milvus/internal/storage/serde_events_test.go
shaoting-huang 88b373b024
enhance: binlog primary key turn off dict encoding (#34358)
issue: #34357 

Go Parquet uses dictionary encoding by default, and it will fall back to
plain encoding if the dictionary size exceeds the dictionary size page
limit. Users can specify custom fallback encoding by using
`parquet.WithEncoding(ENCODING_METHOD)` in writer properties. However,
Go Parquet [fallbacks to plain
encoding](e65c1e295d/go/parquet/file/column_writer_types.gen.go.tmpl (L238))
rather than custom encoding method users provide. Therefore, this patch
only turns off dictionary encoding for the primary key.

With a 5 million auto ID primary key benchmark, the parquet file size
improves from 13.93 MB to 8.36 MB when dictionary encoding is turned
off, reducing primary key storage space by 40%.

Signed-off-by: shaoting-huang <shaoting.huang@zilliz.com>
2024-07-17 17:47:44 +08:00

496 lines
13 KiB
Go

// Licensed to the LF AI & Data foundation under one
// or more contributor license agreements. See the NOTICE file
// distributed with this work for additional information
// regarding copyright ownership. The ASF licenses this file
// to you under the Apache License, Version 2.0 (the
// "License"); you may not use this file except in compliance
// with the License. You may obtain a copy of the License at
//
// http://www.apache.org/licenses/LICENSE-2.0
//
// Unless required by applicable law or agreed to in writing, software
// distributed under the License is distributed on an "AS IS" BASIS,
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
// See the License for the specific language governing permissions and
// limitations under the License.
package storage
import (
"bytes"
"context"
"io"
"strconv"
"testing"
"github.com/apache/arrow/go/v12/arrow"
"github.com/apache/arrow/go/v12/arrow/array"
"github.com/apache/arrow/go/v12/arrow/memory"
"github.com/apache/arrow/go/v12/parquet/file"
"github.com/apache/arrow/go/v12/parquet/pqarrow"
"github.com/stretchr/testify/assert"
"github.com/milvus-io/milvus-proto/go-api/v2/schemapb"
"github.com/milvus-io/milvus/pkg/common"
)
func TestBinlogDeserializeReader(t *testing.T) {
t.Run("test empty data", func(t *testing.T) {
reader, err := NewBinlogDeserializeReader(nil, common.RowIDField)
assert.NoError(t, err)
defer reader.Close()
err = reader.Next()
assert.Equal(t, io.EOF, err)
})
t.Run("test deserialize", func(t *testing.T) {
size := 3
blobs, err := generateTestData(size)
assert.NoError(t, err)
reader, err := NewBinlogDeserializeReader(blobs, common.RowIDField)
assert.NoError(t, err)
defer reader.Close()
for i := 1; i <= size; i++ {
err = reader.Next()
assert.NoError(t, err)
value := reader.Value()
assertTestData(t, i, value)
}
err = reader.Next()
assert.Equal(t, io.EOF, err)
})
}
func TestBinlogStreamWriter(t *testing.T) {
t.Run("test write", func(t *testing.T) {
size := 3
field := arrow.Field{Name: "bool", Type: arrow.FixedWidthTypes.Boolean}
var w bytes.Buffer
rw, err := newSingleFieldRecordWriter(1, field, &w)
assert.NoError(t, err)
builder := array.NewBooleanBuilder(memory.DefaultAllocator)
builder.AppendValues([]bool{true, false, true}, nil)
arr := builder.NewArray()
defer arr.Release()
ar := array.NewRecord(
arrow.NewSchema(
[]arrow.Field{field},
nil,
),
[]arrow.Array{arr},
