Update dependencies (#5518)

This commit is contained in:
hongming
2023-02-12 23:09:20 +08:00
committed by GitHub
parent d3b35fb2da
commit a979342f56
1486 changed files with 126660 additions and 71128 deletions

View File

@@ -6,12 +6,17 @@ package levenshtein
import "unicode/utf8"
// minLengthThreshold is the length of the string beyond which
// an allocation will be made. Strings smaller than this will be
// zero alloc.
const minLengthThreshold = 32
// ComputeDistance computes the levenshtein distance between the two
// strings passed as an argument. The return value is the levenshtein distance
//
// Works on runes (Unicode code points) but does not normalize
// the input strings. See https://blog.golang.org/normalization
// and the golang.org/x/text/unicode/norm pacage.
// and the golang.org/x/text/unicode/norm package.
func ComputeDistance(a, b string) int {
if len(a) == 0 {
return utf8.RuneCountInString(b)
@@ -25,12 +30,10 @@ func ComputeDistance(a, b string) int {
return 0
}
// We need to convert to []rune if the strings are non-ascii.
// We need to convert to []rune if the strings are non-ASCII.
// This could be avoided by using utf8.RuneCountInString
// and then doing some juggling with rune indices.
// The primary challenge is keeping track of the previous rune.
// With a range loop, its not that easy. And with a for-loop
// we need to keep track of the inter-rune width using utf8.DecodeRuneInString
// and then doing some juggling with rune indices,
// but leads to far more bounds checks. It is a reasonable trade-off.
s1 := []rune(a)
s2 := []rune(b)
@@ -41,22 +44,33 @@ func ComputeDistance(a, b string) int {
lenS1 := len(s1)
lenS2 := len(s2)
// init the row
x := make([]int, lenS1+1)
for i := 0; i <= lenS1; i++ {
x[i] = i
// Init the row.
var x []uint16
if lenS1+1 > minLengthThreshold {
x = make([]uint16, lenS1+1)
} else {
// We make a small optimization here for small strings.
// Because a slice of constant length is effectively an array,
// it does not allocate. So we can re-slice it to the right length
// as long as it is below a desired threshold.
x = make([]uint16, minLengthThreshold)
x = x[:lenS1+1]
}
// we start from 1 because index 0 is already 0.
for i := 1; i < len(x); i++ {
x[i] = uint16(i)
}
// make a dummy bounds check to prevent the 2 bounds check down below.
// The one inside the loop is particularly costly.
_ = x[lenS1]
// fill in the rest
for i := 1; i <= lenS2; i++ {
prev := i
var current int
prev := uint16(i)
for j := 1; j <= lenS1; j++ {
if s2[i-1] == s1[j-1] {
current = x[j-1] // match
} else {
current := x[j-1] // match
if s2[i-1] != s1[j-1] {
current = min(min(x[j-1]+1, prev+1), x[j]+1)
}
x[j-1] = prev
@@ -64,10 +78,10 @@ func ComputeDistance(a, b string) int {
}
x[lenS1] = prev
}
return x[lenS1]
return int(x[lenS1])
}
func min(a, b int) int {
func min(a, b uint16) uint16 {
if a < b {
return a
}