-
Notifications
You must be signed in to change notification settings - Fork 0
/
token_internal.go
364 lines (339 loc) · 11.4 KB
/
token_internal.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
package search
import (
"context"
"encoding/json"
"fmt"
"net/url"
"sort"
"strings"
"github.com/pkg/errors"
"gopkg.in/olivere/elastic.v6"
"github.com/Bnei-Baruch/archive-backend/consts"
"github.com/Bnei-Baruch/archive-backend/es"
)
type Span struct {
Start int
End int
}
func MakeSpan(start, end int) *Span {
return &Span{Start: start, End: end}
}
type Token struct {
Token string `json:"token"`
StartOffset int `json:"start_offset"`
EndOffset int `json:"end_offset"`
Type string `json:"type"`
Position int `json:"position"`
PositionLength int `json:"positionLength"`
}
type OriginalTokenNode struct {
OriginalWholePhrase *string
SkippedPrefix *Span // Stopwords before this token that were skipped.
OriginalPhrase string // Original string that was tokenized.
SkippedSuffix *Span // Will be set only for IsEnd = true nodes.
TokenNode *TokenNode
Parents []*OriginalTokenNode
Children []*OriginalTokenNode
}
type TokenNode struct {
Token Token
IsEnd bool
Parents []*TokenNode
Children []*TokenNode
// May be several original phrases per token from several sources that were merged.
OriginalTokenNodes []*OriginalTokenNode
// Caches phrases with origin.
Phrases []*PhrasesWithOrigin
}
func originalPhraseToString(phrase *string, tn *TokenNode, parentTn *TokenNode) string {
isMultiWord := parentTn != nil &&
tn.Token.StartOffset == parentTn.Token.StartOffset &&
tn.Token.EndOffset == parentTn.Token.EndOffset
isSynonym := tn.Token.Type == "SYNONYM"
if isSynonym {
if isMultiWord {
return fmt.Sprintf(" %s", tn.Token.Token)
} else {
return tn.Token.Token
}
} else {
if isMultiWord {
return ""
} else {
return string([]rune(*phrase)[tn.Token.StartOffset:tn.Token.EndOffset])
}
}
}
func (otn *OriginalTokenNode) SkippedSuffixToString() string {
if otn.SkippedSuffix == nil {
return ""
}
return string([]rune(*otn.OriginalWholePhrase)[otn.SkippedSuffix.Start:otn.SkippedSuffix.End])
}
func (otn *OriginalTokenNode) SkippedPrefixToString() string {
if otn.SkippedPrefix == nil {
return ""
}
return string([]rune(*otn.OriginalWholePhrase)[otn.SkippedPrefix.Start:otn.SkippedPrefix.End])
}
func (otn *OriginalTokenNode) OriginalFullPhraseToString() string {
return fmt.Sprintf("%s%s%s",
otn.SkippedPrefixToString(),
otn.OriginalPhrase,
otn.SkippedSuffixToString(),
)
}
func (otn *OriginalTokenNode) OriginalVariableToString() string {
prefix := []rune(otn.SkippedPrefixToString())
if len(prefix) > 0 {
return fmt.Sprintf("%s%s", string(prefix[len(prefix)-1:]), otn.OriginalPhrase)
}
return otn.OriginalPhrase
}
func (otn *OriginalTokenNode) OriginalVariablePrefixToString() string {
prefix := []rune(otn.SkippedPrefixToString())
if len(prefix) > 0 {
return fmt.Sprintf("%s", string(prefix[:len(prefix)-1]))
}
return ""
}
func MakeTokensFromPhrase(phrase string, lang string, esc *elastic.Client, tc *TokensCache) ([]*TokenNode, error) {
if tc != nil && tc.Has(phrase, lang) {
return tc.Get(phrase, lang), nil
}
index := es.IndexNameForServing("prod", consts.ES_RESULTS_INDEX, lang)
tokens, err := MakeTokensFromPhraseIndex(phrase, lang, esc, index, context.TODO())
if err != nil {
return nil, err
}
if tc != nil {
tc.Set(phrase, lang, tokens)
}
return tokens, nil
}
func MakeTokensFromPhraseIndex(phrase string, lang string, esc *elastic.Client, index string, ctx context.Context) ([]*TokenNode, error) {
// TODO: Skip Variables, don't analyze them.
