromTime(t, time.UTC)
case "TIMESTAMP":
d = parser.MakeDTimestamp(t, time.Nanosecond)
case "TIMESTAMP WITH TIME ZONE":
d = parser.MakeDTimestampTZ(t, time.Nanosecond)
default:
panic(errors.Errorf("unknown timestamp type: %s, %v: %s", t, cols[si], coltypes[cols[si]]))
}
ivals[si] = fmt.Sprintf("'%s'", d)
default:
panic(errors.Errorf("unknown field type: %T (%s)", t, cols[si]))
}
}
inserts = append(inserts, ivals)
i++
}
for si, sv := range pk {
b, ok := sv.([]byte)
if ok && strings.HasPrefix(coltypes[pkcols[si]], "STRING") {
// Primary key strings need to be converted to a go string, but not SQL
// encoded since they aren't being written to a text file.
pk[si] = string(b)
}
}
if err := rows.Close(); err != nil {
return err
}
if i == 0 {
break
}
fmt.Fprintf(w, "\nINSERT INTO %s VALUES", tablename)
for idx, values := range inserts {
if idx > 0 {
fmt.Fprint(w, ",")
}
fmt.Fprint(w, "\n\t(")
for vi, v := range values {
if vi > 0 {
fmt.Fprint(w, ", ")
}
fmt.Fprint(w, v)
}
fmt.Fprint(w, ")")
}
fmt.Fprintln(w, ";")
if i < limit {
break
}
}
return nil
}
func TestEncDatumRowCompare(t *testing.T) {
defer leaktest.AfterTest(t)()
v := [5]EncDatum{}
for i := range v {
v[i] = DatumToEncDatum(ColumnType{Kind: ColumnType_INT}, parser.NewDInt(parser.DInt(i)))
}
asc := encoding.Ascending
desc := encoding.Descending
testCases := []struct {
row1, row2 EncDatumRow
ord ColumnOrdering
cmp int
}{
{
row1: EncDatumRow{v[0], v[1], v[2]},
row2: EncDatumRow{v[0], v[1], v[3]},
ord: ColumnOrdering{},
cmp: 0,
},
{
row1: EncDatumRow{v[0], v[1], v[2]},
row2: EncDatumRow{v[0], v[1], v[3]},
ord: ColumnOrdering{{1, desc}},
cmp: 0,
},
{
row1: EncDatumRow{v[0], v[1], v[2]},
row2: EncDatumRow{v[0], v[1], v[3]},
ord: ColumnOrdering{{0, asc}, {1, desc}},
cmp: 0,
},
{
row1: EncDatumRow{v[0], v[1], v[2]},
row2: EncDatumRow{v[0], v[1], v[3]},
ord: ColumnOrdering{{2, asc}},
cmp: -1,
},
{
row1: EncDatumRow{v[0], v[1], v[3]},
row2: EncDatumRow{v[0], v[1], v[2]},
ord: ColumnOrdering{{2, asc}},
cmp: 1,
},
{
row1: EncDatumRow{v[0], v[1], v[2]},
row2: EncDatumRow{v[0], v[1], v[3]},
ord: ColumnOrdering{{2, asc}, {0, asc}, {1, asc}},
cmp: -1,
},
{
row1: EncDatumRow{v[0], v[1], v[2]},
row2: EncDatumRow{v[0], v[1], v[3]},
ord: ColumnOrdering{{0, asc}, {2, desc}},
cmp: 1,
},
{
row1: EncDatumRow{v[0], v[1], v[2]},
row2: EncDatumRow{v[0], v[1], v[3]},
ord: ColumnOrdering{{1, desc}, {0, asc}, {2, desc}},
cmp: 1,
},
{
row1: EncDatumRow{v[2], v[3]},
row2: EncDatumRow{v[1], v[3], v[0]},
ord: ColumnOrdering{{0, asc}},
cmp: 1,
},
{
row1: EncDatumRow{v[2], v[3]},
row2: EncDatumRow{v[1], v[3], v[0]},
ord: ColumnOrdering{{1, desc}, {0, asc}},
cmp: 1,
},
{
row1: EncDatumRow{v[2], v[3]},
row2: EncDatumRow{v[1], v[3], v[0]},
ord: ColumnOrdering{{1, asc}, {0, asc}},
cmp: 1,
},
{
row1: EncDatumRow{v[2], v[3]},
row2: EncDatumRow{v[1], v[3], v[0]},
