config, changefeed (ticdc): add sql mode config for changefeed (#9941)

close #9876

cdc/api/v2/model.go

@@ -193,7 +193,8 @@ type ReplicaConfig struct {
 	Consistent                   *ConsistentConfig          `json:"consistent,omitempty"`
 	Scheduler                    *ChangefeedSchedulerConfig `json:"scheduler"`
 	Integrity                    *IntegrityConfig           `json:"integrity"`
-	ChangefeedErrorStuckDuration *JSONDuration              `json:"changefeed_error_stuck_duration,omitempty" swaggertype:"string"`
+	ChangefeedErrorStuckDuration *JSONDuration              `json:"changefeed_error_stuck_duration,omitempty"`
+	SQLMode                      string                     `json:"sql_mode,omitempty"`
 }
 
 // ToInternalReplicaConfig coverts *v2.ReplicaConfig into *config.ReplicaConfig
@@ -211,6 +212,7 @@ func (c *ReplicaConfig) toInternalReplicaConfigWithOriginConfig(
 	res.CheckGCSafePoint = c.CheckGCSafePoint
 	res.EnableSyncPoint = c.EnableSyncPoint
 	res.IgnoreIneligibleTable = c.IgnoreIneligibleTable
+	res.SQLMode = c.SQLMode
 	if c.SyncPointInterval != nil {
 		res.SyncPointInterval = &c.SyncPointInterval.duration
 	}
@@ -498,6 +500,7 @@ func ToAPIReplicaConfig(c *config.ReplicaConfig) *ReplicaConfig {
 		CheckGCSafePoint:      cloned.CheckGCSafePoint,
 		EnableSyncPoint:       cloned.EnableSyncPoint,
 		BDRMode:               cloned.BDRMode,
+		SQLMode:               cloned.SQLMode,
 	}
 
 	if cloned.SyncPointInterval != nil {

cdc/api/v2/model_test.go

@@ -81,6 +81,7 @@ var defaultAPIConfig = &ReplicaConfig{
 	},
 	ChangefeedErrorStuckDuration: &JSONDuration{*config.
 		GetDefaultReplicaConfig().ChangefeedErrorStuckDuration},
+	SQLMode: config.GetDefaultReplicaConfig().SQLMode,
 }
 
 func TestDefaultReplicaConfig(t *testing.T) {

cdc/model/changefeed.go

@@ -334,6 +334,9 @@ func (info *ChangeFeedInfo) VerifyAndComplete() {
 	if info.Config.ChangefeedErrorStuckDuration == nil {
 		info.Config.ChangefeedErrorStuckDuration = defaultConfig.ChangefeedErrorStuckDuration
 	}
+	if info.Config.SQLMode == "" {
+		info.Config.SQLMode = defaultConfig.SQLMode
+	}
 	info.RmUnusedFields()
 }
 

cdc/owner/ddl_sink.go

@@ -25,6 +25,7 @@ import (
 	"github.com/pingcap/log"
 	"github.com/pingcap/tidb/parser"
 	"github.com/pingcap/tidb/parser/format"
+	"github.com/pingcap/tidb/parser/mysql"
 	"github.com/pingcap/tiflow/cdc/model"
 	"github.com/pingcap/tiflow/cdc/sink/ddlsink"
 	"github.com/pingcap/tiflow/cdc/sink/ddlsink/factory"
@@ -426,7 +427,18 @@ func (s *ddlSinkImpl) close(ctx context.Context) (err error) {
 
