/
scan.go
198 lines (171 loc) · 4.18 KB
/
scan.go
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
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
package executor
import (
"github.com/gotodb/gotodb/connector"
"github.com/gotodb/gotodb/pb"
"github.com/gotodb/gotodb/row"
"github.com/gotodb/gotodb/stage"
"github.com/gotodb/gotodb/util"
"github.com/vmihailenco/msgpack"
"io"
"sync"
)
func (e *Executor) SetInstructionScan(instruction *pb.Instruction) error {
var job stage.ScanJob
if err := msgpack.Unmarshal(instruction.EncodedStageJobBytes, &job); err != nil {
return err
}
//partition info must decode firstly
if err := job.Partition.Decode(); err != nil {
return err
}
e.StageJob = &job
return nil
}
func (e *Executor) RunScan() error {
job := e.StageJob.(*stage.ScanJob)
ctr, err := connector.NewConnector(job.Catalog, job.Schema, job.Table)
if err != nil {
return err
}
//send metadata
for i := 0; i < len(e.Writers); i++ {
if err = util.WriteObject(e.Writers[i], job.Metadata); err != nil {
return err
}
}
colIndexes := job.Metadata.GetColumnIndexes()
inputMetadata := job.Metadata
rbWriters := make([]*row.RowsBuffer, len(e.Writers))
for i, writer := range e.Writers {
rbWriters[i] = row.NewRowsBuffer(job.Metadata, nil, writer)
}
for _, filter := range job.Filters {
if err := filter.Init(job.Metadata); err != nil {
return err
}
}
//send rows
jobs := make(chan *row.RowsGroup)
var wg sync.WaitGroup
for _, rbWriter := range rbWriters {
wg.Add(1)
go func(rbWriter *row.RowsBuffer) {
defer func() {
wg.Done()
}()
for {
rg, ok := <-jobs
if ok {
for _, filter := range job.Filters { //TODO: improve performance, add flag in RowsGroup?
iFlags, err := filter.Result(rg)
if err != nil {
e.AddLogInfo(err, pb.LogLevel_ERR)
break
}
rgTmp := row.NewRowsGroup(job.Metadata)
for i, flag := range iFlags.([]interface{}) {
if flag.(bool) {
rgTmp.AppendRowVals(rg.GetRowVals(i)...)
}
}
rg = rgTmp
}
if err := rbWriter.Write(rg); err != nil {
e.AddLogInfo(err, pb.LogLevel_ERR)
break
}
} else {
break
}
}
}(rbWriter)
}
var filters []string
for _, filter := range job.Filters {
filters = append(filters, filter.Clause)
}
// no partitions
if !job.Partition.IsPartition() {
for _, file := range job.Partition.GetNoPartitionFiles() {
var reader row.GroupReader
reader, err = ctr.GetReader(file, inputMetadata, filters)
if err != nil {
break
}
for err == nil {
var rg *row.RowsGroup
rg, err = reader(colIndexes)
if err == io.EOF {
err = nil
break
}
if err != nil {
break
}
jobs <- rg
}
}
} else { // partitioned
parColNum := job.Partition.GetPartitionColumnNum()
totColNum := inputMetadata.GetColumnNumber()
dataColNum := totColNum - parColNum
var dataCols []int
var parCols []int
for _, index := range colIndexes {
if index < dataColNum {
dataCols = append(dataCols, index) //column from input
} else {
parCols = append(parCols, index-dataColNum) //column from partition
}
}
parMD := inputMetadata.SelectColumnsByIndexes(parCols)
for i := totColNum - 1; i >= dataColNum; i-- {
inputMetadata.DeleteColumnByIndex(i)
}
for i := 0; i < job.Partition.GetPartitionNum(); i++ {
parFullRow := job.Partition.GetPartitionRow(i)
parRow := row.NewRow()
for _, index := range parCols {
parRow.AppendVals(parFullRow.Vals[index])
}
for _, file := range job.Partition.GetPartitionFiles(i) {
var reader row.GroupReader
reader, err = ctr.GetReader(file, inputMetadata, filters)
if err != nil {
break
}
for err == nil {
var dataRG *row.RowsGroup
dataRG, err = reader(dataCols)
if err == io.EOF {
err = nil
break
}
if err != nil {
break
}
parRG := row.NewRowsGroup(parMD)
for i := 0; i < dataRG.GetRowsNumber(); i++ {
parRG.Write(parRow)
}
rg := row.NewRowsGroup(job.Metadata)
rg.ClearColumns()
rg.AppendValColumns(dataRG.Vals...)
rg.AppendValColumns(parRG.Vals...)
jobs <- rg
}
}
}
}
close(jobs)
wg.Wait()
if err != nil {
return err
}
for _, rbWriter := range rbWriters {
if err = rbWriter.Flush(); err != nil {
return err
}
}
return nil
}