int64(size),
)
r := newSimpleArrowRecord(ar, map[FieldID]schemapb.DataType{1: schemapb.DataType_Bool}, map[FieldID]int{1: 0})
defer r.Release()
err = rw.Write(r)
assert.NoError(t, err)
rw.Close()
reader, err := file.NewParquetReader(bytes.NewReader(w.Bytes()))
assert.NoError(t, err)
arrowReader, err := pqarrow.NewFileReader(reader, pqarrow.ArrowReadProperties{BatchSize: 1024}, memory.DefaultAllocator)
assert.NoError(t, err)
rr, err := arrowReader.GetRecordReader(context.Background(), nil, nil)
assert.NoError(t, err)
defer rr.Release()
ok := rr.Next()
assert.True(t, ok)
rec := rr.Record()
defer rec.Release()
assert.Equal(t, int64(size), rec.NumRows())
ok = rr.Next()
assert.False(t, ok)
})
}
func TestBinlogSerializeWriter(t *testing.T) {
t.Run("test empty data", func(t *testing.T) {
reader, err := NewBinlogDeserializeReader(nil, common.RowIDField)
assert.NoError(t, err)
defer reader.Close()
err = reader.Next()
assert.Equal(t, io.EOF, err)
})
t.Run("test serialize", func(t *testing.T) {
size := 16
blobs, err := generateTestData(size)
assert.NoError(t, err)
reader, err := NewBinlogDeserializeReader(blobs, common.RowIDField)
assert.NoError(t, err)
defer reader.Close()
schema := generateTestSchema()
// Copy write the generated data
writers := NewBinlogStreamWriters(0, 0, 0, schema.Fields)
writer, err := NewBinlogSerializeWriter(schema, 0, 0, writers, 7)
assert.NoError(t, err)
for i := 1; i <= size; i++ {
err = reader.Next()
assert.NoError(t, err)
value := reader.Value()
assertTestData(t, i, value)
err := writer.Write(value)
assert.NoError(t, err)
}
for _, f := range schema.Fields {
props := writers[f.FieldID].rw.writerProps
assert.Equal(t, !f.IsPrimaryKey, props.DictionaryEnabled())
}
err = reader.Next()
assert.Equal(t, io.EOF, err)
err = writer.Close()
assert.NoError(t, err)
assert.True(t, writer.WrittenMemorySize() >= 429)
// Read from the written data
newblobs := make([]*Blob, len(writers))
i := 0
for _, w := range writers {
blob, err := w.Finalize()
assert.NoError(t, err)
assert.NotNil(t, blob)
assert.True(t, blob.MemorySize > 0)
newblobs[i] = blob
i++
}
// Both field pk and field 17 are with datatype string and auto id
// in test data. Field pk uses delta byte array encoding, while
// field 17 uses dict encoding.
assert.Less(t, writers[16].buf.Len(), writers[17].buf.Len())
// assert.Equal(t, blobs[0].Value, newblobs[0].Value)
reader, err = NewBinlogDeserializeReader(newblobs, common.RowIDField)
assert.NoError(t, err)
defer reader.Close()
for i := 1; i <= size; i++ {
err = reader.Next()
assert.NoError(t, err, i)
value := reader.Value()
assertTestData(t, i, value)
}
})
}
func TestNull(t *testing.T) {
t.Run("test null", func(t *testing.T) {
schema := generateTestSchema()
// Copy write the generated data
writers := NewBinlogStreamWriters(0, 0, 0, schema.Fields)
writer, err := NewBinlogSerializeWriter(schema, 0, 0, writers, 1024)
assert.NoError(t, err)
m := make(map[FieldID]any)
for _, fs := range schema.Fields {
m[fs.FieldID] = nil
}
m[common.RowIDField] = int64(0)
m[common.TimeStampField] = int64(0)
pk, err := GenPrimaryKeyByRawData(m[common.RowIDField], schemapb.DataType_Int64)
assert.NoError(t, err)
value := &Value{
ID: 0,
PK: pk,
Timestamp: 0,
IsDeleted: false,
Value: m,
}
writer.Write(value)
err = writer.Close()
assert.NoError(t, err)
// Read from the written data