 // addSpecialComment translate tidb feature to comment
 func (s *ddlSinkImpl) addSpecialComment(ddl *model.DDLEvent) (string, error) {
-	stms, _, err := parser.New().Parse(ddl.Query, ddl.Charset, ddl.Collate)
+	p := parser.New()
+	// We need to use the correct SQL mode to parse the DDL query.
+	// Otherwise, the parser may fail to parse the DDL query.
+	// For example, it is needed to parse the following DDL query:
+	//  `alter table "t" add column "c" int default 1;`
+	// by adding `ANSI_QUOTES` to the SQL mode.
+	mode, err := mysql.GetSQLMode(s.info.Config.SQLMode)
+	if err != nil {
+		return "", errors.Trace(err)
+	}
+	p.SetSQLMode(mode)
+	stms, _, err := p.Parse(ddl.Query, ddl.Charset, ddl.Collate)
 	if err != nil {
 		return "", errors.Trace(err)
 	}

cdc/owner/ddl_sink_test.go

@@ -23,6 +23,7 @@ import (
 	"github.com/pingcap/errors"
 	"github.com/pingcap/tiflow/cdc/model"
 	"github.com/pingcap/tiflow/cdc/sink/ddlsink"
+	"github.com/pingcap/tiflow/pkg/config"
 	cerror "github.com/pingcap/tiflow/pkg/errors"
 	"github.com/pingcap/tiflow/pkg/retry"
 	"github.com/stretchr/testify/require"
@@ -61,7 +62,13 @@ func (m *mockSink) GetDDL() *model.DDLEvent {
 
 func newDDLSink4Test(reportErr func(err error), reportWarn func(err error)) (DDLSink, *mockSink) {
 	mockSink := &mockSink{}
-	ddlSink := newDDLSink(model.DefaultChangeFeedID("changefeed-test"), &model.ChangeFeedInfo{}, reportErr, reportWarn)
+	ddlSink := newDDLSink(
+		model.DefaultChangeFeedID("changefeed-test"),
+		&model.ChangeFeedInfo{
+			Config: config.GetDefaultReplicaConfig(),
+		},
+		reportErr,
+		reportWarn)
 	ddlSink.(*ddlSinkImpl).sinkInitHandler = func(ctx context.Context, s *ddlSinkImpl) error {
 		s.sink = mockSink
 		return nil
@@ -545,6 +552,9 @@ func TestAddSpecialComment(t *testing.T) {
 	}
 
 	s := &ddlSinkImpl{}
+	s.info = &model.ChangeFeedInfo{
+		Config: config.GetDefaultReplicaConfig(),
+	}
 	for _, ca := range testCase {
 		re, err := s.addSpecialComment(ca.event)
 		require.Nil(t, err)

pkg/config/config_test_data.go

@@ -82,7 +82,8 @@ const (
     "integrity-check-level": "none",
     "corruption-handle-level": "warn"
  },
-  "changefeed-error-stuck-duration": 1800000000000
+  "changefeed-error-stuck-duration": 1800000000000,
+  "sql-mode":"ONLY_FULL_GROUP_BY,STRICT_TRANS_TABLES,NO_ZERO_IN_DATE,NO_ZERO_DATE,ERROR_FOR_DIVISION_BY_ZERO,NO_AUTO_CREATE_USER,NO_ENGINE_SUBSTITUTION"
 }`
 
 	testCfgTestServerConfigMarshal = `{
@@ -319,7 +320,8 @@ const (
     "integrity-check-level": "none",
     "corruption-handle-level": "warn"
   },
-  "changefeed-error-stuck-duration": 1800000000000
+  "changefeed-error-stuck-duration": 1800000000000,
+  "sql-mode":"ONLY_FULL_GROUP_BY,STRICT_TRANS_TABLES,NO_ZERO_IN_DATE,NO_ZERO_DATE,ERROR_FOR_DIVISION_BY_ZERO,NO_AUTO_CREATE_USER,NO_ENGINE_SUBSTITUTION"
 }`
 
