I am trying to solve leetcode problem permutations.
But when i test with -benchmem, i found it allocs too much which reach 1957 allocs/op when permute([]int{1,2,3,4,5,6})
I found it escape to heap when generating sub-nums target. Even i try to allocate [6]int, and use unsafe package to build the slice, it still moved to heap.
My question is, why the slice escape to heap, and how could i allocate the slice on stack?
Here's my code:
package main
import (
"fmt"
"reflect"
"unsafe"
)
func permute(nums []int) [][]int {
resLen := 1
for i := 1; i<= len(nums);i ++{
resLen *= i
}
// pre allocate
res := make([][]int, resLen)
for i := range res{
res[i] = make([]int, 0, len(nums))
}
build(res, nums)
return res
}
func build(res [][]int,targets []int){
step := len(res) / len(targets)
for i := range targets{
for j := i*step; j < (i+1) * step; j ++{
res[j] = append(res[j], targets[i])
}
if len(targets) != 1{
var ab = [6]int{}
var buff []int
var bp *reflect.SliceHeader
bp = (*reflect.SliceHeader)(unsafe.Pointer(&buff))
bp.Data = uintptr(unsafe.Pointer(&ab))
bp.Cap = 6
buff = append(buff, targets[:i]...)
buff = append(buff, targets[i+1:]...)
build(res[i*step:(i+1)*step], buff)
}
}
return
}
func main() {
nums := []int{1,2,3}
res := permute(nums)
fmt.Println(res)
}
build function without unsafe but escapes to heap:
func build(res [][]int, targets []int) {
step := len(res) / len(targets)
for i := range targets {
for j := i * step; j < (i+1)*step; j++ {
res[j] = append(res[j], targets[i])
}
if len(targets) != 1 {
buff := make([]int, 0, 6) // make([]int, 0, 6) escapes to heap
buff = append(buff, targets[:i]...)
buff = append(buff, targets[i+1:]...)
build(res[i*step:(i+1)*step], buff)
}
}
return
}
And my test case:
package main
import "testing"
func Benchmark(b *testing.B){
for i:=0;i<b.N;i++{
permute([]int{1,2,3,4,5,6})
}
}
When i run go build -gcflags="-m", it reports ./main.go:32:8: moved to heap: ab
Trying to subvert the compiler using unsafe.Pointer is only making it harder for the escape analysis to do its job, preventing the slice from being stack allocated. Simply allocate a single slice and reuse it for each loop iteration:
func build(res [][]int, targets []int) {
buff := make([]int, 0, 6)
step := len(res) / len(targets)
for i := range targets {
buff = buff[:0]
for j := i * step; j < (i+1)*step; j++ {
res[j] = append(res[j], targets[i])
}
if len(targets) != 1 {
buff = append(buff, targets[:i]...)
buff = append(buff, targets[i+1:]...)
build(res[i*step:(i+1)*step], buff)
}
}
return
}
This can be correctly optimized by the compiler
./main.go:26:17: make([]int, 0, 6) does not escape
And will result in only the desired allocations:
Benchmark-8 44607 26838 ns/op 52992 B/op 721 allocs/op
Related
I'm benchmarking unmarshaling from string to int and uint with this code:
package main
import (
"strconv"
"testing"
)
func BenchmarkUnmarshalInt(b *testing.B) {
for i := 0; i < b.N; i++ {
UnmarshalInt("123456")
}
}
func BenchmarkUnmarshalUint(b *testing.B) {
for i := 0; i < b.N; i++ {
UnmarshalUint("123456")
}
}
func UnmarshalInt(v string) int {
i, _ := strconv.Atoi(v)
return i
}
func UnmarshalUint(v string) uint {
i, _ := strconv.ParseUint(v, 10, 64)
return uint(i)
}
Result:
Running tool: C:\Go\bin\go.exe test -benchmem -run=^$ myBench/main -bench .
goos: windows
goarch: amd64
pkg: myBench/main
BenchmarkUnmarshalInt-8 99994166 11.7 ns/op 0 B/op 0 allocs/op
BenchmarkUnmarshalUint-8 54550413 21.0 ns/op 0 B/op 0 allocs/op
Is it possible that the second (uint) is almost twice as slow as the first (int)?
