mirror of
https://github.com/golang/go
synced 2024-11-22 03:34:40 -07:00
index/suffixarray: support for serialization
R=r CC=golang-dev https://golang.org/cl/5040041
This commit is contained in:
parent
f5181ae9d7
commit
bd80b1198b
@ -26,7 +26,7 @@ package suffixarray
|
|||||||
|
|
||||||
import "sort"
|
import "sort"
|
||||||
|
|
||||||
func qsufsort(data []byte) []int {
|
func qsufsort(data []byte) []int32 {
|
||||||
// initial sorting by first byte of suffix
|
// initial sorting by first byte of suffix
|
||||||
sa := sortedByFirstByte(data)
|
sa := sortedByFirstByte(data)
|
||||||
if len(sa) < 2 {
|
if len(sa) < 2 {
|
||||||
@ -39,20 +39,20 @@ func qsufsort(data []byte) []int {
|
|||||||
// the index starts 1-ordered
|
// the index starts 1-ordered
|
||||||
sufSortable := &suffixSortable{sa, inv, 1}
|
sufSortable := &suffixSortable{sa, inv, 1}
|
||||||
|
|
||||||
for sa[0] > -len(sa) { // until all suffixes are one big sorted group
|
for int(sa[0]) > -len(sa) { // until all suffixes are one big sorted group
|
||||||
// The suffixes are h-ordered, make them 2*h-ordered
|
// The suffixes are h-ordered, make them 2*h-ordered
|
||||||
pi := 0 // pi is first position of first group
|
pi := 0 // pi is first position of first group
|
||||||
sl := 0 // sl is negated length of sorted groups
|
sl := 0 // sl is negated length of sorted groups
|
||||||
for pi < len(sa) {
|
for pi < len(sa) {
|
||||||
if s := sa[pi]; s < 0 { // if pi starts sorted group
|
if s := int(sa[pi]); s < 0 { // if pi starts sorted group
|
||||||
pi -= s // skip over sorted group
|
pi -= s // skip over sorted group
|
||||||
sl += s // add negated length to sl
|
sl += s // add negated length to sl
|
||||||
} else { // if pi starts unsorted group
|
} else { // if pi starts unsorted group
|
||||||
if sl != 0 {
|
if sl != 0 {
|
||||||
sa[pi+sl] = sl // combine sorted groups before pi
|
sa[pi+sl] = int32(sl) // combine sorted groups before pi
|
||||||
sl = 0
|
sl = 0
|
||||||
}
|
}
|
||||||
pk := inv[s] + 1 // pk-1 is last position of unsorted group
|
pk := int(inv[s]) + 1 // pk-1 is last position of unsorted group
|
||||||
sufSortable.sa = sa[pi:pk]
|
sufSortable.sa = sa[pi:pk]
|
||||||
sort.Sort(sufSortable)
|
sort.Sort(sufSortable)
|
||||||
sufSortable.updateGroups(pi)
|
sufSortable.updateGroups(pi)
|
||||||
@ -60,19 +60,19 @@ func qsufsort(data []byte) []int {
|
|||||||
}
|
}
|
||||||
}
|
}
|
||||||
if sl != 0 { // if the array ends with a sorted group
|
if sl != 0 { // if the array ends with a sorted group
|
||||||
sa[pi+sl] = sl // combine sorted groups at end of sa
|
sa[pi+sl] = int32(sl) // combine sorted groups at end of sa
|
||||||
}
|
}
|
||||||
|
|
||||||
sufSortable.h *= 2 // double sorted depth
|
sufSortable.h *= 2 // double sorted depth
|
||||||
}
|
}
|
||||||
|
|
||||||
for i := range sa { // reconstruct suffix array from inverse
|
for i := range sa { // reconstruct suffix array from inverse
|
||||||
sa[inv[i]] = i
|
sa[inv[i]] = int32(i)
|
||||||
}
|
}
|
||||||
return sa
|
return sa
|
||||||
}
|
}
|
||||||
|
|
||||||
func sortedByFirstByte(data []byte) []int {
|
func sortedByFirstByte(data []byte) []int32 {
|
||||||
// total byte counts
|
// total byte counts
|
||||||
var count [256]int
|
var count [256]int
|