blobs := make([]*Blob, len(writers))
i := 0
for _, w := range writers {
blob, err := w.Finalize()
assert.NoError(t, err)
assert.NotNil(t, blob)
blobs[i] = blob
i++
}
reader, err := NewBinlogDeserializeReader(blobs, common.RowIDField)
assert.NoError(t, err)
defer reader.Close()
err = reader.Next()
assert.NoError(t, err)
readValue := reader.Value()
assert.Equal(t, value, readValue)
})
}
func generateTestDeltalogData(size int) (*Blob, error) {
codec := NewDeleteCodec()
pks := make([]int64, size)
tss := make([]uint64, size)
for i := 0; i < size; i++ {
pks[i] = int64(i)
tss[i] = uint64(i + 1)
}
data := &DeleteData{}
for i := range pks {
data.Append(NewInt64PrimaryKey(pks[i]), tss[i])
}
return codec.Serialize(0, 0, 0, data)
}
func assertTestDeltalogData(t *testing.T, i int, value *DeleteLog) {
assert.Equal(t, &Int64PrimaryKey{int64(i)}, value.Pk)
assert.Equal(t, uint64(i+1), value.Ts)
}
func TestDeltalogDeserializeReader(t *testing.T) {
t.Run("test empty data", func(t *testing.T) {
reader, err := NewDeltalogDeserializeReader(nil)
assert.NoError(t, err)
defer reader.Close()
err = reader.Next()
assert.Equal(t, io.EOF, err)
})
t.Run("test deserialize", func(t *testing.T) {
size := 3
blob, err := generateTestDeltalogData(size)
assert.NoError(t, err)
reader, err := NewDeltalogDeserializeReader([]*Blob{blob})
assert.NoError(t, err)
defer reader.Close()
for i := 0; i < size; i++ {
err = reader.Next()
assert.NoError(t, err)
value := reader.Value()
assertTestDeltalogData(t, i, value)
}
err = reader.Next()
assert.Equal(t, io.EOF, err)
})
}
func TestDeltalogSerializeWriter(t *testing.T) {
t.Run("test empty data", func(t *testing.T) {
reader, err := NewDeltalogDeserializeReader(nil)
assert.NoError(t, err)
defer reader.Close()
err = reader.Next()
assert.Equal(t, io.EOF, err)
})
t.Run("test serialize", func(t *testing.T) {
size := 16
blob, err := generateTestDeltalogData(size)
assert.NoError(t, err)
reader, err := NewDeltalogDeserializeReader([]*Blob{blob})
assert.NoError(t, err)
defer reader.Close()
// Copy write the generated data
eventWriter := NewDeltalogStreamWriter(0, 0, 0)
writer, err := NewDeltalogSerializeWriter(0, 0, eventWriter, 7)
assert.NoError(t, err)
for i := 0; i < size; i++ {
err = reader.Next()
assert.NoError(t, err)
value := reader.Value()
assertTestDeltalogData(t, i, value)
err := writer.Write(value)
assert.NoError(t, err)
}
err = reader.Next()
assert.Equal(t, io.EOF, err)
err = writer.Close()
assert.NoError(t, err)
// Read from the written data
newblob, err := eventWriter.Finalize()
assert.NoError(t, err)
assert.NotNil(t, newblob)
// assert.Equal(t, blobs[0].Value, newblobs[0].Value)
reader, err = NewDeltalogDeserializeReader([]*Blob{newblob})
assert.NoError(t, err)
defer reader.Close()
for i := 0; i < size; i++ {
err = reader.Next()
assert.NoError(t, err, i)
value := reader.Value()
assertTestDeltalogData(t, i, value)
}
})
}
func TestDeltalogPkTsSeparateFormat(t *testing.T) {
t.Run("test empty data", func(t *testing.T) {
eventWriter := NewMultiFieldDeltalogStreamWriter(0, 0, 0, nil)
writer, err := NewDeltalogMultiFieldWriter(0, 0, eventWriter, 7)
assert.NoError(t, err)
defer writer.Close()
err = writer.Close()
assert.NoError(t, err)
blob, err := eventWriter.Finalize()
assert.NoError(t, err)
assert.NotNil(t, blob)
})
testCases := []struct {
name string