Yes, it's possible. strconv.Atoi has a fast path when the input string length is less than 19 (or 10 if int is 32 bit). This allows it to be a lot faster because it doesn't need to check for overflow.
If you change your test number to "1234567890123456789" (assuming 64 bit int), then your int benchmark is slightly slower than the uint benchmark because the fast path can't be used. On my machine, it takes 37.6 ns/op for the signed version vs 31.5 ns/op for the unsigned version.
Here's the modified benchmark code (note I added a variable that sums up the parsed results, just in case the compiler got clever and optimized it away).
package main
import (
"fmt"
"strconv"
"testing"
)
const X = "1234567890123456789"
func BenchmarkUnmarshalInt(b *testing.B) {
var T int
for i := 0; i < b.N; i++ {
T += UnmarshalInt(X)
}
fmt.Println(T)
}
func BenchmarkUnmarshalUint(b *testing.B) {
var T uint
for i := 0; i < b.N; i++ {
T += UnmarshalUint(X)
}
fmt.Println(T)
}
func UnmarshalInt(v string) int {
i, _ := strconv.Atoi(v)
return i
}
func UnmarshalUint(v string) uint {
i, _ := strconv.ParseUint(v, 10, 64)
return uint(i)
}
For reference, the code for strconv.Atoi in the standard library is currently as follows:
func Atoi(s string) (int, error) {
const fnAtoi = "Atoi"
sLen := len(s)
if intSize == 32 && (0 < sLen && sLen < 10) ||
intSize == 64 && (0 < sLen && sLen < 19) {
// Fast path for small integers that fit int type.
s0 := s
if s[0] == '-' || s[0] == '+' {
s = s[1:]
if len(s) < 1 {
return 0, &NumError{fnAtoi, s0, ErrSyntax}
}
}
n := 0
for _, ch := range []byte(s) {
ch -= '0'
if ch > 9 {
return 0, &NumError{fnAtoi, s0, ErrSyntax}
}
n = n*10 + int(ch)
}
if s0[0] == '-' {
n = -n
}
return n, nil
}
// Slow path for invalid, big, or underscored integers.
i64, err := ParseInt(s, 10, 0)
if nerr, ok := err.(*NumError); ok {
nerr.Func = fnAtoi
}
return int(i64), err
}
I tested simple string concatenation in Go with “+” and bytes.Buffer (both “WriteString” and “Write(bytes)”. The result shows that “+” is much slower than the other two, which makes sense.
However, when I use the three ways to implement Fibonacci-like string concatenation (i.e. a, b, ab, bab, abbab, bababbab, abbabbababbab), “+” performs the best. The sample codes and the benchmarking results are shown as follows.
String “+”
func Fibonacci(n int) string {
FiboResult := ""
prev_result := "a"
next_result := "b"
if n == 1{
FiboResult = "a"
}else if n == 2 {
FiboResult = "b"
}else{
for i := 3; i <= n; i++ {
FiboResult = prev_result + next_result
prev_result = next_result
next_result = FiboResult
}
}
return FiboResult
}
bytes.Buffer (WriteString)
func Fibonacci(n int) bytes.Buffer {
var FiboResult bytes.Buffer
var prev_result bytes.Buffer
prev_result.WriteString("a")
var next_result bytes.Buffer
next_result.WriteString("b")
if n == 1{
FiboResult.WriteString("a")
}else if n == 2 {
FiboResult.WriteString("b")
}else{
for i := 3; i <= n; i++ {
FiboResult.Reset()
FiboResult.WriteString(prev_result.String())
FiboResult.WriteString(next_result.String())
prev_result.Reset()
prev_result.WriteString(next_result.String())
next_result.Reset()
next_result.WriteString(FiboResult.String())
}
}
return FiboResult
}
the benchmarking results
I believe it is the overhead of bytes.Buffer.String() that make this happen. But I could not figure out how to use bytes.Buffer correctly in this case. Or how could I modify my code to avoid the problem? Hints, sample codes, or explanations are all appreciated. Many thanks in advance!
In Go, use the testing package for benchmarks.
