1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
|
package logstore
import (
"context"
"encoding/binary"
"fmt"
"github.com/parquet-go/parquet-go"
"github.com/seaweedfs/seaweedfs/weed/filer"
"github.com/seaweedfs/seaweedfs/weed/mq/schema"
"github.com/seaweedfs/seaweedfs/weed/mq/topic"
"github.com/seaweedfs/seaweedfs/weed/pb/filer_pb"
"github.com/seaweedfs/seaweedfs/weed/pb/mq_pb"
"github.com/seaweedfs/seaweedfs/weed/util/chunk_cache"
"github.com/seaweedfs/seaweedfs/weed/util/log_buffer"
"google.golang.org/protobuf/proto"
"io"
"math"
"strings"
)
var (
chunkCache = chunk_cache.NewChunkCacheInMemory(256) // 256 entries, 8MB max per entry
)
func GenParquetReadFunc(filerClient filer_pb.FilerClient, t topic.Topic, p topic.Partition) log_buffer.LogReadFromDiskFuncType {
partitionDir := topic.PartitionDir(t, p)
lookupFileIdFn := filer.LookupFn(filerClient)
// read topic conf from filer
var topicConf *mq_pb.ConfigureTopicResponse
var err error
if err := filerClient.WithFilerClient(false, func(client filer_pb.SeaweedFilerClient) error {
topicConf, err = t.ReadConfFile(client)
return err
}); err != nil {
return nil
}
recordType := topicConf.GetRecordType()
recordType = schema.NewRecordTypeBuilder(recordType).
WithField(SW_COLUMN_NAME_TS, schema.TypeInt64).
WithField(SW_COLUMN_NAME_KEY, schema.TypeBytes).
RecordTypeEnd()
parquetSchema, err := schema.ToParquetSchema(t.Name, recordType)
if err != nil {
return nil
}
parquetLevels, err := schema.ToParquetLevels(recordType)
if err != nil {
return nil
}
// eachFileFn reads a parquet file and calls eachLogEntryFn for each log entry
eachFileFn := func(entry *filer_pb.Entry, eachLogEntryFn log_buffer.EachLogEntryFuncType, starTsNs, stopTsNs int64) (processedTsNs int64, err error) {
// create readerAt for the parquet file
fileSize := filer.FileSize(entry)
visibleIntervals, _ := filer.NonOverlappingVisibleIntervals(context.Background(), lookupFileIdFn, entry.Chunks, 0, int64(fileSize))
chunkViews := filer.ViewFromVisibleIntervals(visibleIntervals, 0, int64(fileSize))
readerCache := filer.NewReaderCache(32, chunkCache, lookupFileIdFn)
readerAt := filer.NewChunkReaderAtFromClient(readerCache, chunkViews, int64(fileSize))
// create parquet reader
parquetReader := parquet.NewReader(readerAt, parquetSchema)
rows := make([]parquet.Row, 128)
for {
rowCount, readErr := parquetReader.ReadRows(rows)
for i := 0; i < rowCount; i++ {
row := rows[i]
// convert parquet row to schema_pb.RecordValue
recordValue, err := schema.ToRecordValue(recordType, parquetLevels, row)
if err != nil {
return processedTsNs, fmt.Errorf("ToRecordValue failed: %v", err)
}
processedTsNs = recordValue.Fields[SW_COLUMN_NAME_TS].GetInt64Value()
if processedTsNs <= starTsNs {
continue
}
if stopTsNs != 0 && processedTsNs >= stopTsNs {
return processedTsNs, nil
}
data, marshalErr := proto.Marshal(recordValue)
if marshalErr != nil {
return processedTsNs, fmt.Errorf("marshal record value: %v", marshalErr)
}
logEntry := &filer_pb.LogEntry{
Key: recordValue.Fields[SW_COLUMN_NAME_KEY].GetBytesValue(),
TsNs: processedTsNs,
Data: data,
}
// fmt.Printf(" parquet entry %s ts %v\n", string(logEntry.Key), time.Unix(0, logEntry.TsNs).UTC())
if _, err = eachLogEntryFn(logEntry); err != nil {
return processedTsNs, fmt.Errorf("process log entry %v: %v", logEntry, err)
}
}
if readErr != nil {
if readErr == io.EOF {
return processedTsNs, nil
}
return processedTsNs, readErr
}
}
return
}
return func(startPosition log_buffer.MessagePosition, stopTsNs int64, eachLogEntryFn log_buffer.EachLogEntryFuncType) (lastReadPosition log_buffer.MessagePosition, isDone bool, err error) {
startFileName := startPosition.UTC().Format(topic.TIME_FORMAT)
startTsNs := startPosition.Time.UnixNano()
var processedTsNs int64
err = filerClient.WithFilerClient(false, func(client filer_pb.SeaweedFilerClient) error {
return filer_pb.SeaweedList(context.Background(), client, partitionDir, "", func(entry *filer_pb.Entry, isLast bool) error {
if entry.IsDirectory {
return nil
}
if !strings.HasSuffix(entry.Name, ".parquet") {
return nil
}
if len(entry.Extended) == 0 {
return nil
}
// read minTs from the parquet file
minTsBytes := entry.Extended["min"]
if len(minTsBytes) != 8 {
return nil
}
minTsNs := int64(binary.BigEndian.Uint64(minTsBytes))
// read max ts
maxTsBytes := entry.Extended["max"]
if len(maxTsBytes) != 8 {
return nil
}
maxTsNs := int64(binary.BigEndian.Uint64(maxTsBytes))
if stopTsNs != 0 && stopTsNs <= minTsNs {
isDone = true
return nil
}
if maxTsNs < startTsNs {
return nil
}
if processedTsNs, err = eachFileFn(entry, eachLogEntryFn, startTsNs, stopTsNs); err != nil {
return err
}
return nil
}, startFileName, true, math.MaxInt32)
})
lastReadPosition = log_buffer.NewMessagePosition(processedTsNs, -2)
return
}
}
|