encodedIndex := url.QueryEscape(index)
res, err := esc.PerformRequest(ctx, elastic.PerformRequestOptions{
Method: "GET",
Path: fmt.Sprintf("/%s/_analyze", encodedIndex),
Body: struct {
Text string `json:"text"`
Analyzer string `json:"analyzer"`
}{
Text: phrase,
Analyzer: consts.ANALYZERS[lang],
},
})
if err != nil {
return nil, errors.Wrapf(err, "Error analyzing [%s] in %s with analyzer %s, index [%s]",
phrase, lang, consts.ANALYZERS[lang], index)
}
tokens := struct {
Tokens []Token `json:"tokens"`
}{Tokens: []Token{}}
if err = json.Unmarshal(res.Body, &tokens); err != nil {
return nil, errors.Wrapf(err, "Error unmarshling analyze body while analyzing [%s] in %s with analyzer %s, index [%s]",
phrase, lang, consts.ANALYZERS[lang], index)
}
tokenNodes := makeTokenForest(tokens.Tokens, phrase)
return tokenNodes, nil
}
func makeTokenForest(tokens []Token, phrase string) []*TokenNode {
tokenRoot := []*TokenNode{}
tokenEnd := [][]*TokenNode{}
minPos := -1
for i := range tokens {
// Update PositionLength, 0 basically means 1.
if tokens[i].PositionLength == 0 {
tokens[i].PositionLength = 1
}
if tokens[i].Position < minPos || minPos == -1 {
minPos = tokens[i].Position
}
}
for _, t := range tokens {
node := TokenNode{Token: t}
if t.Position == minPos {
tokenRoot = append(tokenRoot, &node)
}
for len(tokenEnd) <= t.Position+t.PositionLength {
tokenEnd = append(tokenEnd, []*TokenNode{})
}
tokenEnd[t.Position+t.PositionLength] = append(tokenEnd[t.Position+t.PositionLength], &node)
if t.Position > 0 {
for i := range tokenEnd[t.Position] {
tokenEnd[t.Position][i].Children = append(tokenEnd[t.Position][i].Children, &node)
node.Parents = append(node.Parents, tokenEnd[t.Position][i])
}
}
}
// Skip gaps if exist, as positions might jump.
i := 0
for i < len(tokenEnd) {
if len(tokenEnd[i]) == 0 {
if i > 0 && i < len(tokenEnd)-1 {
// connect
for a := range tokenEnd[i-1] {
for b := range tokenEnd[i+1] {
tokenEnd[i-1][a].Children = append(tokenEnd[i-1][a].Children, tokenEnd[i+1][b])
tokenEnd[i+1][b].Parents = append(tokenEnd[i+1][b].Parents, tokenEnd[i-1][a])
}
}
}
tokenEnd = append(tokenEnd[:i], tokenEnd[i+1:]...)
} else {
i++
}
}
// Set end tokens.
if len(tokenEnd) > 0 {
for i := range tokenEnd[len(tokenEnd)-1] {
tokenEnd[len(tokenEnd)-1][i].IsEnd = true
}
}
// Fill skipped phrases
fillOriginalTokens(tokenRoot, nil, &phrase, 0, make(map[*TokenNode]bool))
// Sort
tokenRoot = sortTokenGraph(tokenRoot, nil, make(map[*TokenNode]bool), 0)
originalTokenRoot := []*OriginalTokenNode{}
for i := range tokenRoot {
originalTokenRoot = append(originalTokenRoot, tokenRoot[i].OriginalTokenNodes...)
}
sortOriginalTokenGraph(originalTokenRoot, nil, make(map[*OriginalTokenNode]bool), 0)
// Uncommend for debug.
// depth := 0
// iterator := tokenRoot
// for iterator != nil && len(iterator) > 0 {
// log.Infof("\n\n======= DEPTH %d ======\n", depth)
// for i := range iterator {
// log.Infof("%sTN: %p %p %+v", strings.Repeat(" ", depth), &iterator, &iterator[i], iterator[i])
// //log.Infof("OTN: %+v", iterator[i].OriginalTokenNodes)
// for j := range iterator[i].OriginalTokenNodes {
// log.Infof(" %sOTN[%d]: %p %+v\n", strings.Repeat(" ", depth), j, iterator[i].OriginalTokenNodes[j], iterator[i].OriginalTokenNodes[j])
// }
// }
// iterator = iterator[0].Children
// depth++
// }
return tokenRoot
}
// Merges second token to first if they have the same Token.