 	testCfgTestReplicaConfigMarshal2 = `{
@@ -465,6 +467,7 @@ const (
     "integrity-check-level": "none",
     "corruption-handle-level": "warn"
   },
-  "changefeed-error-stuck-duration": 1800000000000
+  "changefeed-error-stuck-duration": 1800000000000,
+  "sql-mode":"ONLY_FULL_GROUP_BY,STRICT_TRANS_TABLES,NO_ZERO_IN_DATE,NO_ZERO_DATE,ERROR_FOR_DIVISION_BY_ZERO,NO_AUTO_CREATE_USER,NO_ENGINE_SUBSTITUTION"
 }`
 )

pkg/config/replica_config.go

@@ -23,6 +23,7 @@ import (
 
 	"github.com/pingcap/errors"
 	"github.com/pingcap/log"
+	"github.com/pingcap/tidb/parser/mysql"
 	"github.com/pingcap/tiflow/pkg/config/outdated"
 	cerror "github.com/pingcap/tiflow/pkg/errors"
 	"github.com/pingcap/tiflow/pkg/integrity"
@@ -38,6 +39,12 @@ const (
 	// minSyncPointRetention is the minimum of SyncPointRetention can be set.
 	minSyncPointRetention           = time.Hour * 1
 	minChangeFeedErrorStuckDuration = time.Minute * 30
+	// The default SQL Mode of TiDB: "ONLY_FULL_GROUP_BY,
+	// STRICT_TRANS_TABLES,NO_ZERO_IN_DATE,NO_ZERO_DATE,ERROR_FOR_DIVISION_BY_ZERO,
+	// NO_AUTO_CREATE_USER,NO_ENGINE_SUBSTITUTION"
+	// Note: The SQL Mode of TiDB is not the same as ORACLE.
+	// If you want to use the same SQL Mode as ORACLE, you need to add "ORACLE" to the SQL Mode.
+	defaultSQLMode = mysql.DefaultSQLMode
 )
 
 var defaultReplicaConfig = &ReplicaConfig{
@@ -89,6 +96,7 @@ var defaultReplicaConfig = &ReplicaConfig{
 		CorruptionHandleLevel: integrity.CorruptionHandleLevelWarn,
 	},
 	ChangefeedErrorStuckDuration: util.AddressOf(time.Minute * 30),
+	SQLMode:                      defaultSQLMode,
 }
 
 // GetDefaultReplicaConfig returns the default replica config.
@@ -140,6 +148,7 @@ type replicaConfig struct {
 	// Integrity is only available when the downstream is MQ.
 	Integrity                    *integrity.Config `toml:"integrity" json:"integrity"`
 	ChangefeedErrorStuckDuration *time.Duration    `toml:"changefeed-error-stuck-duration" json:"changefeed-error-stuck-duration,omitempty"`
+	SQLMode                      string            `toml:"sql-mode" json:"sql-mode"`
 }
 
 // Value implements the driver.Valuer interface

pkg/filter/expr_filter.go

@@ -14,13 +14,15 @@
 package filter
 
 import (
+	"fmt"
 	"strings"
 	"sync"
 
 	"github.com/pingcap/errors"
 	"github.com/pingcap/log"
 	"github.com/pingcap/tidb/expression"
 	"github.com/pingcap/tidb/parser"
+	"github.com/pingcap/tidb/parser/mysql"
 	"github.com/pingcap/tidb/planner/core"
 	"github.com/pingcap/tidb/sessionctx"
 	"github.com/pingcap/tidb/types"
@@ -77,10 +79,16 @@ func newExprFilterRule(
 
