Dan Willemsen | ebae302 | 2017-01-13 23:01:08 -0800 | [diff] [blame] | 1 | // Copyright 2016 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 bytes |
| 6 | |
| 7 | //go:noescape |
| 8 | |
| 9 | // indexShortStr returns the index of the first instance of sep in s, |
| 10 | // or -1 if sep is not present in s. |
| 11 | // indexShortStr requires 2 <= len(sep) <= shortStringLen |
| 12 | func indexShortStr(s, c []byte) int // ../runtime/asm_s390x.s |
| 13 | |
| 14 | // supportsVX reports whether the vector facility is available. |
| 15 | // indexShortStr must not be called if the vector facility is not |
| 16 | // available. |
| 17 | func supportsVX() bool // ../runtime/asm_s390x.s |
| 18 | |
| 19 | var shortStringLen = -1 |
| 20 | |
| 21 | func init() { |
| 22 | if supportsVX() { |
| 23 | shortStringLen = 64 |
| 24 | } |
| 25 | } |
| 26 | |
| 27 | // Index returns the index of the first instance of sep in s, or -1 if sep is not present in s. |
| 28 | func Index(s, sep []byte) int { |
| 29 | n := len(sep) |
| 30 | switch { |
| 31 | case n == 0: |
| 32 | return 0 |
| 33 | case n == 1: |
| 34 | return IndexByte(s, sep[0]) |
| 35 | case n == len(s): |
| 36 | if Equal(sep, s) { |
| 37 | return 0 |
| 38 | } |
| 39 | return -1 |
| 40 | case n > len(s): |
| 41 | return -1 |
| 42 | case n <= shortStringLen: |
| 43 | // Use brute force when s and sep both are small |
| 44 | if len(s) <= 64 { |
| 45 | return indexShortStr(s, sep) |
| 46 | } |
| 47 | c := sep[0] |
| 48 | i := 0 |
| 49 | t := s[:len(s)-n+1] |
| 50 | fails := 0 |
| 51 | for i < len(t) { |
| 52 | if t[i] != c { |
| 53 | // IndexByte skips 16/32 bytes per iteration, |
| 54 | // so it's faster than indexShortStr. |
| 55 | o := IndexByte(t[i:], c) |
| 56 | if o < 0 { |
| 57 | return -1 |
| 58 | } |
| 59 | i += o |
| 60 | } |
| 61 | if Equal(s[i:i+n], sep) { |
| 62 | return i |
| 63 | } |
| 64 | fails++ |
| 65 | i++ |
| 66 | // Switch to indexShortStr when IndexByte produces too many false positives. |
| 67 | // Too many means more that 1 error per 8 characters. |
| 68 | // Allow some errors in the beginning. |
| 69 | if fails > (i+16)/8 { |
| 70 | r := indexShortStr(s[i:], sep) |
| 71 | if r >= 0 { |
| 72 | return r + i |
| 73 | } |
| 74 | return -1 |
| 75 | } |
| 76 | } |
| 77 | return -1 |
| 78 | } |
Dan Willemsen | a322328 | 2018-02-27 19:41:43 -0800 | [diff] [blame^] | 79 | return indexRabinKarp(s, sep) |
Dan Willemsen | ebae302 | 2017-01-13 23:01:08 -0800 | [diff] [blame] | 80 | } |
| 81 | |
Dan Willemsen | d279748 | 2017-07-26 13:13:13 -0700 | [diff] [blame] | 82 | // Count counts the number of non-overlapping instances of sep in s. |
Dan Willemsen | a322328 | 2018-02-27 19:41:43 -0800 | [diff] [blame^] | 83 | // If sep is an empty slice, Count returns 1 + the number of UTF-8-encoded code points in s. |
Dan Willemsen | d279748 | 2017-07-26 13:13:13 -0700 | [diff] [blame] | 84 | func Count(s, sep []byte) int { |
| 85 | return countGeneric(s, sep) |
| 86 | } |