/
table.go
381 lines (329 loc) · 9.92 KB
/
table.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
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
// Licensed to the Apache Software Foundation (ASF) 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 array
import (
"errors"
"fmt"
"math"
"sync/atomic"
"github.com/apache/arrow/go/v12/arrow"
"github.com/apache/arrow/go/v12/arrow/internal/debug"
)
// NewColumnSlice returns a new zero-copy slice of the column with the indicated
// indices i and j, corresponding to the column's array[i:j].
// The returned column must be Release()'d after use.
//
// NewColSlice panics if the slice is outside the valid range of the column's array.
// NewColSlice panics if j < i.
func NewColumnSlice(col *arrow.Column, i, j int64) *arrow.Column {
slice := NewChunkedSlice(col.Data(), i, j)
defer slice.Release()
return arrow.NewColumn(col.Field(), slice)
}
// NewChunkedSlice constructs a zero-copy slice of the chunked array with the indicated
// indices i and j, corresponding to array[i:j].
// The returned chunked array must be Release()'d after use.
//
// NewSlice panics if the slice is outside the valid range of the input array.
// NewSlice panics if j < i.
func NewChunkedSlice(a *arrow.Chunked, i, j int64) *arrow.Chunked {
if j > int64(a.Len()) || i > j || i > int64(a.Len()) {
panic("arrow/array: index out of range")
}
var (
cur = 0
beg = i
sz = j - i
chunks = make([]arrow.Array, 0, len(a.Chunks()))
)
for cur < len(a.Chunks()) && beg >= int64(a.Chunks()[cur].Len()) {
beg -= int64(a.Chunks()[cur].Len())
cur++
}
for cur < len(a.Chunks()) && sz > 0 {
arr := a.Chunks()[cur]
end := beg + sz
if end > int64(arr.Len()) {
end = int64(arr.Len())
}
chunks = append(chunks, NewSlice(arr, beg, end))
sz -= int64(arr.Len()) - beg
beg = 0
cur++
}
chunks = chunks[:len(chunks):len(chunks)]
defer func() {
for _, chunk := range chunks {
chunk.Release()
}
}()
return arrow.NewChunked(a.DataType(), chunks)
}
// simpleTable is a basic, non-lazy in-memory table.
type simpleTable struct {
refCount int64
rows int64
cols []arrow.Column
schema *arrow.Schema
}
// NewTable returns a new basic, non-lazy in-memory table.
// If rows is negative, the number of rows will be inferred from the height
// of the columns.
//
// NewTable panics if the columns and schema are inconsistent.
// NewTable panics if rows is larger than the height of the columns.
func NewTable(schema *arrow.Schema, cols []arrow.Column, rows int64) *simpleTable {
tbl := simpleTable{
refCount: 1,
rows: rows,
cols: cols,
schema: schema,
}
if tbl.rows < 0 {
switch len(tbl.cols) {
case 0:
tbl.rows = 0
default:
tbl.rows = int64(tbl.cols[0].Len())
}
}
// validate the table and its constituents.
// note we retain the columns after having validated the table
// in case the validation fails and panics (and would otherwise leak
// a ref-count on the columns.)
tbl.validate()
for i := range tbl.cols {
tbl.cols[i].Retain()
}
return &tbl
}
// NewTableFromSlice is a convenience function to create a table from a slice
// of slices of arrow.Array.
//
// Like other NewTable functions this can panic if:
// - len(schema.Fields) != len(data)
// - the total length of each column's array slice (ie: number of rows
// in the column) aren't the same for all columns.
func NewTableFromSlice(schema *arrow.Schema, data [][]arrow.Array) *simpleTable {
if len(data) != len(schema.Fields()) {
panic("array/table: mismatch in number of columns and data for creating a table")
}
cols := make([]arrow.Column, len(schema.Fields()))
for i, arrs := range data {
field := schema.Field(i)
chunked := arrow.NewChunked(field.Type, arrs)
cols[i] = *arrow.NewColumn(field, chunked)
chunked.Release()
}
tbl := simpleTable{
refCount: 1,
schema: schema,
cols: cols,
rows: int64(cols[0].Len()),
}
defer func() {
if r := recover(); r != nil {
// if validate panics, let's release the columns
// so that we don't leak them, then propagate the panic
for _, c := range cols {
c.Release()
}
panic(r)
}
}()
// validate the table and its constituents.
tbl.validate()
return &tbl
}
// NewTableFromRecords returns a new basic, non-lazy in-memory table.