 // verifyAndInitRule will verify and init the rule.
 // It should only be called in dmlExprFilter's verify method.
-func (r *dmlExprFilterRule) verify(tableInfos []*model.TableInfo) error {
+func (r *dmlExprFilterRule) verify(tableInfos []*model.TableInfo, sqlMode string) error {
 	// verify expression filter rule syntax.
 	p := parser.New()
-	_, _, err := p.ParseSQL(completeExpression(r.config.IgnoreInsertValueExpr))
+	mode, err := mysql.GetSQLMode(sqlMode)
+	if err != nil {
+		log.Error("failed to get sql mode", zap.Error(err))
+		return cerror.ErrInvalidReplicaConfig.FastGenByArgs(fmt.Sprintf("invalid sqlMode %s", sqlMode))
+	}
+	p.SetSQLMode(mode)
+	_, _, err = p.ParseSQL(completeExpression(r.config.IgnoreInsertValueExpr))
 	if err != nil {
 		log.Error("failed to parse expression", zap.Error(err))
 		return cerror.ErrExpressionParseFailed.
@@ -347,14 +355,18 @@ func getColumnFromError(err error) string {
 
 // dmlExprFilter is a filter that filters DML events by SQL expression.
 type dmlExprFilter struct {
-	rules []*dmlExprFilterRule
+	rules   []*dmlExprFilterRule
+	sqlMODE string
 }
 
 func newExprFilter(
 	timezone string,
 	cfg *config.FilterConfig,
+	sqlMODE string,
 ) (*dmlExprFilter, error) {
-	res := &dmlExprFilter{}
+	res := &dmlExprFilter{
+		sqlMODE: sqlMODE,
+	}
 	sessCtx := utils.NewSessionCtx(map[string]string{
 		"time_zone": timezone,
 	})
@@ -382,7 +394,7 @@ func (f *dmlExprFilter) addRule(
 // verify checks if all rules in this filter is valid.
 func (f *dmlExprFilter) verify(tableInfos []*model.TableInfo) error {
 	for _, rule := range f.rules {
-		err := rule.verify(tableInfos)
+		err := rule.verify(tableInfos, f.sqlMODE)
 		if err != nil {
 			log.Error("failed to verify expression filter rule", zap.Error(err))
 			return errors.Trace(err)

pkg/filter/expr_filter_bench_test.go

@@ -57,7 +57,7 @@ func BenchmarkSkipDML(b *testing.B) {
 	sessCtx := utils.NewSessionCtx(map[string]string{
 		"time_zone": "",
 	})
-	f, err := newExprFilter("", cfg)
+	f, err := newExprFilter("", cfg, config.GetDefaultReplicaConfig().SQLMode)
 	require.Nil(b, err)
 
 	type innerCase struct {

pkg/filter/expr_filter_test.go

@@ -324,7 +324,7 @@ func TestShouldSkipDMLBasic(t *testing.T) {
 
 	for _, tc := range testCases {
 		tableInfo := helper.execDDL(tc.ddl)
-		f, err := newExprFilter("", tc.cfg)
+		f, err := newExprFilter("", tc.cfg, config.GetDefaultReplicaConfig().SQLMode)
 		require.Nil(t, err)
 		for _, c := range tc.cases {
 			rowDatums, err := utils.AdjustBinaryProtocolForDatum(sessCtx, c.row, tableInfo.Columns)
@@ -441,7 +441,7 @@ func TestShouldSkipDMLError(t *testing.T) {
 
 	for _, tc := range testCases {
 		tableInfo := helper.execDDL(tc.ddl)
-		f, err := newExprFilter("", tc.cfg)
+		f, err := newExprFilter("", tc.cfg, config.GetDefaultReplicaConfig().SQLMode)
 		require.Nil(t, err)
 		for _, c := range tc.cases {
 			rowDatums, err := utils.AdjustBinaryProtocolForDatum(sessCtx, c.row, tableInfo.Columns)
@@ -634,7 +634,7 @@ func TestShouldSkipDMLTableUpdated(t *testing.T) {
 