||||||
for _, b := range data {
|
for _, b := range data {
|
||||||
@ -84,17 +84,17 @@ func sortedByFirstByte(data []byte) []int {
|
|||||||
count[b], sum = sum, count[b]+sum
|
count[b], sum = sum, count[b]+sum
|
||||||
}
|
}
|
||||||
// iterate through bytes, placing index into the correct spot in sa
|
// iterate through bytes, placing index into the correct spot in sa
|
||||||
sa := make([]int, len(data))
|
sa := make([]int32, len(data))
|
||||||
for i, b := range data {
|
for i, b := range data {
|
||||||
sa[count[b]] = i
|
sa[count[b]] = int32(i)
|
||||||
count[b]++
|
count[b]++
|
||||||
}
|
}
|
||||||
return sa
|
return sa
|
||||||
}
|
}
|
||||||
|
|
||||||
func initGroups(sa []int, data []byte) []int {
|
func initGroups(sa []int32, data []byte) []int32 {
|
||||||
// label contiguous same-letter groups with the same group number
|
// label contiguous same-letter groups with the same group number
|
||||||
inv := make([]int, len(data))
|
inv := make([]int32, len(data))
|
||||||
prevGroup := len(sa) - 1
|
prevGroup := len(sa) - 1
|
||||||
groupByte := data[sa[prevGroup]]
|
groupByte := data[sa[prevGroup]]
|
||||||
for i := len(sa) - 1; i >= 0; i-- {
|
for i := len(sa) - 1; i >= 0; i-- {
|
||||||
@ -105,7 +105,7 @@ func initGroups(sa []int, data []byte) []int {
|
|||||||
groupByte = b
|
groupByte = b
|
||||||
prevGroup = i
|
prevGroup = i
|
||||||
}
|
}
|
||||||
inv[sa[i]] = prevGroup
|
inv[sa[i]] = int32(prevGroup)
|
||||||
if prevGroup == 0 {
|
if prevGroup == 0 {
|
||||||
sa[0] = -1
|
sa[0] = -1
|
||||||
}
|
}
|
||||||
@ -120,9 +120,9 @@ func initGroups(sa []int, data []byte) []int {
|
|||||||
if data[sa[i]] == lastByte && s == -1 {
|
if data[sa[i]] == lastByte && s == -1 {
|
||||||
s = i
|
s = i
|
||||||
}
|
}
|
||||||
if sa[i] == len(sa)-1 {
|
if int(sa[i]) == len(sa)-1 {
|
||||||
sa[i], sa[s] = sa[s], sa[i]
|
sa[i], sa[s] = sa[s], sa[i]
|
||||||
inv[sa[s]] = s
|
inv[sa[s]] = int32(s)
|
||||||
sa[s] = -1 // mark it as an isolated sorted group
|
sa[s] = -1 // mark it as an isolated sorted group
|
||||||
break
|
break
|
||||||
}
|
}
|
||||||
@ -132,9 +132,9 @@ func initGroups(sa []int, data []byte) []int {
|
|||||||
}
|
}
|
||||||
|
|
||||||
type suffixSortable struct {
|
type suffixSortable struct {
|
||||||
sa []int
|
sa []int32
|
||||||
inv []int
|
inv []int32
|
||||||
h int
|
h int32
|
||||||
}
|
}
|
||||||
|
|
||||||
func (x *suffixSortable) Len() int { return len(x.sa) }
|
func (x *suffixSortable) Len() int { return len(x.sa) }
|
||||||
@ -156,7 +156,7 @@ func (x *suffixSortable) updateGroups(offset int) {
|
|||||||
prev := 0
|
prev := 0
|
||||||
for _, b := range bounds {
|
for _, b := range bounds {
|
||||||
for i := prev; i < b; i++ {
|
for i := prev; i < b; i++ {
|
||||||
x.inv[x.sa[i]] = offset + b - 1
|
x.inv[x.sa[i]] = int32(offset + b - 1)
|
||||||
}
|
}
|
||||||
if b-prev == 1 {
|
if b-prev == 1 {
|
||||||
x.sa[prev] = -1
|
x.sa[prev] = -1
|
||||||
|
@ -18,14 +18,17 @@ package suffixarray
|
|||||||
|
|
||||||
import (
|
import (
|
||||||
"bytes"
|
"bytes"
|
||||||
|
"encoding/binary"
|
||||||
"exp/regexp"
|
"exp/regexp"
|
||||||
|
"io"
|
||||||
|
"os"
|
||||||
"sort"
|
"sort"
|
||||||
)
|
)
|
||||||
|
|
||||||
// Index implements a suffix array for fast substring search.
