1 // Copyright 2010 The Go Authors. All rights reserved.
2 // Use of this source code is governed by a BSD-style
3 // license that can be found in the LICENSE file.
4 5 // Package suffixarray implements substring search in logarithmic time using
6 // an in-memory suffix array.
7 //
8 // Example use:
9 //
10 // // create index for some data
11 // index := suffixarray.New(data)
12 //
13 // // lookup byte slice s
14 // offsets1 := index.Lookup(s, -1) // the list of all indices where s occurs in data
15 // offsets2 := index.Lookup(s, 3) // the list of at most 3 indices where s occurs in data
16 package suffixarray
17 18 import (
19 "bytes"
20 "encoding/binary"
21 "errors"
22 "io"
23 "math"
24 "regexp"
25 "slices"
26 "sort"
27 )
28 29 // Can change for testing
30 var maxData32 int = realMaxData32
31 32 const realMaxData32 = math.MaxInt32
33 34 // Index implements a suffix array for fast substring search.
35 type Index struct {
36 data []byte
37 sa ints // suffix array for data; sa.len() == len(data)
38 }
39 40 // An ints is either an []int32 or an []int64.
41 // That is, one of them is empty, and one is the real data.
42 // The int64 form is used when len(data) > maxData32
43 type ints struct {
44 int32 []int32
45 int64 []int64
46 }
47 48 func (a *ints) len() int {
49 return len(a.int32) + len(a.int64)
50 }
51 52 func (a *ints) get(i int) int64 {
53 if a.int32 != nil {
54 return int64(a.int32[i])
55 }
56 return a.int64[i]
57 }
58 59 func (a *ints) set(i int, v int64) {
60 if a.int32 != nil {
61 a.int32[i] = int32(v)
62 } else {
63 a.int64[i] = v
64 }
65 }
66 67 func (a *ints) slice(i, j int) ints {
68 if a.int32 != nil {
69 return ints{a.int32[i:j], nil}
70 }
71 return ints{nil, a.int64[i:j]}
72 }
73 74 // New creates a new [Index] for data.
75 // [Index] creation time is O(N) for N = len(data).
76 func New(data []byte) *Index {
77 ix := &Index{data: data}
78 if len(data) <= maxData32 {
79 ix.sa.int32 = []int32{:len(data)}
80 text_32(data, ix.sa.int32)
81 } else {
82 ix.sa.int64 = []int64{:len(data)}
83 text_64(data, ix.sa.int64)
84 }
85 return ix
86 }
87 88 // writeInt writes an int x to w using buf to buffer the write.
89 func writeInt(w io.Writer, buf []byte, x int) error {
90 binary.PutVarint(buf, int64(x))
91 _, err := w.Write(buf[0:binary.MaxVarintLen64])
92 return err
93 }
94 95 // readInt reads an int x from r using buf to buffer the read and returns x.
96 func readInt(r io.Reader, buf []byte) (int64, error) {
97 _, err := io.ReadFull(r, buf[0:binary.MaxVarintLen64]) // ok to continue with error
98 x, _ := binary.Varint(buf)
99 return x, err
100 }
101 102 // writeSlice writes data[:n] to w and returns n.
103 // It uses buf to buffer the write.
104 func writeSlice(w io.Writer, buf []byte, data ints) (n int, err error) {
105 // encode as many elements as fit into buf
106 p := binary.MaxVarintLen64
107 m := data.len()
108 for ; n < m && p+binary.MaxVarintLen64 <= len(buf); n++ {
109 p += binary.PutUvarint(buf[p:], uint64(data.get(n)))
110 }
111 112 // update buffer size
113 binary.PutVarint(buf, int64(p))
114 115 // write buffer
116 _, err = w.Write(buf[0:p])
117 return
118 }
119 120 var errTooBig = errors.New("suffixarray: data too large")
121 122 // readSlice reads data[:n] from r and returns n.
123 // It uses buf to buffer the read.