Write reasonably efficient Go functions. Don't perform unnecessary conversions. Minimize allocations and copies. And so on. Allow for non-ASCII characters, for example Chinese characters. Allow for strings of more than one character. Consider using a byte slice. For example,
func fibonacciN(n int) uint64 {
f := uint64(0)
a, b := uint64(0), uint64(1)
for i := 0; i < n; i++ {
f, a, b = a, b, a+b
if a > b {
break
}
}
return f
}
func Fibonacci(a, b string, n int) string {
if n < 0 {
n = 0
}
switch n {
case 0:
return ""
case 1:
return a
case 2:
return b
}
f := make([]byte, len(a)*int(fibonacciN(n-1))+len(b)*int(fibonacciN(n)))
ab := a + b
copy(f[len(f)-len(ab):], ab)
for i := 4; i <= n; i++ {
end := len(f) - (len(a)*int(fibonacciN(i-3)) + len(b)*int(fibonacciN(i-2)))
start := len(f) - (len(a)*int(fibonacciN(i-1)) + len(b)*int(fibonacciN(i)))
copy(f[start:end], f[end:])
}
return string(f)
}
Benchmark functions. For example, with n = 20,
$ go test fib_test.go -bench=. -benchmem
goos: linux
goarch: amd64
BenchmarkPeterSO-8 1000000 1851 ns/op 13568 B/op 2 allocs/op
BenchmarkPlus-8 500000 2493 ns/op 18832 B/op 18 allocs/op
BenchmarkBuffer-8 100000 12773 ns/op 90256 B/op 60 allocs/op
PASS
$
fib_test.go:
package main
import (
"bytes"
"testing"
)
var benchN = 20
func fibonacciN(n int) uint64 {
f := uint64(0)
a, b := uint64(0), uint64(1)
for i := 0; i < n; i++ {
f, a, b = a, b, a+b
if a > b {
break
}
}
return f
}
func FibonacciPeterSO(a, b string, n int) string {
if n < 0 {
n = 0
}
switch n {
case 0:
return ""
case 1:
return a
case 2:
return b
}
f := make([]byte, len(a)*int(fibonacciN(n-1))+len(b)*int(fibonacciN(n)))
ab := a + b
copy(f[len(f)-len(ab):], ab)
for i := 4; i <= n; i++ {
end := len(f) - (len(a)*int(fibonacciN(i-3)) + len(b)*int(fibonacciN(i-2)))
start := len(f) - (len(a)*int(fibonacciN(i-1)) + len(b)*int(fibonacciN(i)))
copy(f[start:end], f[end:])
}
return string(f)
}
func BenchmarkPeterSO(b *testing.B) {
for i := 0; i < b.N; i++ {
FibonacciPeterSO("a", "b", benchN)
}
}
func FibonacciPlus(n int) string {
FiboResult := ""
prev_result := "a"
next_result := "b"
if n == 1 {
FiboResult = "a"
} else if n == 2 {
FiboResult = "b"
} else {
for i := 3; i <= n; i++ {
FiboResult = prev_result + next_result
prev_result = next_result
next_result = FiboResult
}
}
return FiboResult
}
func BenchmarkPlus(b *testing.B) {
for i := 0; i < b.N; i++ {
FibonacciPlus(benchN)
}
}
func FibonacciBuffer(n int) bytes.Buffer {
var FiboResult bytes.Buffer
var prev_result bytes.Buffer
prev_result.WriteString("a")
var next_result bytes.Buffer
next_result.WriteString("b")
if n == 1 {
FiboResult.WriteString("a")
} else if n == 2 {
FiboResult.WriteString("b")
} else {
for i := 3; i <= n; i++ {
FiboResult.Reset()
FiboResult.WriteString(prev_result.String())
FiboResult.WriteString(next_result.String())
prev_result.Reset()
prev_result.WriteString(next_result.String())
next_result.Reset()
next_result.WriteString(FiboResult.String())
}
}
return FiboResult
}
func BenchmarkBuffer(b *testing.B) {
for i := 0; i < b.N; i++ {
FibonacciBuffer(benchN)
}
}
var testN = benchN
func TestPeterSO(t *testing.T) {
for n := 0; n <= testN; n++ {
got := FibonacciPeterSO("a", "b", n)
want := FibonacciPlus(n)
if want != got {
t.Errorf("want: %s got: %s", want, got)
}
}
}
bytes.Buffer (or the newer and faster strings.Builder) wins over simple + string concatenation if you want to append "many" values, and obtain the result once in the end, because intermediate allocations are not needed compared to using + multiple times.
