First, we need to have a comparison object. In view of the limited computer disk space, this article compares the differences in file reading at the three levels of KB, MB, and GB.
package main
import (
"bufio"
"math/rand"
"os"
"time"
)
const charset = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"
var seededRand = rand.New(rand.NewSource(time.Now().UnixNano()))
func StringWithCharset(length int) string {
b := make([]byte, length)
for i := range b {
b[i] = charset[seededRand.Intn(len(charset))]
}
return string(b)
}
func main() {
files := map[string]int{"4KB.txt": 4, "4MB.txt": 4096, "4GB.txt": 4194304, "16GB.txt": 16777216}
for name, number := range files {
file, err := os.OpenFile(name, os.O_WRONLY|os.O_CREATE, 0666)
if err != nil {
panic(err)
}
write := bufio.NewWriter(file)
for i := 0; i < number; i++ {
s := StringWithCharset(1023) + "\n"
write.WriteString(s)
}
file.Close()
}
}Executing the above code, we get files of 4KB, 4MB, 4GB, and 16GB in sequence, which are composed of the contents of random strings of 1KB per line.
$ ls -alh 4kb.txt 4MB.txt 4GB.txt 16GB.txt -rw-r--r-- 1 slp staff 16G Mar 6 15:57 16GB.txt -rw-r--r-- 1 slp staff 4.0G Mar 6 15:54 4GB.txt -rw-r--r-- 1 slp staff 4.0M Mar 6 15:53 4MB.txt -rw-r--r-- 1 slp staff 4.0K Mar 6 15:16 4kb.txt
Next, we use different ways to read the contents of these files.
Go provides methods to read the file contents at once: os.ReadFile and ioutil.ReadFile. Starting in Go 1.16, ioutil.ReadFile is equivalent to os.ReadFile.
func BenchmarkOsReadFile4KB(b *testing.B) {
for i := 0; i < b.N; i++ {
_, err := os.ReadFile("./4KB.txt")
if err != nil {
b.Fatal(err)
}
}
}
func BenchmarkOsReadFile4MB(b *testing.B) {
for i := 0; i < b.N; i++ {
_, err := os.ReadFile("./4MB.txt")
if err != nil {
b.Fatal(err)
}
}
}
func BenchmarkOsReadFile4GB(b *testing.B) {
for i := 0; i < b.N; i++ {
_, err := os.ReadFile("./4GB.txt")
if err != nil {
b.Fatal(err)
}
}
}
func BenchmarkOsReadFile16GB(b *testing.B) {
for i := 0; i < b.N; i++ {
_, err := os.ReadFile("./16GB.txt")
if err != nil {
b.Fatal(err)
}
}
}The advantages and disadvantages of one-time loading of files are very obvious. It can reduce the number of IOs, but it will load all the file contents into the memory. For large files, there is a possibility of memory overflow. risk.
在很多情况下，例如日志分析，对文件的处理都是按行进行的。Go 中 bufio.Reader 对象提供了一个 ReadLine() 方法，但其实我们更多地是使用 ReadBytes('\n') 或者 ReadString('\n') 代替。
// ReadLine is a low-level line-reading primitive. Most callers should use // ReadBytes('\n') or ReadString('\n') instead or use a Scanner.
我们以 ReadString('\n') 为例，对 4 个文件分别进行逐行读取
func ReadLines(filename string) {
fi, err := os.Open(filename)
if err != nil{
panic(err)
}
defer fi.Close()
reader := bufio.NewReader(fi)
for {
_, err = reader.ReadString('\n')
if err != nil {
if err == io.EOF {
break
}
panic(err)
}
}
}
func BenchmarkReadLines4KB(b *testing.B) {
for i := 0; i < b.N; i++ {
ReadLines("./4KB.txt")
}
}
func BenchmarkReadLines4MB(b *testing.B) {
for i := 0; i < b.N; i++ {
ReadLines("./4MB.txt")
}
}
func BenchmarkReadLines4GB(b *testing.B) {
for i := 0; i < b.N; i++ {
ReadLines("./4GB.txt")
}
}
func BenchmarkReadLines16GB(b *testing.B) {
for i := 0; i < b.N; i++ {
ReadLines("./16GB.txt")
}
}块读取也称为分片读取，这也很好理解，我们可以将内容分成一块块的，每次读取指定大小的块内容。这里，我们将块大小设置为 4KB。
func ReadChunk(filename string) {
f, err := os.Open(filename)
if err != nil {
panic(err)
}
defer f.Close()
buf := make([]byte, 4*1024)
r := bufio.NewReader(f)
for {
_, err = r.Read(buf)
if err != nil {
if err == io.EOF {
break
}
panic(err)
}
}
}
func BenchmarkReadChunk4KB(b *testing.B) {
for i := 0; i < b.N; i++ {
ReadChunk("./4KB.txt")
}
}
func BenchmarkReadChunk4MB(b *testing.B) {
for i := 0; i < b.N; i++ {
ReadChunk("./4MB.txt")
}
}
func BenchmarkReadChunk4GB(b *testing.B) {
for i := 0; i < b.N; i++ {
ReadChunk("./4GB.txt")
}
}
func BenchmarkReadChunk16GB(b *testing.B) {
for i := 0; i < b.N; i++ {
ReadChunk("./16GB.txt")
}
}汇总结果
BenchmarkOsReadFile4KB-8 92877 12491 ns/op BenchmarkOsReadFile4MB-8 1620 744460 ns/op BenchmarkOsReadFile4GB-8 1 7518057733 ns/op signal: killed BenchmarkReadLines4KB-8 90846 13184 ns/op BenchmarkReadLines4MB-8 493 2338170 ns/op BenchmarkReadLines4GB-8 1 3072629047 ns/op BenchmarkReadLines16GB-8 1 12472749187 ns/op BenchmarkReadChunk4KB-8 99848 12262 ns/op BenchmarkReadChunk4MB-8 913 1233216 ns/op BenchmarkReadChunk4GB-8 1 2095515009 ns/op BenchmarkReadChunk16GB-8 1 8547054349 ns/op
在本文的测试条件下（每行数据 1KB），对于小对象 4KB 的读取，三种方式差距并不大；在 MB 级别的读取中，直接加载最快，但块读取也慢不了多少；上了 GB 后，块读取方式会最快。
且有一点可以注意到的是，在整个文件加载的方式中，对于 16 GB 的文件数据（测试机器运行内存为 8GB），会内存耗尽出错，没法执行。
不管是什么大小的文件，均不推荐整个文件加载的方式，因为它在小文件时的速度优势并没有那么大，相较于安全隐患，不值得选择它。
块读取是优先选择，尤其对于一些没有换行符的文件，例如音视频等。通过设定合适的块读取大小，能让速度和内存得到很好的平衡。且在读取过程中，往往伴随着处理内容的逻辑。每块内容可以赋给一个工作 goroutine 来处理，能更好地并发。
------------------- End -------------- -----
Structure of the basics of Go language (Winter)
An article will take you to understand the basic map of Go language
#Welcome everyonelike,leave a message,forward,reprint,Thank you all Companion and support
If you want to join the Go study group, please reply in the background [ Join the group]
Love is always the same across thousands of rivers and mountains, can you click [在看]
The above is the detailed content of How to choose a Go file reading solution. For more information, please follow other related articles on the PHP Chinese website!
![[Web front-end] Node.js quick start](https://img.php.cn/upload/course/000/000/067/662b5d34ba7c0227.png)
