Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

executor: load data/batch insert improvement reducing memory allocation/free #11284

Merged
merged 2 commits into from
Jul 23, 2019
Merged

executor: load data/batch insert improvement reducing memory allocation/free #11284

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
3ea1f68
optimize load data/batch insert memory usage, alloc/free times, reduc…
cfzjywxk Jul 16, 2019
67a3b77
Merge branch 'master' into p4348
cfzjywxk Jul 23, 2019
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
3 changes: 3 additions & 0 deletions executor/builder.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -681,6 +681,9 @@ func (b *executorBuilder) buildLoadData(v *plannercore.LoadData) Executor {
},
}

var defaultLoadDataBatchCnt uint64 = 20000 // TODO this will be changed to variable in another pr
loadDataExec.loadDataInfo.SetMaxRowsInBatch(defaultLoadDataBatchCnt)

return loadDataExec
}

Expand Down
21 changes: 18 additions & 3 deletions executor/insert_common.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -35,6 +35,7 @@ type InsertValues struct {
batchChecker

rowCount uint64
curBatchCnt uint64
maxRowsInBatch uint64
lastInsertID uint64
hasRefCols bool
Expand Down Expand Up @@ -378,6 +379,20 @@ func (e *InsertValues) getRow(ctx context.Context, vals []types.Datum) ([]types.
return e.fillRow(ctx, row, hasValue)
}

func (e *InsertValues) getRowInPlace(ctx context.Context, vals []types.Datum, rowBuf []types.Datum) ([]types.Datum, error) {
hasValue := make([]bool, len(e.Table.Cols()))
for i, v := range vals {
casted, err := table.CastValue(e.ctx, v, e.insertColumns[i].ToInfo())
if e.filterErr(err) != nil {
return nil, err
}
offset := e.insertColumns[i].Offset
rowBuf[offset] = casted
hasValue[offset] = true
}
return e.fillRow(ctx, rowBuf, hasValue)
}

func (e *InsertValues) filterErr(err error) error {
if err == nil {
return nil
Expand Down Expand Up @@ -547,25 +562,25 @@ func (e *InsertValues) batchCheckAndInsert(rows [][]types.Datum, addRecord func(
}
// append warnings and get no duplicated error rows
for i, r := range e.toBeCheckedRows {
skip := false
if r.handleKey != nil {
if _, found := e.dupKVs[string(r.handleKey.newKV.key)]; found {
rows[i] = nil
e.ctx.GetSessionVars().StmtCtx.AppendWarning(r.handleKey.dupErr)
continue
}
}
for _, uk := range r.uniqueKeys {
if _, found := e.dupKVs[string(uk.newKV.key)]; found {
// If duplicate keys were found in BatchGet, mark row = nil.
rows[i] = nil
e.ctx.GetSessionVars().StmtCtx.AppendWarning(uk.dupErr)
skip = true
break
}
}
// If row was checked with no duplicate keys,
// it should be add to values map for the further row check.
// There may be duplicate keys inside the insert statement.
if rows[i] != nil {
if !skip {
e.ctx.GetSessionVars().StmtCtx.AddCopiedRows(1)
_, err = addRecord(rows[i])
if err != nil {
Expand Down
23 changes: 15 additions & 8 deletions executor/load_data.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,7 @@ import (
"github.com/pingcap/tidb/table"
"github.com/pingcap/tidb/types"
"github.com/pingcap/tidb/util/chunk"
"github.com/pingcap/tidb/util/hack"
"github.com/pingcap/tidb/util/logutil"
"github.com/pingcap/tidb/util/stringutil"
"go.uber.org/zap"
Expand Down Expand Up @@ -112,7 +113,12 @@ type LoadDataInfo struct {
// SetMaxRowsInBatch sets the max number of rows to insert in a batch.
func (e *LoadDataInfo) SetMaxRowsInBatch(limit uint64) {
e.maxRowsInBatch = limit
e.rows = make([][]types.Datum, 0, limit)
if uint64(cap(e.rows)) < limit {
e.rows = make([][]types.Datum, 0, limit)
for i := 0; uint64(i) < limit; i++ {
e.rows = append(e.rows, make([]types.Datum, len(e.Table.Cols())))
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

maybe we only need new limit - cap(e.rows) types.Datum here if the limit can be dynamical change

and the other optimize maybe or not work is make([]types.Datum, (limit -cap(e.rows)) * len(e.Table.Cols())) once and then cut slice from this big array to append

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

seems trace not support load data now ?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@lysu currently init will be called only once. in another pr making variable settable will do this

}
}
}

// getValidData returns prevData and curData that starts from starting symbol.
Expand Down Expand Up @@ -174,7 +180,7 @@ func (e *LoadDataInfo) getLine(prevData, curData []byte) ([]byte, []byte, bool)
}
endIdx := -1
if len(curData) >= curStartIdx {
endIdx = strings.Index(string(curData[curStartIdx:]), e.LinesInfo.Terminated)
endIdx = strings.Index(string(hack.String(curData[curStartIdx:])), e.LinesInfo.Terminated)
}
if endIdx == -1 {
// no terminated symbol
Expand Down Expand Up @@ -253,8 +259,9 @@ func (e *LoadDataInfo) InsertData(ctx context.Context, prevData, curData []byte)
if err != nil {
return nil, false, err
}
e.rows = append(e.rows, e.colsToRow(ctx, cols))
e.colsToRow(ctx, cols)
e.rowCount++
e.curBatchCnt++
if e.maxRowsInBatch != 0 && e.rowCount%e.maxRowsInBatch == 0 {
reachLimit = true
logutil.BgLogger().Info("batch limit hit when inserting rows", zap.Int("maxBatchRows", e.maxChunkSize),
Expand All @@ -268,15 +275,15 @@ func (e *LoadDataInfo) InsertData(ctx context.Context, prevData, curData []byte)
// CheckAndInsertOneBatch is used to commit one transaction batch full filled data
func (e *LoadDataInfo) CheckAndInsertOneBatch() error {
var err error
if len(e.rows) == 0 {
if e.curBatchCnt == 0 {
return err
}
e.ctx.GetSessionVars().StmtCtx.AddRecordRows(uint64(len(e.rows)))
err = e.batchCheckAndInsert(e.rows, e.addRecordLD)
e.ctx.GetSessionVars().StmtCtx.AddRecordRows(e.curBatchCnt)
err = e.batchCheckAndInsert(e.rows[0:e.curBatchCnt], e.addRecordLD)
if err != nil {
return err
}
e.rows = e.rows[:0]
e.curBatchCnt = 0
return err
}

Expand Down Expand Up @@ -312,7 +319,7 @@ func (e *LoadDataInfo) colsToRow(ctx context.Context, cols []field) []types.Datu
e.row[i].SetString(string(cols[i].str))
}
}
row, err := e.getRow(ctx, e.row)
row, err := e.getRowInPlace(ctx, e.row, e.rows[e.curBatchCnt])
if err != nil {
e.handleWarning(err)
return nil
Expand Down