And you are not using bytes.Buffer this way: you just write one string into it, and you obtain its content and you reset it. That's just a roundtrip which turns out to be an overhead.
The problem here is that generating the Fibonacci string you are seeking, that requires prepending text to the buffer, not appending to it. And bytes.Buffer only supports appending to it, so using it like this is not a good fit at all.
Generating reverse with bytes.Buffer
Note that a prepend operation is basically an append operation if you generate the reverse of a string. Which means if we first would generate the reverse of the result, we could use bytes.Buffer to perform an append when otherwise a prepend would be needed. Of course the appended string would have to also be the reverse of what otherwise would be prepended.
And of course when we're done, we have to reverse the result to get what we originally wanted.
Also note that when building result in an iterative way, the successive intermediate result is the concatenation of the previous and the one before that. So to obtain the nth result, we can simply append the substring of what we already have! This is a nice optimization.
Here's how it would look like:
func FibonacciReverseBuf(n int) string {
switch n {
case 0:
return ""
case 1:
return "a"
case 2:
return "b"
}
prev, prev2 := 1, 1
buf := bytes.NewBufferString("ba")
for i := 3; i < n; i++ {
buf.Write(buf.Bytes()[:buf.Len()-prev2])
prev2, prev = prev, prev+prev2
}
// Reverse
b := buf.Bytes()
for i, j := 0, len(b)-1; i < j; i, j = i+1, j-1 {
b[i], b[j] = b[j], b[i]
}
return string(b)
}
Generating reverse with []byte and append()
Also note that since we're only appending, we can just as easily use a []byte and use the builtin append() function:
func FibonacciReverse(n int) string {
switch n {
case 0:
return ""
case 1:
return "a"
case 2:
return "b"
}
prev, prev2 := 1, 1
b := []byte("ba")
for i := 3; i < n; i++ {
b = append(b, b[:len(b)-prev2]...)
prev2, prev = prev, prev+prev2
}
// Reverse
for i, j := 0, len(b)-1; i < j; i, j = i+1, j-1 {
b[i], b[j] = b[j], b[i]
}
return string(b)
}
Preallocating and using copy() in a single []byte
Still, using append() may cause reallocations, because we don't know how big the buffer (the result) will be. So we start with a small buffer, and append() will increase it as needed. Also append() requires slice value (slice header) assignments. And we also have to reverse the result.
A much faster solution would be to get rid of those cons.
First let's calculate how big the result will be (this is essentially calculating the Fibonacci numbers), and allocate the required byte slice in one step.
If we do so, we can do the "prepend" operations by copying parts of our buffer (which is a []byte) to specific positions. So no append(), no reallocations, no reversing.
This is how it will look like:
func Fibonacci(n int) string {
switch n {
case 0:
return ""
case 1:
return "a"
case 2:
return "b"
}
fibs := make([]int, n)
fibs[0], fibs[1] = 1, 1
for i := 2; i < n; i++ {
fibs[i] = fibs[i-1] + fibs[i-2]
}
l := fibs[n-1]
b := make([]byte, l)
b[l-2], b[l-1] = 'a', 'b'
for i := 3; i < n; i++ {
copy(b[l-fibs[i]:], b[l-fibs[i-2]:])
}
return string(b)
}
Testing the output
To test if the above functions give the result we expect them to give, we may use the following testing function:
func TestFibonacci(t *testing.T) {
cases := []struct {
n int
exp string
}{
{0, ""},
{1, "a"},
{2, "b"},
{3, "ab"},
{4, "bab"},
{5, "abbab"},
{6, "bababbab"},
{7, "abbabbababbab"},
}
funcs := []struct {
name string
f func(int) string
}{
{"FibonacciReverseBuf", FibonacciReverseBuf},
{"FibonacciReverse", FibonacciReverse},
{"Fibonacci", Fibonacci},
}
for _, c := range cases {
for _, f := range funcs {
if got := f.f(c.n); got != c.exp {
t.Errorf("%s: Expected: %s, got: %s, n: %d",
f.name, c.exp, got, c.n)
}
}
}
}
Benchmark results
Benchmarking with n = 20:
BenchmarkFibonacciReverseBuf-4 200000 10739 ns/op 18024 B/op 10 allocs/op
BenchmarkFibonacciReverse-4 100000 13208 ns/op 28864 B/op 10 allocs/op
BenchmarkFibonacci-4 500000 3383 ns/op 13728 B/op 3 allocs/op
BenchmarkPeterSO-4 300000 4417 ns/op 13568 B/op 2 allocs/op
BenchmarkPlus-4 200000 6072 ns/op 18832 B/op 18 allocs/op
BenchmarkBuffer-4 50000 29608 ns/op 90256 B/op 60 allocs/op
We can see that this use of bytes.Buffer was much better than yours. Still, using concatenation was faster, because there aren't many concatenations here, they are small ones, and that doesn't require reversing in the end.