ord: ColumnOrdering{{1, asc}, {0, desc}},
cmp: -1,
},
{
row1: EncDatumRow{v[2], v[3]},
row2: EncDatumRow{v[1], v[3], v[0]},
ord: ColumnOrdering{{0, desc}, {1, asc}},
cmp: -1,
},
}
a := &DatumAlloc{}
for _, c := range testCases {
cmp, err := c.row1.Compare(a, c.ord, c.row2)
if err != nil {
t.Error(err)
} else if cmp != c.cmp {
t.Errorf("%s cmp %s ordering %v got %d, expected %d",
c.row1, c.row2, c.ord, cmp, c.cmp)
}
}
}
func TestJoinReader(t *testing.T) {
defer leaktest.AfterTest(t)()
s, sqlDB, kvDB := serverutils.StartServer(t, base.TestServerArgs{})
defer s.Stopper().Stop()
// Create a table where each row is:
//
// | a | b | sum | s |
// |-----------------------------------------------------------------|
// | rowId/10 | rowId%10 | rowId/10 + rowId%10 | IntToEnglish(rowId) |
aFn := func(row int) parser.Datum {
return parser.NewDInt(parser.DInt(row / 10))
}
bFn := func(row int) parser.Datum {
return parser.NewDInt(parser.DInt(row % 10))
}
sumFn := func(row int) parser.Datum {
return parser.NewDInt(parser.DInt(row/10 + row%10))
}
sqlutils.CreateTable(t, sqlDB, "t",
"a INT, b INT, sum INT, s STRING, PRIMARY KEY (a,b), INDEX bs (b,s)",
99,
sqlutils.ToRowFn(aFn, bFn, sumFn, sqlutils.RowEnglishFn))
td := sqlbase.GetTableDescriptor(kvDB, "test", "t")
testCases := []struct {
spec JoinReaderSpec
input [][]parser.Datum
expected string
}{
{
spec: JoinReaderSpec{
OutputColumns: []uint32{0, 1, 2},
},
input: [][]parser.Datum{
{aFn(2), bFn(2)},
{aFn(5), bFn(5)},
{aFn(10), bFn(10)},
{aFn(15), bFn(15)},
},
expected: "[[0 2 2] [0 5 5] [1 0 1] [1 5 6]]",
},
{
spec: JoinReaderSpec{
Filter: Expression{Expr: "$2 <= 5"}, // sum <= 5
OutputColumns: []uint32{3},
},
input: [][]parser.Datum{
{aFn(1), bFn(1)},
{aFn(25), bFn(25)},
{aFn(5), bFn(5)},
{aFn(21), bFn(21)},
{aFn(34), bFn(34)},
{aFn(13), bFn(13)},
{aFn(51), bFn(51)},
{aFn(50), bFn(50)},
},
expected: "[['one'] ['five'] ['two-one'] ['one-three'] ['five-zero']]",
},
}
for _, c := range testCases {
js := c.spec
js.Table = *td
txn := client.NewTxn(context.Background(), *kvDB)
flowCtx := FlowCtx{
Context: context.Background(),
evalCtx: &parser.EvalContext{},
txn: txn,
}
in := &RowBuffer{}
for _, row := range c.input {
encRow := make(sqlbase.EncDatumRow, len(row))
for i, d := range row {
encRow[i].SetDatum(sqlbase.ColumnType_INT, d)
}
in.rows = append(in.rows, encRow)
}
out := &RowBuffer{}
jr, err := newJoinReader(&flowCtx, &js, in, out)
if err != nil {
t.Fatal(err)
}
jr.Run(nil)
if out.err != nil {
t.Fatal(out.err)
}
if !in.done {
t.Fatal("joinReader stopped accepting rows")
}
if !out.closed {
t.Fatalf("output RowReceiver not closed")
}
if result := out.rows.String(); result != c.expected {
t.Errorf("invalid results: %s, expected %s'", result, c.expected)
}
}
}
// decodeOidDatum decodes bytes with specified Oid and format code into
// a datum.
