为什么这个程序这么慢？我认为代码已经相当优化，但在我的根文件系统上使用时，它比 find 命令花费的时间要长得多。
它大约需要 4 分钟，而 find 命令大约需要 40 秒。
我尝试删除排序算法，但并没有加快程序速度。
package main
import (
"fmt"
"io"
"io/fs"
"log"
"os"
"sort"
"sync"
"github.com/google/fscrypt/filesystem"
"github.com/sirupsen/logrus"
"gopkg.in/alecthomas/kingpin.v2"
)
var (
mountpoint = kingpin.Flag("mount", "The mount to find the largest file usages. Can be a subath of mount").Required().String()
limit = kingpin.Flag("limit", "The maximum number of files return to the display").Default("10").Short('l').Int()
)
var device string
type fileDisplay struct {
Size int64
Path string
}
type bySize []fileDisplay
func (a bySize) Len() int { return len(a) }
func (a bySize) Less(i, j int) bool { return a[i].Size < a[j].Size }
func (a bySize) Swap(i, j int) { a[i], a[j] = a[j], a[i] }
var fileChan = make(chan fileDisplay)
var files []fileDisplay
func main() {
log.SetOutput(io.Discard)
kingpin.Version("0.0.1")
kingpin.Parse()
//Define limit after parsing
logrus.SetLevel(logrus.FatalLevel)
if (*mountpoint)[len(*mountpoint)-1:] != "/" {
*mountpoint = *mountpoint + "/"
}
fmt.Println("Finding the top", *limit, "largest files on filesystem", *mountpoint, "\n================================================")
mount, err := filesystem.FindMount(*mountpoint)
if err != nil {
logrus.Fatal(err)
}
device = mount.Device
entries, err := os.ReadDir(*mountpoint)
if err != nil {
logrus.Fatal(err)
}
var wg sync.WaitGroup
getFiles(*mountpoint, entries, &wg)
go func() {
defer close(fileChan)
wg.Wait()
}()
var last int64
for file := range fileChan {
if file.Size > last {
files = append(files, file)
} else {
files = append([]fileDisplay{file}, files...)
}
}
sort.Sort(bySize(files))
var shortFiles []fileDisplay
if len(files) > *limit {
shortFiles = files[len(files)-*limit:]
} else {
shortFiles = files
}
for _, file := range shortFiles {
fmt.Println(file.Path, file.DisplaySizeIEC())
}
}
func getFiles(start string, entries []fs.DirEntry, wg *sync.WaitGroup) {
for _, entry := range entries {
wg.Add(1)
go handleEntry(start, entry, wg)
}
}
func handleEntry(start string, entry fs.DirEntry, wg *sync.WaitGroup) {
defer wg.Done()
var file fileDisplay
mount, err := filesystem.FindMount(start + entry.Name())
if err != nil {
logrus.Fatalln(err, start+entry.Name())
return
}
if mount.Device == device {
if entry.Type().IsRegular() {
fileInfo, err := os.Stat(start + entry.Name())
if err != nil {
logrus.Fatalln(err, start+entry.Name())
return
}
file.Path = start + entry.Name()
file.Size = fileInfo.Size()
fileChan <- file
} else if entry.IsDir() {
entries, err := os.ReadDir(start + entry.Name())
if err != nil {
logrus.Fatalln(err, start+entry.Name())
return
}
logrus.Info("Searching ", start+entry.Name())
getFiles(start+entry.Name()+"/", entries, wg)
}
}
}
func (f *fileDisplay) DisplaySizeIEC() string {
const unit = 1024
b := f.Size
if b < unit {
return fmt.Sprintf("%dB", b)
}
div, exp := int64(unit), 0
for n := b / unit; n >= unit; n /= unit {
div *= unit
exp++
}
return fmt.Sprintf("%.2f%ciB",
float64(b)/float64(div), "KMGTPE"[exp])
}
编辑：我尝试删除通道并仅附加到切片。这加快了速度，但并不安全，因为多个例程可能正在访问它。