On the other hand my Fibonacci() solution outperformed all other presented solutions.
I'm actually in a bit of a trouble...
I have a calculator, but when I want to divide nubers with them, I have a panic err saying that you can't divide by 0.
Well, I know that in maths we can't divide by 0, but I don't put 0 in my ints.
Any idea of the problem ?
Here is the code :
package main
import (
"fmt"
"os"
"strconv"
)
func mult(nums ...int) {
result := 0
total := 1
for _, num := range nums {
result = total * num
total = result
}
fmt.Println(result)
}
func add(nums ...int){
result := 0
total := 0
for _, num := range nums {
result = total + num
total = result
}
fmt.Println(result)
}
func sub(nums ...int){
result := 0
total := 0
for _, num := range nums {
result = num - total
total = result
}
fmt.Println(result)
}
func div(nums ...int){
result := 1
total := 1
for _, num := range nums {
result = num / total
total = result
}
fmt.Println(result)
}
func main() {
var d [] int
var args= os.Args[1:]
nums := make([]int, len(args))
for i := 0; i < len(args); i++ {
nums[i], _ = strconv.Atoi(args[i]);
strconv.Atoi(args[i])
d = append(d, nums[i])
}
num := d
if os.Args[1] == "*"{
mult(num...)
} else if os.Args[1] == "+"{
add(num...)
} else if os.Args[1] == "-"{
sub(num...)
} else if os.Args[1] == "/"{
div(num...)
} else {
fmt.Println("Well well well, you didn't entered a right operand ! Try with +, -, /, or * between double quotes")
}
}
The command I want to run this go code is :
go run calc.exe / 3 2 [Infinite args,...]
If your first parameter will always be a operator select, you can do something like that in your main func, you have a two problems in your main, you are ignoring the convertion error of a string to int and then this index of your array are setted with 0, and you are defining the array larger than you need because your first parameter it's not a number to your div func
nums := make([]int, len(args)-1)
for i := 0; i < len(args); i++ {
ret, errAtoi := strconv.Atoi(args[i])
if errAtoi != nil {
fmt.Println(errAtoi.Error())
} else {
nums[i-1] = ret
d = append(d, nums[i-1])
}
}
I have a bytes.Buffer type variable which I filled with Unicode characters:
var mbuff bytes.Buffer
unicodeSource := 'کیا حال ھے؟'
for i,r := range(unicodeSource) {
mbuff.WriteRune(r)
}
Note: I iterated over a Unicode literals here, but really the source is an infinite loop of user input characters.
Now, I want to remove a Unicode character from any position in the buffer mbuff. The problem is that characters may be of variable byte sizes. So I cannot just pick out the ith byte from mbuff.String() as it might be the beginning, middle, or end of a character. This is my trivial (and horrendous) solution:
// removing Unicode character at position n
var tempString string
currChar := 0
for _, ch := range(mbuff.String()) { // iterate over Unicode chars
if currChar != n { // skip concatenating nth char
tempString += ch
}
currChar++
}
mbuff.Reset() // empty buffer
mbuff.WriteString(tempString) // write new string
This is bad in many ways. For one, I convert buffer to string, remove ith element, and write a new string back into the buffer. Too many operations. Second, I use the += operator in the loop to concatenate Unicode characters into a new string. I am using buffers in the first place exactly to avoid concatenation using += which is slow as this answer points out.