func decodeOidDatum(id oid.Oid, code formatCode, b []byte) (parser.Datum, error) {
var d parser.Datum
switch id {
case oid.T_bool:
switch code {
case formatText:
v, err := strconv.ParseBool(string(b))
if err != nil {
return d, err
}
d = parser.MakeDBool(parser.DBool(v))
case formatBinary:
switch b[0] {
case 0:
d = parser.MakeDBool(false)
case 1:
d = parser.MakeDBool(true)
default:
return d, errors.Errorf("unsupported binary bool: %q", b)
}
default:
return d, errors.Errorf("unsupported bool format code: %d", code)
}
case oid.T_int2:
switch code {
case formatText:
i, err := strconv.ParseInt(string(b), 10, 64)
if err != nil {
return d, err
}
d = parser.NewDInt(parser.DInt(i))
case formatBinary:
if len(b) < 2 {
return d, errors.Errorf("int2 requires 2 bytes for binary format")
}
i := int16(binary.BigEndian.Uint16(b))
d = parser.NewDInt(parser.DInt(i))
default:
return d, errors.Errorf("unsupported int2 format code: %d", code)
}
case oid.T_int4:
switch code {
case formatText:
i, err := strconv.ParseInt(string(b), 10, 64)
if err != nil {
return d, err
}
d = parser.NewDInt(parser.DInt(i))
case formatBinary:
if len(b) < 4 {
return d, errors.Errorf("int4 requires 4 bytes for binary format")
}
i := int32(binary.BigEndian.Uint32(b))
d = parser.NewDInt(parser.DInt(i))
default:
return d, errors.Errorf("unsupported int4 format code: %d", code)
}
case oid.T_int8:
switch code {
case formatText:
i, err := strconv.ParseInt(string(b), 10, 64)
if err != nil {
return d, err
}
d = parser.NewDInt(parser.DInt(i))
case formatBinary:
if len(b) < 8 {
return d, errors.Errorf("int8 requires 8 bytes for binary format")
}
i := int64(binary.BigEndian.Uint64(b))
d = parser.NewDInt(parser.DInt(i))
default:
return d, errors.Errorf("unsupported int8 format code: %d", code)
}
case oid.T_float4:
switch code {
case formatText:
f, err := strconv.ParseFloat(string(b), 64)
if err != nil {
return d, err
}
d = parser.NewDFloat(parser.DFloat(f))
case formatBinary:
if len(b) < 4 {
return d, errors.Errorf("float4 requires 4 bytes for binary format")
}
f := math.Float32frombits(binary.BigEndian.Uint32(b))
d = parser.NewDFloat(parser.DFloat(f))
default:
return d, errors.Errorf("unsupported float4 format code: %d", code)
}
case oid.T_float8:
switch code {
case formatText:
f, err := strconv.ParseFloat(string(b), 64)
if err != nil {
return d, err
}
d = parser.NewDFloat(parser.DFloat(f))
case formatBinary:
if len(b) < 8 {
return d, errors.Errorf("float8 requires 8 bytes for binary format")
}
f := math.Float64frombits(binary.BigEndian.Uint64(b))
d = parser.NewDFloat(parser.DFloat(f))
default:
return d, errors.Errorf("unsupported float8 format code: %d", code)
}
case oid.T_numeric:
switch code {
case formatText:
dd := &parser.DDecimal{}
if _, ok := dd.SetString(string(b)); !ok {
return nil, errors.Errorf("could not parse string %q as decimal", b)
}
d = dd
case formatBinary:
r := bytes.NewReader(b)
alloc := struct {
pgNum pgNumeric
i16 int16
dd parser.DDecimal
}{}
for _, ptr := range []interface{}{
&alloc.pgNum.ndigits,
&alloc.pgNum.weight,
&alloc.pgNum.sign,
&alloc.pgNum.dscale,
} {
if err := binary.Read(r, binary.BigEndian, ptr); err != nil {
return d, err
}
}
if alloc.pgNum.ndigits > 0 {
decDigits := make([]byte, 0, alloc.pgNum.ndigits*pgDecDigits)
nextDigit := func() error {
if err := binary.Read(r, binary.BigEndian, &alloc.i16); err != nil {
return err
}
numZeroes := pgDecDigits
for i16 := alloc.i16; i16 > 0; i16 /= 10 {
numZeroes--
}
for ; numZeroes > 0; numZeroes-- {
decDigits = append(decDigits, '0')
}
return nil
}
for i := int16(0); i < alloc.pgNum.ndigits-1; i++ {
if err := nextDigit(); err != nil {
return d, err
}
if alloc.i16 > 0 {
decDigits = strconv.AppendUint(decDigits, uint64(alloc.i16), 10)
}
}
// The last digit may contain padding, which we need to deal with.