我的最终草案涉及删除通道并使用sync.RWMutex锁定列表和自定义附加函数来附加锁定。这使我能够放弃通道并使用附加，而无需冒多个例程编辑同一切片的风险。
我放弃了该通道，因为这会导致例程保持打开状态，直到打开通道上的 for 循环可以到达它们的消息。我的通道操作被阻塞。因此，例程导致它减慢了 for 循环在通道上迭代的速度。
您可以在这里看到差异：
package main
import (
"fmt"
"io"
"io/fs"
"log"
"os"
"sort"
"sync"
"github.com/google/fscrypt/filesystem"
"github.com/sirupsen/logrus"
"gopkg.in/alecthomas/kingpin.v2"
)
var (
mountpoint = kingpin.Flag("mount", "The mount to find the largest file usages. Can be a subath of mount").Required().String()
limit = kingpin.Flag("limit", "The maximum number of files return to the display").Default("10").Short('l').Int()
)
var device string
type fileDisplays struct {
sync.RWMutex
Files []fileDisplay
}
var files fileDisplays
type fileDisplay struct {
Size int64
Path string
}
type bySize []fileDisplay
func (a bySize) Len() int { return len(a) }
func (a bySize) Less(i, j int) bool { return a[i].Size < a[j].Size }
func (a bySize) Swap(i, j int) { a[i], a[j] = a[j], a[i] }
func main() {
log.SetOutput(io.Discard)
kingpin.Version("0.0.1")
kingpin.Parse()
//Define limit after parsing
logrus.SetLevel(logrus.FatalLevel)
if (*mountpoint)[len(*mountpoint)-1:] != "/" {
*mountpoint = *mountpoint + "/"
}
fmt.Println("Finding the top", *limit, "largest files on filesystem", *mountpoint, "\n================================================")
mount, err := filesystem.FindMount(*mountpoint)
if err != nil {
logrus.Fatal(err)
}
device = mount.Device
entries, err := os.ReadDir(*mountpoint)
if err != nil {
logrus.Fatal(err)
}
var wg sync.WaitGroup
getFiles(*mountpoint, entries, &wg)
wg.Wait()
sort.Sort(bySize(files.Files))
var shortFiles []fileDisplay
if len(files.Files) > *limit {
shortFiles = files.Files[len(files.Files)-*limit:]
} else {
shortFiles = files.Files
}
for _, file := range shortFiles {
fmt.Println(file.Path, file.DisplaySizeIEC())
}
}
func getFiles(start string, entries []fs.DirEntry, wg *sync.WaitGroup) {
for _, entry := range entries {
wg.Add(1)
go handleEntry(start, entry, wg)
}
}
func handleEntry(start string, entry fs.DirEntry, wg *sync.WaitGroup) {
defer wg.Done()
var file fileDisplay
mount, err := filesystem.FindMount(start + entry.Name())
if err != nil {
logrus.Errorln(err, start+entry.Name())
return
}
if mount.Device == device {
if entry.Type().IsRegular() {
fileInfo, err := os.Stat(start + entry.Name())
if err != nil {
logrus.Errorln(err, start+entry.Name())
return
}
file.Path = start + entry.Name()
file.Size = fileInfo.Size()
files.Append(file)
} else if entry.IsDir() {
entries, err := os.ReadDir(start + entry.Name())
if err != nil {
logrus.Errorln(err, start+entry.Name())
return
}
logrus.Info("Searching ", start+entry.Name())
getFiles(start+entry.Name()+"/", entries, wg)
}
}
}
func (f *fileDisplay) DisplaySizeIEC() string {
const unit = 1024
b := f.Size
if b < unit {
return fmt.Sprintf("%dB", b)
}
div, exp := int64(unit), 0
for n := b / unit; n >= unit; n /= unit {
div *= unit
exp++
}
return fmt.Sprintf("%.2f%ciB",
float64(b)/float64(div), "KMGTPE"[exp])
}
func (fd *fileDisplays) Append(item fileDisplay) {
fd.Lock()
defer fd.Unlock()
fd.Files = append(fd.Files, item)
}