|
// Index implements a suffix array for fast substring search.
|
||||||
type Index struct {
|
type Index struct {
|
||||||
data []byte
|
data []byte
|
||||||
sa []int // suffix array for data
|
sa []int32 // suffix array for data; len(sa) == len(data)
|
||||||
}
|
}
|
||||||
|
|
||||||
// New creates a new Index for data.
|
// New creates a new Index for data.
|
||||||
@ -34,6 +37,48 @@ func New(data []byte) *Index {
|
|||||||
return &Index{data, qsufsort(data)}
|
return &Index{data, qsufsort(data)}
|
||||||
}
|
}
|
||||||
|
|
||||||
|
// Read reads the index from r into x; x must not be nil.
|
||||||
|
func (x *Index) Read(r io.Reader) os.Error {
|
||||||
|
var n int32
|
||||||
|
if err := binary.Read(r, binary.LittleEndian, &n); err != nil {
|
||||||
|
return err
|
||||||
|
}
|
||||||
|
if 2*n < int32(cap(x.data)) || int32(cap(x.data)) < n {
|
||||||
|
// new data is significantly smaller or larger then
|
||||||
|
// existing buffers - allocate new ones
|
||||||
|
x.data = make([]byte, n)
|
||||||
|
x.sa = make([]int32, n)
|
||||||
|
} else {
|
||||||
|
// re-use existing buffers
|
||||||
|
x.data = x.data[0:n]
|
||||||
|
x.sa = x.sa[0:n]
|
||||||
|
}
|
||||||
|
|
||||||
|
if err := binary.Read(r, binary.LittleEndian, x.data); err != nil {
|
||||||
|
return err
|
||||||
|
}
|
||||||
|
if err := binary.Read(r, binary.LittleEndian, x.sa); err != nil {
|
||||||
|
return err
|
||||||
|
}
|
||||||
|
|
||||||
|
return nil
|
||||||
|
}
|
||||||
|
|
||||||
|
// Write writes the index x to w.
|
||||||
|
func (x *Index) Write(w io.Writer) os.Error {
|
||||||
|
n := int32(len(x.data))
|
||||||
|
if err := binary.Write(w, binary.LittleEndian, n); err != nil {
|
||||||
|
return err
|
||||||
|
}
|
||||||
|
if err := binary.Write(w, binary.LittleEndian, x.data); err != nil {
|
||||||
|
return err
|
||||||
|
}
|
||||||
|
if err := binary.Write(w, binary.LittleEndian, x.sa); err != nil {
|
||||||
|
return err
|
||||||
|
}
|
||||||
|
return nil
|
||||||
|
}
|
||||||
|
|
||||||
// Bytes returns the data over which the index was created.
|
// Bytes returns the data over which the index was created.
|
||||||
// It must not be modified.
|
// It must not be modified.
|
||||||
//
|
//
|
||||||
@ -47,7 +92,7 @@ func (x *Index) at(i int) []byte {
|
|||||||
|
|
||||||
// lookupAll returns a slice into the matching region of the index.
|
// lookupAll returns a slice into the matching region of the index.
|
||||||
// The runtime is O(log(N)*len(s)).
|
// The runtime is O(log(N)*len(s)).