pkType schemapb.DataType
assertPk func(t *testing.T, i int, value *DeleteLog)
}{
{
name: "test int64 pk",
pkType: schemapb.DataType_Int64,
assertPk: func(t *testing.T, i int, value *DeleteLog) {
assert.Equal(t, NewInt64PrimaryKey(int64(i)), value.Pk)
assert.Equal(t, uint64(i+1), value.Ts)
},
},
{
name: "test varchar pk",
pkType: schemapb.DataType_VarChar,
assertPk: func(t *testing.T, i int, value *DeleteLog) {
assert.Equal(t, NewVarCharPrimaryKey(strconv.Itoa(i)), value.Pk)
assert.Equal(t, uint64(i+1), value.Ts)
},
},
}
for _, tc := range testCases {
t.Run(tc.name, func(t *testing.T) {
// serialize data
size := 10
blob, err := writeDeltalogNewFormat(size, tc.pkType, 7)
assert.NoError(t, err)
// Deserialize data
reader, err := NewDeltalogDeserializeReader([]*Blob{blob})
assert.NoError(t, err)
defer reader.Close()
for i := 0; i < size; i++ {
err = reader.Next()
assert.NoError(t, err)
value := reader.Value()
tc.assertPk(t, i, value)
}
err = reader.Next()
assert.Equal(t, io.EOF, err)
})
}
}
func BenchmarkDeltalogReader(b *testing.B) {
size := 1000000
blob, err := generateTestDeltalogData(size)
assert.NoError(b, err)
b.Run("one string format reader", func(b *testing.B) {
b.ResetTimer()
for i := 0; i < b.N; i++ {
readDeltaLog(size, blob)
}
})
blob, err = writeDeltalogNewFormat(size, schemapb.DataType_Int64, size)
assert.NoError(b, err)
b.Run("pk ts separate format reader", func(b *testing.B) {
b.ResetTimer()
for i := 0; i < b.N; i++ {
readDeltaLog(size, blob)
}
})
}
func BenchmarkDeltalogFormatWriter(b *testing.B) {
size := 1000000
b.Run("one string format writer", func(b *testing.B) {
b.ResetTimer()
for i := 0; i < b.N; i++ {
eventWriter := NewDeltalogStreamWriter(0, 0, 0)
writer, _ := NewDeltalogSerializeWriter(0, 0, eventWriter, size)
var value *DeleteLog
for j := 0; j < size; j++ {
value = NewDeleteLog(NewInt64PrimaryKey(int64(j)), uint64(j+1))
writer.Write(value)
}
writer.Close()
eventWriter.Finalize()
}
b.ReportAllocs()
})
b.Run("pk and ts separate format writer", func(b *testing.B) {
b.ResetTimer()
for i := 0; i < b.N; i++ {
writeDeltalogNewFormat(size, schemapb.DataType_Int64, size)
}
b.ReportAllocs()
})
}
func writeDeltalogNewFormat(size int, pkType schemapb.DataType, batchSize int) (*Blob, error) {
var err error
eventWriter := NewMultiFieldDeltalogStreamWriter(0, 0, 0, []*schemapb.FieldSchema{
{FieldID: common.RowIDField, Name: "pk", DataType: pkType},
{FieldID: common.TimeStampField, Name: "ts", DataType: schemapb.DataType_Int64},
})
writer, err := NewDeltalogMultiFieldWriter(0, 0, eventWriter, batchSize)
if err != nil {
return nil, err
}
var value *DeleteLog
for i := 0; i < size; i++ {
switch pkType {
case schemapb.DataType_Int64:
value = NewDeleteLog(NewInt64PrimaryKey(int64(i)), uint64(i+1))
case schemapb.DataType_VarChar:
value = NewDeleteLog(NewVarCharPrimaryKey(strconv.Itoa(i)), uint64(i+1))
}
if err = writer.Write(value); err != nil {
return nil, err
}
}
if err = writer.Close(); err != nil {
return nil, err
}
blob, err := eventWriter.Finalize()
if err != nil {
return nil, err
}
return blob, nil
}
func readDeltaLog(size int, blob *Blob) error {
reader, err := NewDeltalogDeserializeReader([]*Blob{blob})
if err != nil {
return err
}
defer reader.Close()
for j := 0; j < size; j++ {
err = reader.Next()
_ = reader.Value()
if err != nil {
return err
}
}
return nil
}