What is an efficient method to remove the ith Unicode character in a bytes.Buffer?
Also what is an efficient way to insert a Unicode character after i-1 Unicode characters (i.e. in the ith place)?
To remove the ith rune from a slice of bytes, loop through the slice counting runes. When the ith rune is found, copy the bytes following the rune down to the position of the ith rune:
func removeAtBytes(p []byte, i int) []byte {
j := 0
k := 0
for k < len(p) {
_, n := utf8.DecodeRune(p[k:])
if i == j {
p = p[:k+copy(p[k:], p[k+n:])]
}
j++
k += n
}
return p
}
This function modifies the backing array of the argument slice, but it does not allocate memory.
Use this function to remove a rune from a bytes.Buffer.
p := removeAtBytes(mbuf.Bytes(), i)
mbuf.Truncate(len(p)) // backing bytes were updated, adjust length
playground example
To remove the ith rune from a string, loop through the string counting runes. When the ith rune is found, create a string by concatenating the segment of the string before the rune with the segment of the string after the rune.
func removeAt(s string, i int) string {
j := 0 // count of runes
k := 0 // index in string of current rune
for k < len(s) {
_, n := utf8.DecodeRuneInString(s[k:])
if i == j {
return s[:k] + s[k+n:]
}
j++
k += n
}
return s
}
This function allocates a single string, the result. DecodeRuneInString is a function in the standard library unicode/utf8 package.
Taking a step back, go often works on Readers and Writers, so an alternative solution would be to use the text/transform package. You create a Transformer, attach it to a Reader and use the new Reader to produce a transformed string. For example here's a skipper:
func main() {
src := strings.NewReader("کیا حال ھے؟")
skipped := transform.NewReader(src, NewSkipper(5))
var buf bytes.Buffer
io.Copy(&buf, skipped)
fmt.Println("RESULT:", buf.String())
}
And here's the implementation:
package main
import (
"bytes"
"fmt"
"io"
"strings"
"unicode/utf8"
"golang.org/x/text/transform"
)
type skipper struct {
pos int
cnt int
}
// NewSkipper creates a text transformer which will remove the rune at pos
func NewSkipper(pos int) transform.Transformer {
return &skipper{pos: pos}
}
func (s *skipper) Transform(dst, src []byte, atEOF bool) (nDst, nSrc int, err error) {
for utf8.FullRune(src) {
_, sz := utf8.DecodeRune(src)
// not enough space in the dst
if len(dst) < sz {
return nDst, nSrc, transform.ErrShortDst
}
if s.pos != s.cnt {
copy(dst[:sz], src[:sz])
// track that we stored in dst
dst = dst[sz:]
nDst += sz
}
// track that we read from src
src = src[sz:]
nSrc += sz
// on to the next rune
s.cnt++
}
if len(src) > 0 && !atEOF {
return nDst, nSrc, transform.ErrShortSrc
}
return nDst, nSrc, nil
}
func (s *skipper) Reset() {
s.cnt = 0
}
There may be bugs with this code, but hopefully you can see the idea.
The benefit of this approach is it could work on a potentially infinite amount of data without having to store all of it in memory. For example you could transform a file this way.
Edit:
Remove the ith rune in the buffer:
A: Shift all runes one location to the left (Here A is faster than B), try it on The Go Playground:
func removeRuneAt(s string, runePosition int) string {
if runePosition < 0 {
return s
}
r := []rune(s)
if runePosition >= len(r) {
return s
}
copy(r[runePosition:], r[runePosition+1:])
return string(r[:len(r)-1])
}
B: Copy to new buffer, try it on The Go Playground
func removeRuneAt(s string, runePosition int) string {
if runePosition < 0 {
return s // avoid allocation
}
r := []rune(s)
if runePosition >= len(r) {
return s // avoid allocation
}
t := make([]rune, len(r)-1) // Apply replacements to buffer.