|
||||||
func (x *Index) lookupAll(s []byte) []int {
|
func (x *Index) lookupAll(s []byte) []int32 {
|
||||||
// find matching suffix index range [i:j]
|
// find matching suffix index range [i:j]
|
||||||
// find the first index where s would be the prefix
|
// find the first index where s would be the prefix
|
||||||
i := sort.Search(len(x.sa), func(i int) bool { return bytes.Compare(x.at(i), s) >= 0 })
|
i := sort.Search(len(x.sa), func(i int) bool { return bytes.Compare(x.at(i), s) >= 0 })
|
||||||
@ -65,12 +110,15 @@ func (x *Index) lookupAll(s []byte) []int {
|
|||||||
func (x *Index) Lookup(s []byte, n int) (result []int) {
|
func (x *Index) Lookup(s []byte, n int) (result []int) {
|
||||||
if len(s) > 0 && n != 0 {
|
if len(s) > 0 && n != 0 {
|
||||||
matches := x.lookupAll(s)
|
matches := x.lookupAll(s)
|
||||||
if len(matches) < n || n < 0 {
|
if n < 0 || len(matches) < n {
|
||||||
n = len(matches)
|
n = len(matches)
|
||||||
}
|
}
|
||||||
|
// 0 <= n <= len(matches)
|
||||||
if n > 0 {
|
if n > 0 {
|
||||||
result = make([]int, n)
|
result = make([]int, n)
|
||||||
copy(result, matches)
|
for i, x := range matches[0:n] {
|
||||||
|
result[i] = int(x)
|
||||||
|
}
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
return
|
return
|
||||||
|
@ -213,7 +213,33 @@ func (a *index) at(i int) []byte { return a.data[a.sa[i]:] }
|
|||||||
|
|
||||||
func testConstruction(t *testing.T, tc *testCase, x *Index) {
|
func testConstruction(t *testing.T, tc *testCase, x *Index) {
|
||||||
if !sort.IsSorted((*index)(x)) {
|
if !sort.IsSorted((*index)(x)) {
|
||||||
t.Errorf("testConstruction failed %s", tc.name)
|
t.Errorf("failed testConstruction %s", tc.name)
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
func equal(x, y *Index) bool {
|
||||||
|
if !bytes.Equal(x.data, y.data) {
|
||||||
|
return false
|
||||||
|
}
|
||||||
|
for i, j := range x.sa {
|
||||||
|
if j != y.sa[i] {
|
||||||
|
return false
|
||||||
|
}
|
||||||
|
}
|
||||||
|
return true
|
||||||
|
}
|
||||||
|
|
||||||
|
func testSaveRestore(t *testing.T, tc *testCase, x *Index) {
|
||||||
|
var buf bytes.Buffer
|
||||||
|
if err := x.Write(&buf); err != nil {
|
||||||
|
t.Errorf("failed writing index %s (%s)", tc.name, err)
|
||||||
|
}
|
||||||
|
var y Index
|
||||||
|
if err := y.Read(&buf); err != nil {
|
||||||
|
t.Errorf("failed reading index %s (%s)", tc.name, err)
|
||||||
|
}
|
||||||
|
if !equal(x, &y) {
|
||||||
|
t.Errorf("restored index doesn't match saved index %s", tc.name)
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
||||||
@ -221,6 +247,7 @@ func TestIndex(t *testing.T) {
|
|||||||
for _, tc := range testCases {
|
for _, tc := range testCases {
|
||||||
x := New([]byte(tc.source))
|
x := New([]byte(tc.source))
|
||||||
testConstruction(t, &tc, x)
|
testConstruction(t, &tc, x)
|
||||||
|
testSaveRestore(t, &tc, x)
|
||||||
testLookups(t, &tc, x, 0)
|
testLookups(t, &tc, x, 0)
|
||||||
testLookups(t, &tc, x, 1)
|
testLookups(t, &tc, x, 1)
|
||||||
testLookups(t, &tc, x, 10)
|
testLookups(t, &tc, x, 10)
|
||||||
|
Loading…
Reference in New Issue
Block a user