//
// NewTableFromRecords panics if the records and schema are inconsistent.
func NewTableFromRecords(schema *arrow.Schema, recs []arrow.Record) *simpleTable {
arrs := make([]arrow.Array, len(recs))
cols := make([]arrow.Column, len(schema.Fields()))
defer func(cols []arrow.Column) {
for i := range cols {
cols[i].Release()
}
}(cols)
for i := range cols {
field := schema.Field(i)
for j, rec := range recs {
arrs[j] = rec.Column(i)
}
chunk := arrow.NewChunked(field.Type, arrs)
cols[i] = *arrow.NewColumn(field, chunk)
chunk.Release()
}
return NewTable(schema, cols, -1)
}
func (tbl *simpleTable) Schema() *arrow.Schema { return tbl.schema }
func (tbl *simpleTable) NumRows() int64 { return tbl.rows }
func (tbl *simpleTable) NumCols() int64 { return int64(len(tbl.cols)) }
func (tbl *simpleTable) Column(i int) *arrow.Column { return &tbl.cols[i] }
func (tbl *simpleTable) validate() {
if len(tbl.cols) != len(tbl.schema.Fields()) {
panic(errors.New("arrow/array: table schema mismatch"))
}
for i, col := range tbl.cols {
if !col.Field().Equal(tbl.schema.Field(i)) {
panic(fmt.Errorf("arrow/array: column field %q is inconsistent with schema", col.Name()))
}
if int64(col.Len()) < tbl.rows {
panic(fmt.Errorf("arrow/array: column %q expected length >= %d but got length %d", col.Name(), tbl.rows, col.Len()))
}
}
}
// Retain increases the reference count by 1.
// Retain may be called simultaneously from multiple goroutines.
func (tbl *simpleTable) Retain() {
atomic.AddInt64(&tbl.refCount, 1)
}
// Release decreases the reference count by 1.
// When the reference count goes to zero, the memory is freed.
// Release may be called simultaneously from multiple goroutines.
func (tbl *simpleTable) Release() {
debug.Assert(atomic.LoadInt64(&tbl.refCount) > 0, "too many releases")
if atomic.AddInt64(&tbl.refCount, -1) == 0 {
for i := range tbl.cols {
tbl.cols[i].Release()
}
tbl.cols = nil
}
}
// TableReader is a Record iterator over a (possibly chunked) Table
type TableReader struct {
refCount int64
tbl arrow.Table
cur int64 // current row
max int64 // total number of rows
rec arrow.Record // current Record
chksz int64 // chunk size
chunks []*arrow.Chunked
slots []int // chunk indices
offsets []int64 // chunk offsets
}
// NewTableReader returns a new TableReader to iterate over the (possibly chunked) Table.
// if chunkSize is <= 0, the biggest possible chunk will be selected.
func NewTableReader(tbl arrow.Table, chunkSize int64) *TableReader {
ncols := tbl.NumCols()
tr := &TableReader{
refCount: 1,
tbl: tbl,
cur: 0,
max: int64(tbl.NumRows()),
chksz: chunkSize,
chunks: make([]*arrow.Chunked, ncols),
slots: make([]int, ncols),
offsets: make([]int64, ncols),
}
tr.tbl.Retain()
if tr.chksz <= 0 {
tr.chksz = math.MaxInt64
}
for i := range tr.chunks {
col := tr.tbl.Column(i)
tr.chunks[i] = col.Data()
tr.chunks[i].Retain()
}
return tr
}
func (tr *TableReader) Schema() *arrow.Schema { return tr.tbl.Schema() }
func (tr *TableReader) Record() arrow.Record { return tr.rec }
func (tr *TableReader) Next() bool {
if tr.cur >= tr.max {
return false
}
if tr.rec != nil {
tr.rec.Release()
}
// determine the minimum contiguous slice across all columns
chunksz := imin64(tr.max, tr.chksz)
chunks := make([]arrow.Array, len(tr.chunks))
for i := range chunks {
j := tr.slots[i]
chunk := tr.chunks[i].Chunk(j)
remain := int64(chunk.Len()) - tr.offsets[i]
if remain < chunksz {
chunksz = remain
}
chunks[i] = chunk
}
// slice the chunks, advance each chunk slot as appropriate.
batch := make([]arrow.Array, len(tr.chunks))
for i, chunk := range chunks {
var slice arrow.Array
offset := tr.offsets[i]
switch int64(chunk.Len()) - offset {
case chunksz:
tr.slots[i]++
tr.offsets[i] = 0
if offset > 0 {
// need to slice
slice = NewSlice(chunk, offset, offset+chunksz)
} else {
// no need to slice
slice = chunk
slice.Retain()
}
default:
tr.offsets[i] += chunksz
slice = NewSlice(chunk, offset, offset+chunksz)
}
batch[i] = slice
}
tr.cur += chunksz
tr.rec = NewRecord(tr.tbl.Schema(), batch, chunksz)
for _, arr := range batch {
arr.Release()
}
return true
}
// Retain increases the reference count by 1.
// Retain may be called simultaneously from multiple goroutines.
func (tr *TableReader) Retain() {
atomic.AddInt64(&tr.refCount, 1)
}
// Release decreases the reference count by 1.
// When the reference count goes to zero, the memory is freed.
// Release may be called simultaneously from multiple goroutines.
func (tr *TableReader) Release() {
debug.Assert(atomic.LoadInt64(&tr.refCount) > 0, "too many releases")
if atomic.AddInt64(&tr.refCount, -1) == 0 {
tr.tbl.Release()
for _, chk := range tr.chunks {
chk.Release()
}
if tr.rec != nil {
tr.rec.Release()
}
tr.tbl = nil
tr.chunks = nil
tr.slots = nil
tr.offsets = nil
}
}
func (tr *TableReader) Err() error { return nil }
func imin64(a, b int64) int64 {
if a < b {
return a
}
return b
}
var (
_ arrow.Table = (*simpleTable)(nil)
_ RecordReader = (*TableReader)(nil)
)