 	for _, tc := range testCases {
 		tableInfo := helper.execDDL(tc.ddl)
-		f, err := newExprFilter("", tc.cfg)
+		f, err := newExprFilter("", tc.cfg, config.GetDefaultReplicaConfig().SQLMode)
 		require.Nil(t, err)
 		for _, c := range tc.cases {
 			if c.updateDDl != "" {
@@ -754,7 +754,7 @@ func TestVerify(t *testing.T) {
 			ti := helper.execDDL(ddl)
 			tableInfos = append(tableInfos, ti)
 		}
-		f, err := newExprFilter("", tc.cfg)
+		f, err := newExprFilter("", tc.cfg, config.GetDefaultReplicaConfig().SQLMode)
 		require.Nil(t, err)
 		err = f.verify(tableInfos)
 		require.True(t, errors.ErrorEqual(tc.err, err), "case: %+v", tc, err)

pkg/filter/filter.go

@@ -102,11 +102,11 @@ func NewFilter(cfg *config.ReplicaConfig, tz string) (Filter, error) {
 		f = tfilter.CaseInsensitive(f)
 	}
 
-	dmlExprFilter, err := newExprFilter(tz, cfg.Filter)
+	dmlExprFilter, err := newExprFilter(tz, cfg.Filter, cfg.SQLMode)
 	if err != nil {
 		return nil, err
 	}
-	sqlEventFilter, err := newSQLEventFilter(cfg.Filter)
+	sqlEventFilter, err := newSQLEventFilter(cfg.Filter, cfg.SQLMode)
 	if err != nil {
 		return nil, err
 	}

pkg/filter/sql_event_filter.go

@@ -14,13 +14,15 @@
 package filter
 
 import (
+	"fmt"
 	"sync"
 
 	"github.com/pingcap/errors"
 	"github.com/pingcap/log"
 	bf "github.com/pingcap/tidb-tools/pkg/binlog-filter"
 	"github.com/pingcap/tidb/parser"
 	timodel "github.com/pingcap/tidb/parser/model"
+	"github.com/pingcap/tidb/parser/mysql"
 	tfilter "github.com/pingcap/tidb/util/table-filter"
 	"github.com/pingcap/tiflow/cdc/model"
 	"github.com/pingcap/tiflow/pkg/config"
@@ -107,9 +109,17 @@ type sqlEventFilter struct {
 	rules     []*sqlEventRule
 }
 
-func newSQLEventFilter(cfg *config.FilterConfig) (*sqlEventFilter, error) {
+func newSQLEventFilter(cfg *config.FilterConfig, sqlMode string) (*sqlEventFilter, error) {
+	p := parser.New()
+	mode, err := mysql.GetSQLMode(sqlMode)
+	if err != nil {
+		log.Error("failed to get sql mode", zap.Error(err))
+		return nil, cerror.ErrInvalidReplicaConfig.FastGenByArgs(fmt.Sprintf("invalid sqlMode %s", sqlMode))
+	}
+	p.SetSQLMode(mode)
+
 	res := &sqlEventFilter{
-		ddlParser: parser.New(),
+		ddlParser: p,
 	}
 	for _, rule := range cfg.EventFilters {
 		if err := res.addRule(rule); err != nil {

pkg/filter/sql_event_filter_test.go

@@ -171,7 +171,7 @@ func TestShouldSkipDDL(t *testing.T) {
 	}
 
 	for _, tc := range testCases {
-		f, err := newSQLEventFilter(tc.cfg)
+		f, err := newSQLEventFilter(tc.cfg, config.GetDefaultReplicaConfig().SQLMode)
 		require.True(t, errors.ErrorEqual(err, tc.err), "case: %+s", err)
 		for _, c := range tc.cases {
 			ddl := &model.DDLEvent{
@@ -298,7 +298,7 @@ func TestShouldSkipDML(t *testing.T) {
 		tc := tc
 		t.Run(tc.name, func(t *testing.T) {
 			t.Parallel()
-			f, err := newSQLEventFilter(tc.cfg)
+			f, err := newSQLEventFilter(tc.cfg, config.GetDefaultReplicaConfig().SQLMode)
 			require.NoError(t, err)
 			for _, c := range tc.cases {
 				event := &model.RowChangedEvent{