if err := nextDigit(); err != nil {
return d, err
}
dscale := (alloc.pgNum.ndigits - (alloc.pgNum.weight + 1)) * pgDecDigits
if overScale := dscale - alloc.pgNum.dscale; overScale > 0 {
dscale -= overScale
for i := int16(0); i < overScale; i++ {
alloc.i16 /= 10
}
}
decDigits = strconv.AppendUint(decDigits, uint64(alloc.i16), 10)
decString := string(decDigits)
if _, ok := alloc.dd.UnscaledBig().SetString(decString, 10); !ok {
return nil, errors.Errorf("could not parse string %q as decimal", decString)
}
alloc.dd.SetScale(inf.Scale(dscale))
}
switch alloc.pgNum.sign {
case pgNumericPos:
case pgNumericNeg:
alloc.dd.Neg(&alloc.dd.Dec)
default:
return d, errors.Errorf("unsupported numeric sign: %d", alloc.pgNum.sign)
}
d = &alloc.dd
default:
return d, errors.Errorf("unsupported numeric format code: %d", code)
}
case oid.T_text, oid.T_varchar:
switch code {
case formatText, formatBinary:
d = parser.NewDString(string(b))
default:
return d, errors.Errorf("unsupported text format code: %d", code)
}
case oid.T_bytea:
switch code {
case formatText:
// http://www.postgresql.org/docs/current/static/datatype-binary.html#AEN5667
// Code cribbed from github.com/lib/pq.
// We only support hex encoding.
if len(b) >= 2 && bytes.Equal(b[:2], []byte("\\x")) {
b = b[2:] // trim off leading "\\x"
result := make([]byte, hex.DecodedLen(len(b)))
_, err := hex.Decode(result, b)
if err != nil {
return d, err
}
d = parser.NewDBytes(parser.DBytes(result))
} else {
return d, errors.Errorf("unsupported bytea encoding: %q", b)
}
case formatBinary:
d = parser.NewDBytes(parser.DBytes(b))
default:
return d, errors.Errorf("unsupported bytea format code: %d", code)
}
case oid.T_timestamp:
switch code {
case formatText:
ts, err := parseTs(string(b))
if err != nil {
return d, errors.Errorf("could not parse string %q as timestamp", b)
}
d = parser.MakeDTimestamp(ts, time.Microsecond)
case formatBinary:
if len(b) < 8 {
return d, errors.Errorf("timestamp requires 8 bytes for binary format")
}
i := int64(binary.BigEndian.Uint64(b))
d = parser.MakeDTimestamp(pgBinaryToTime(i), time.Microsecond)
default:
return d, errors.Errorf("unsupported timestamp format code: %d", code)
}
case oid.T_timestamptz:
switch code {
case formatText:
ts, err := parseTs(string(b))
if err != nil {
return d, errors.Errorf("could not parse string %q as timestamp", b)
}
d = parser.MakeDTimestampTZ(ts, time.Microsecond)
case formatBinary:
if len(b) < 8 {
return d, errors.Errorf("timestamptz requires 8 bytes for binary format")
}
i := int64(binary.BigEndian.Uint64(b))
d = parser.MakeDTimestampTZ(pgBinaryToTime(i), time.Microsecond)
default:
return d, errors.Errorf("unsupported timestamptz format code: %d", code)
}
case oid.T_date:
switch code {
case formatText:
ts, err := parseTs(string(b))
if err != nil {
res, err := parser.ParseDDate(string(b), time.UTC)
if err != nil {
return d, errors.Errorf("could not parse string %q as date", b)
}
d = res
} else {
daysSinceEpoch := ts.Unix() / secondsInDay
d = parser.NewDDate(parser.DDate(daysSinceEpoch))
}
case formatBinary:
if len(b) < 4 {
return d, errors.Errorf("date requires 4 bytes for binary format")
}
i := int32(binary.BigEndian.Uint32(b))
d = pgBinaryToDate(i)
default:
return d, errors.Errorf("unsupported date format code: %d", code)
}
case oid.T_interval:
switch code {
case formatText:
d, err := parser.ParseDInterval(string(b))
if err != nil {
return d, errors.Errorf("could not parse string %q as interval", b)
}
return d, nil
default:
return d, errors.Errorf("unsupported interval format code: %d", code)
}
default:
return d, errors.Errorf("unsupported OID: %v", id)
}
return d, nil
}
// UnmarshalColumnValue decodes the value from a key-value pair using the type
// expected by the column. An error is returned if the value's type does not
// match the column's type.