// Returns true if were merged. False otherwise.
func mergeSortTokens(first *TokenNode, second *TokenNode) {
for i := range second.Parents {
for j := range second.Parents[i].Children {
if second.Parents[i].Children[j] == second {
second.Parents[i].Children = append(second.Parents[i].Children[:j], second.Parents[i].Children[j+1:]...)
break
}
}
}
for i := range second.Children {
for j := range second.Children[i].Parents {
if second.Children[i].Parents[j] == second {
second.Children[i].Parents = append(second.Children[i].Parents[:j], second.Children[i].Parents[j+1:]...)
break
}
}
}
parentsMap := make(map[*TokenNode]bool)
for i := range first.Parents {
parentsMap[first.Parents[i]] = true
}
childrenMap := make(map[*TokenNode]bool)
for i := range first.Children {
parentsMap[first.Children[i]] = true
}
for i := range second.Parents {
if _, ok := parentsMap[second.Parents[i]]; !ok {
first.Parents = append(first.Parents, second.Parents[i])
}
}
for i := range second.Children {
if _, ok := childrenMap[second.Children[i]]; !ok {
first.Children = append(first.Children, second.Children[i])
}
}
for i := range second.OriginalTokenNodes {
first.OriginalTokenNodes = append(first.OriginalTokenNodes, second.OriginalTokenNodes[i])
}
}
func sortTokenGraph(children []*TokenNode, parent *TokenNode, visited map[*TokenNode]bool, depth int) []*TokenNode {
for i := range children {
tn := children[i]
if _, ok := visited[tn]; !ok {
tn.Children = sortTokenGraph(tn.Children, tn, visited, depth+1)
visited[tn] = true
}
}
// Sort and merch children after all their children are sorted and merged.
sort.SliceStable(children, func(i, j int) bool {
return strings.Compare(children[i].Token.Token, children[j].Token.Token) < 0
})
return children
}
func sortOriginalTokenGraph(children []*OriginalTokenNode, parent *OriginalTokenNode, visited map[*OriginalTokenNode]bool, depth int) []*OriginalTokenNode {
for i := range children {
otn := children[i]
if _, ok := visited[otn]; !ok {
otn.Children = sortOriginalTokenGraph(otn.Children, otn, visited, depth+1)
visited[otn] = true
}
}
// Sort and merch children after all their children are sorted and merged.
sort.SliceStable(children, func(i, j int) bool {
return strings.Compare(children[i].OriginalFullPhraseToString(), children[j].OriginalFullPhraseToString()) < 0
})
return children
}
func makeOriginalTokenNodes(tokenNode *TokenNode, parent *TokenNode, phrase *string, doneIndex int) {
prefixSpan := (*Span)(nil)
if doneIndex < tokenNode.Token.StartOffset {
prefix := string([]rune(*phrase)[doneIndex:tokenNode.Token.StartOffset])
if prefix != "" {
prefixSpan = MakeSpan(doneIndex, tokenNode.Token.StartOffset)
}
}
suffixSpan := (*Span)(nil)
if tokenNode.IsEnd && tokenNode.Token.EndOffset < len([]rune(*phrase)) {
suffix := string([]rune(*phrase)[tokenNode.Token.EndOffset:])
if suffix != "" {
suffixSpan = MakeSpan(tokenNode.Token.EndOffset, len([]rune(*phrase)))
}
}
parentOriginalTokenNodes := []*OriginalTokenNode{nil}
if parent != nil {
parentOriginalTokenNodes = parent.OriginalTokenNodes
}
for i := range parentOriginalTokenNodes {
originalTokenNode := (*OriginalTokenNode)(nil)
if len(tokenNode.OriginalTokenNodes) == 1 {
originalTokenNode = tokenNode.OriginalTokenNodes[0]
} else if len(tokenNode.OriginalTokenNodes) > 1 {
panic("Should never be more than 1")
} else {
originalTokenNode = &OriginalTokenNode{
OriginalWholePhrase: phrase,
SkippedPrefix: prefixSpan,
OriginalPhrase: originalPhraseToString(phrase, tokenNode, parent),
SkippedSuffix: suffixSpan,
TokenNode: tokenNode,
}
tokenNode.OriginalTokenNodes = append(tokenNode.OriginalTokenNodes, originalTokenNode)
}
if parentOriginalTokenNodes[i] != nil {
originalTokenNode.Parents = append(originalTokenNode.Parents, parentOriginalTokenNodes[i])
parentOriginalTokenNodes[i].Children = append(parentOriginalTokenNodes[i].Children, originalTokenNode)
}
}
}
func fillOriginalTokens(children []*TokenNode, parent *TokenNode, phrase *string, doneIndex int, visited map[*TokenNode]bool) {
for i := range children {
makeOriginalTokenNodes(children[i], parent, phrase, doneIndex)
if _, ok := visited[children[i]]; !ok {
visited[children[i]] = true
fillOriginalTokens(children[i].Children, children[i], phrase, children[i].Token.EndOffset, visited)
}
}
}