strings.Contains() 比 Python3 慢 2 倍？

Question

我正在将文本模式扫描仪从 Python3 转换为 Go1.10，但令我惊讶的是它实际上慢了 2 倍。根据分析，罪魁祸首在 strings.Contains()。请参阅下面的简单基准。我错过了什么吗？你能推荐一种在这种情况下表现更好的更快的 Go 模式搜索算法吗？我不在乎启动时间，相同的模式将用于扫描数百万个文件。

Py3 基准测试：

import time
import re

RUNS = 10000

if __name__ == '__main__':
    with open('data.php') as fh:
        testString = fh.read()

    def do():
        return "576ad4f370014dfb1d0f17b0e6855f22" in testString

    start = time.time()
    for i in range(RUNS):
        _ = do()
    duration = time.time() - start
    print("Python: %.2fs" % duration)

Go1.10 基准测试：

package main

import (
    "fmt"
    "io/ioutil"
    "log"
    "strings"
    "time"
)

const (
    runs = 10000
)

func main() {
    fname := "data.php"
    testdata := readFile(fname)
    needle := "576ad4f370014dfb1d0f17b0e6855f22"
    start := time.Now()

    for i := 0; i < runs; i++ {
        _ = strings.Contains(testdata, needle)

    }
    duration := time.Now().Sub(start)
    fmt.Printf("Go: %.2fs\n", duration.Seconds())
}

func readFile(fname string) string {
    data, err := ioutil.ReadFile(fname)
    if err != nil {
        log.Fatal(err)
    }
    return string(data)
}

data.php是一个528KB的文件，可以found here.

输出：

Go:     1.98s
Python: 0.84s

Answer 1

为什么 Python 3 (24.79s) 比 Go (5.47s) 慢 4.5 倍？你得到了什么结果？

Python:

$ cat contains.py
import time
import re

RUNS = 10000

if __name__ == '__main__':
    # The Complete Works of William Shakespeare by William Shakespeare
    # http://www.gutenberg.org/files/100/100-0.txt
    file = '/home/peter/shakespeare.100-0.txt' # 'data.php'
    with open(file) as fh:
        testString = fh.read()

    def do():
        return "Means to immure herself and not be seen." in testString

    start = time.time()
    for i in range(RUNS):
        _ = do()
    duration = time.time() - start
    print("Python: %.2fs" % duration)
    print(do())
$ python3 --version
Python 3.6.5
$ python3 contains.py
Python: 24.79s
True
$

去:

$ cat contains.go
package main

import (
    "fmt"
    "io/ioutil"
    "log"
    "strings"
    "time"
)

const (
    runs = 10000
)

func main() {
    // The Complete Works of William Shakespeare by William Shakespeare
    // http://www.gutenberg.org/files/100/100-0.txt
    fname := `/home/peter/shakespeare.100-0.txt` // "data.php"
    testdata := readFile(fname)
    needle := "Means to immure herself and not be seen."
    start := time.Now()

    for i := 0; i < runs; i++ {
        _ = strings.Contains(testdata, needle)

    }
    duration := time.Now().Sub(start)
    fmt.Printf("Go: %.2fs\n", duration.Seconds())

    fmt.Println(strings.Contains(testdata, needle))
    fmt.Println(strings.Index(testdata, needle))

}

func readFile(fname string) string {
    data, err := ioutil.ReadFile(fname)
    if err != nil {
        log.Fatal(err)
    }
    return string(data)
}
$ go version
go version devel +5332b5e75a Tue Jul 31 15:44:37 2018 +0000 linux/amd64
$ go run contains.go
Go: 5.47s
true
5837178
$

Answer 2

我对在 Wikipedia 上发现的各种字符串搜索实现进行了更多基准测试，例如：

https://github.com/cloudflare/ahocorasick
https://github.com/cubicdaiya/bms
https://github.com/kkdai/kmp
https://github.com/paddie/gokmp
https://github.com/hillu/go-yara（Yara 似乎在底层实现了 Aho & Corasick）。

基准测试结果（code here）：

BenchmarkStringsContains-4         10000        208055 ns/op
BenchmarkBMSSearch-4                1000       1856732 ns/op
BenchmarkPaddieKMP-4                2000       1069495 ns/op
BenchmarkKkdaiKMP-4                 1000       1440147 ns/op
BenchmarkAhocorasick-4              2000        935885 ns/op
BenchmarkYara-4                     1000       1237887 ns/op

然后，我针对原生（strings.Contains 和 regexp）针对 500KB 不匹配文件测试了大约 1100 个签名（100 个正则表达式，1000 个文字）的实际用例进行了基准测试和基于 C 的 Yara 实现：

BenchmarkScanNative-4              2     824328504 ns/op
BenchmarkScanYara-4              300       5338861 ns/op

尽管 Go 中的 C 调用据说很昂贵，但在这些 "heavy" 操作中，利润是可观的。侧面观察：Yara 只需要 CPU 5 倍的时间来匹配 1100 个签名而不是 1 个。

strings.Contains() 比 Python3 慢 2 倍？

Go strings.Contains() 2x slower than Python3?

algorithm

performance

go

python-3.x

yara