124 func readSlice(r io.Reader, buf []byte, data ints) (n int, err error) {
125 // read buffer size
126 var size64 int64
127 size64, err = readInt(r, buf)
128 if err != nil {
129 return
130 }
131 if int64(int(size64)) != size64 || int(size64) < 0 {
132 // We never write chunks this big anyway.
133 return 0, errTooBig
134 }
135 size := int(size64)
136 137 // read buffer w/o the size
138 if _, err = io.ReadFull(r, buf[binary.MaxVarintLen64:size]); err != nil {
139 return
140 }
141 142 // decode as many elements as present in buf
143 for p := binary.MaxVarintLen64; p < size; n++ {
144 x, w := binary.Uvarint(buf[p:])
145 data.set(n, int64(x))
146 p += w
147 }
148 149 return
150 }
151 152 const bufSize = 16 << 10 // reasonable for BenchmarkSaveRestore
153 154 // Read reads the index from r into x; x must not be nil.
155 func (x *Index) Read(r io.Reader) error {
156 // buffer for all reads
157 buf := []byte{:bufSize}
158 159 // read length
160 n64, err := readInt(r, buf)
161 if err != nil {
162 return err
163 }
164 if int64(int(n64)) != n64 || int(n64) < 0 {
165 return errTooBig
166 }
167 n := int(n64)
168 169 // allocate space
170 if 2*n < cap(x.data) || cap(x.data) < n || x.sa.int32 != nil && n > maxData32 || x.sa.int64 != nil && n <= maxData32 {
171 // new data is significantly smaller or larger than
172 // existing buffers - allocate new ones
173 x.data = []byte{:n}
174 x.sa.int32 = nil
175 x.sa.int64 = nil
176 if n <= maxData32 {
177 x.sa.int32 = []int32{:n}
178 } else {
179 x.sa.int64 = []int64{:n}
180 }
181 } else {
182 // re-use existing buffers
183 x.data = x.data[0:n]
184 x.sa = x.sa.slice(0, n)
185 }
186 187 // read data
188 if _, err := io.ReadFull(r, x.data); err != nil {
189 return err
190 }
191 192 // read index
193 sa := x.sa
194 for sa.len() > 0 {
195 n, err := readSlice(r, buf, sa)
196 if err != nil {
197 return err
198 }
199 sa = sa.slice(n, sa.len())
200 }
201 return nil
202 }
203 204 // Write writes the index x to w.
205 func (x *Index) Write(w io.Writer) error {
206 // buffer for all writes
207 buf := []byte{:bufSize}
208 209 // write length
210 if err := writeInt(w, buf, len(x.data)); err != nil {
211 return err
212 }
213 214 // write data
215 if _, err := w.Write(x.data); err != nil {
216 return err
217 }
218 219 // write index
220 sa := x.sa
221 for sa.len() > 0 {
222 n, err := writeSlice(w, buf, sa)
223 if err != nil {
224 return err
225 }
226 sa = sa.slice(n, sa.len())
227 }
228 return nil
229 }
230 231 // Bytes returns the data over which the index was created.
232 // It must not be modified.
233 func (x *Index) Bytes() []byte {
234 return x.data
235 }
236 237 func (x *Index) at(i int) []byte {
238 return x.data[x.sa.get(i):]
239 }
240 241 // lookupAll returns a slice into the matching region of the index.
242 // The runtime is O(log(N)*len(s)).
243 func (x *Index) lookupAll(s []byte) ints {
244 // find matching suffix index range [i:j]
245 // find the first index where s would be the prefix
246 i := sort.Search(x.sa.len(), func(i int) bool { return bytes.Compare(x.at(i), s) >= 0 })
247 // starting at i, find the first index at which s is not a prefix
248 j := i + sort.Search(x.sa.len()-i, func(j int) bool { return !bytes.HasPrefix(x.at(j+i), s) })
249 return x.sa.slice(i, j)
250 }
251 252 // Lookup returns an unsorted list of at most n indices where the byte string s
253 // occurs in the indexed data. If n < 0, all occurrences are returned.
254 // The result is nil if s is empty, s is not found, or n == 0.
255 // Lookup time is O(log(N)*len(s) + len(result)) where N is the
256 // size of the indexed data.