func UnmarshalColumnValue(
a *DatumAlloc, kind ColumnType_Kind, value *roachpb.Value,
) (parser.Datum, error) {
if value == nil {
return parser.DNull, nil
}
switch kind {
case ColumnType_BOOL:
v, err := value.GetBool()
if err != nil {
return nil, err
}
return parser.MakeDBool(parser.DBool(v)), nil
case ColumnType_INT:
v, err := value.GetInt()
if err != nil {
return nil, err
}
return a.NewDInt(parser.DInt(v)), nil
case ColumnType_FLOAT:
v, err := value.GetFloat()
if err != nil {
return nil, err
}
return a.NewDFloat(parser.DFloat(v)), nil
case ColumnType_DECIMAL:
v, err := value.GetDecimal()
if err != nil {
return nil, err
}
dd := a.NewDDecimal(parser.DDecimal{})
dd.Set(v)
return dd, nil
case ColumnType_STRING:
v, err := value.GetBytes()
if err != nil {
return nil, err
}
return a.NewDString(parser.DString(v)), nil
case ColumnType_BYTES:
v, err := value.GetBytes()
if err != nil {
return nil, err
}
return a.NewDBytes(parser.DBytes(v)), nil
case ColumnType_DATE:
v, err := value.GetInt()
if err != nil {
return nil, err
}
return a.NewDDate(parser.DDate(v)), nil
case ColumnType_TIMESTAMP:
v, err := value.GetTime()
if err != nil {
return nil, err
}
return a.NewDTimestamp(parser.DTimestamp{Time: v}), nil
case ColumnType_TIMESTAMPTZ:
v, err := value.GetTime()
if err != nil {
return nil, err
}
return a.NewDTimestampTZ(parser.DTimestampTZ{Time: v}), nil
case ColumnType_INTERVAL:
d, err := value.GetDuration()
if err != nil {
return nil, err
}
return a.NewDInterval(parser.DInterval{Duration: d}), nil
default:
return nil, errors.Errorf("unsupported column type: %s", kind)
}
}
// ShowIndex returns all the indexes for a table.
// Privileges: Any privilege on table.
// Notes: postgres does not have a SHOW INDEXES statement.
// mysql requires some privilege for any column.
func (p *planner) ShowIndex(n *parser.ShowIndex) (planNode, error) {
tn, err := n.Table.NormalizeWithDatabaseName(p.session.Database)
if err != nil {
return nil, err
}
desc, err := p.mustGetTableDesc(tn)
if err != nil {
return nil, err
}
if err := p.anyPrivilege(desc); err != nil {
return nil, err
}
columns := ResultColumns{
{Name: "Table", Typ: parser.TypeString},
{Name: "Name", Typ: parser.TypeString},
{Name: "Unique", Typ: parser.TypeBool},
{Name: "Seq", Typ: parser.TypeInt},
{Name: "Column", Typ: parser.TypeString},
{Name: "Direction", Typ: parser.TypeString},
{Name: "Storing", Typ: parser.TypeBool},
}
return &delayedNode{
p: p,
name: "SHOW INDEX FROM " + tn.String(),
columns: columns,
constructor: func(p *planner) (planNode, error) {
v := p.newContainerValuesNode(columns, 0)
appendRow := func(index sqlbase.IndexDescriptor, colName string, sequence int,
direction string, isStored bool) error {
newRow := parser.DTuple{
parser.NewDString(tn.Table()),
parser.NewDString(index.Name),
parser.MakeDBool(parser.DBool(index.Unique)),
parser.NewDInt(parser.DInt(sequence)),
parser.NewDString(colName),
parser.NewDString(direction),
parser.MakeDBool(parser.DBool(isStored)),
}
_, err := v.rows.AddRow(newRow)
return err
}
for _, index := range append([]sqlbase.IndexDescriptor{desc.PrimaryIndex}, desc.Indexes...) {
sequence := 1
for i, col := range index.ColumnNames {
if err := appendRow(index, col, sequence, index.ColumnDirections[i].String(), false); err != nil {
v.rows.Close()
return nil, err
}
sequence++
}
for _, col := range index.StoreColumnNames {
if err := appendRow(index, col, sequence, "N/A", true); err != nil {
v.rows.Close()
return nil, err
}
sequence++
}
}
return v, nil
},
}, nil
}
// RowIdxFn is a GenValueFn that returns the row number as a DInt
func RowIdxFn(row int) parser.Datum {
return parser.NewDInt(parser.DInt(row))
}