257 func (x *Index) Lookup(s []byte, n int) (result []int) {
258 if len(s) > 0 && n != 0 {
259 matches := x.lookupAll(s)
260 count := matches.len()
261 if n < 0 || count < n {
262 n = count
263 }
264 // 0 <= n <= count
265 if n > 0 {
266 result = []int{:n}
267 if matches.int32 != nil {
268 for i := range result {
269 result[i] = int(matches.int32[i])
270 }
271 } else {
272 for i := range result {
273 result[i] = int(matches.int64[i])
274 }
275 }
276 }
277 }
278 return
279 }
280 281 // FindAllIndex returns a sorted list of non-overlapping matches of the
282 // regular expression r, where a match is a pair of indices specifying
283 // the matched slice of x.Bytes(). If n < 0, all matches are returned
284 // in successive order. Otherwise, at most n matches are returned and
285 // they may not be successive. The result is nil if there are no matches,
286 // or if n == 0.
287 func (x *Index) FindAllIndex(r *regexp.Regexp, n int) (result [][]int) {
288 // a non-empty literal prefix is used to determine possible
289 // match start indices with Lookup
290 prefix, complete := r.LiteralPrefix()
291 lit := []byte(prefix)
292 293 // worst-case scenario: no literal prefix
294 if prefix == "" {
295 return r.FindAllIndex(x.data, n)
296 }
297 298 // if regexp is a literal just use Lookup and convert its
299 // result into match pairs
300 if complete {
301 // Lookup returns indices that may belong to overlapping matches.
302 // After eliminating them, we may end up with fewer than n matches.
303 // If we don't have enough at the end, redo the search with an
304 // increased value n1, but only if Lookup returned all the requested
305 // indices in the first place (if it returned fewer than that then
306 // there cannot be more).
307 for n1 := n; ; n1 += 2 * (n - len(result)) /* overflow ok */ {
308 indices := x.Lookup(lit, n1)
309 if len(indices) == 0 {
310 return
311 }
312 slices.Sort(indices)
313 pairs := []int{:2*len(indices)}
314 result = [][]int{:len(indices)}
315 count := 0
316 prev := 0
317 for _, i := range indices {
318 if count == n {
319 break
320 }
321 // ignore indices leading to overlapping matches
322 if prev <= i {
323 j := 2 * count
324 pairs[j+0] = i
325 pairs[j+1] = i + len(lit)
326 result[count] = pairs[j : j+2]
327 count++
328 prev = i + len(lit)
329 }
330 }
331 result = result[0:count]
332 if len(result) >= n || len(indices) != n1 {
333 // found all matches or there's no chance to find more
334 // (n and n1 can be negative)
335 break
336 }
337 }
338 if len(result) == 0 {
339 result = nil
340 }
341 return
342 }
343 344 // regexp has a non-empty literal prefix; Lookup(lit) computes
345 // the indices of possible complete matches; use these as starting
346 // points for anchored searches
347 // (regexp "^" matches beginning of input, not beginning of line)
348 r = regexp.MustCompile("^" + r.String()) // compiles because r compiled
349 350 // same comment about Lookup applies here as in the loop above
351 for n1 := n; ; n1 += 2 * (n - len(result)) /* overflow ok */ {
352 indices := x.Lookup(lit, n1)
353 if len(indices) == 0 {
354 return
355 }
356 slices.Sort(indices)
357 result = result[0:0]
358 prev := 0
359 for _, i := range indices {
360 if len(result) == n {
361 break
362 }
363 m := r.FindIndex(x.data[i:]) // anchored search - will not run off
364 // ignore indices leading to overlapping matches
365 if m != nil && prev <= i {
366 m[0] = i // correct m
367 m[1] += i
368 result = append(result, m)
369 prev = m[1]
370 }
371 }
372 if len(result) >= n || len(indices) != n1 {
373 // found all matches or there's no chance to find more
374 // (n and n1 can be negative)
375 break
376 }
377 }
378 if len(result) == 0 {
379 result = nil
380 }
381 return
382 }
383