w := copy(t, r[:runePosition])
w += copy(t[w:], r[runePosition+1:])
return string(t[:w])
}
C: Try it on The Go Playground:
package main
import (
"bytes"
"fmt"
)
func main() {
str := "hello"
fmt.Println(str)
fmt.Println(removeRuneAt(str, 1))
buf := bytes.NewBuffer([]byte(str))
fmt.Println(buf.Bytes())
buf = bytes.NewBuffer([]byte(removeRuneAt(buf.String(), 1)))
fmt.Println(buf.Bytes())
}
func removeRuneAt(s string, runePosition int) string {
if runePosition < 0 {
return s // avoid allocation
}
r := []rune(s)
if runePosition >= len(r) {
return s // avoid allocation
}
t := make([]rune, len(r)-1) // Apply replacements to buffer.
w := copy(t, r[0:runePosition])
w += copy(t[w:], r[runePosition+1:])
return string(t[0:w])
}
D: Benchmark:
A: 745.0426ms
B: 1.0160581s
for 2000000 iterations
1- Short Answer: to replace all (n) instances of a character (or even a string):
n := -1
newR := ""
old := "µ"
buf = bytes.NewBuffer([]byte(strings.Replace(buf.String(), old, newR, n)))
2- For replacing the character(string) in the ith instance in the buffer, you may use:
buf = bytes.NewBuffer([]byte(Replace(buf.String(), oldString, newOrEmptyString, ith)))
See:
// Replace returns a copy of the string s with the ith
// non-overlapping instance of old replaced by new.
func Replace(s, old, new string, ith int) string {
if len(old) == 0 || old == new || ith < 0 {
return s // avoid allocation
}
i, j := 0, 0
for ; ith >= 0; ith-- {
j = strings.Index(s[i:], old)
if j < 0 {
return s // avoid allocation
}
j += i
i = j + len(old)
}
t := make([]byte, len(s)+(len(new)-len(old))) // Apply replacements to buffer.
w := copy(t, s[0:j])
w += copy(t[w:], new)
w += copy(t[w:], s[j+len(old):])
return string(t[0:w])
}
Try it on The Go Playground:
package main
import (
"bytes"
"fmt"
"strings"
)
func main() {
str := `How are you?µ`
fmt.Println(str)
fmt.Println(Replace(str, "µ", "", 0))
buf := bytes.NewBuffer([]byte(str))
fmt.Println(buf.Bytes())
buf = bytes.NewBuffer([]byte(Replace(buf.String(), "µ", "", 0)))
fmt.Println(buf.Bytes())
}
func Replace(s, old, new string, ith int) string {
if len(old) == 0 || old == new || ith < 0 {
return s // avoid allocation
}
i, j := 0, 0
for ; ith >= 0; ith-- {
j = strings.Index(s[i:], old)
if j < 0 {
return s // avoid allocation
}
j += i
i = j + len(old)
}
t := make([]byte, len(s)+(len(new)-len(old))) // Apply replacements to buffer.
w := copy(t, s[0:j])
w += copy(t[w:], new)
w += copy(t[w:], s[j+len(old):])
return string(t[0:w])
}
3- If you want to remove all instances of Unicode character (old string) from any position in the string, you may use:
strings.Replace(str, old, "", -1)
4- Also this works fine for removing from bytes.buffer:
strings.Replace(buf.String(), old, newR, -1)
Like so:
buf = bytes.NewBuffer([]byte(strings.Replace(buf.String(), old, newR, -1)))
Here is the complete working code (try it on The Go Playground):
package main
import (
"bytes"
"fmt"
"strings"
)
func main() {
str := `کیا حال ھے؟` //How are you?
old := `ک`
newR := ""
fmt.Println(strings.Replace(str, old, newR, -1))
buf := bytes.NewBuffer([]byte(str))
// for _, r := range str {
// buf.WriteRune(r)
// }
fmt.Println(buf.Bytes())
bs := []byte(strings.Replace(buf.String(), old, newR, -1))
buf = bytes.NewBuffer(bs)
fmt.Println(" ", buf.Bytes())
}
output:
یا حال ھے؟
[218 169 219 140 216 167 32 216 173 216 167 217 132 32 218 190 219 146 216 159]
[219 140 216 167 32 216 173 216 167 217 132 32 218 190 219 146 216 159]
5- strings.Replace is very efficient, see inside:
// Replace returns a copy of the string s with the first n
// non-overlapping instances of old replaced by new.
// If old is empty, it matches at the beginning of the string
// and after each UTF-8 sequence, yielding up to k+1 replacements
// for a k-rune string.
// If n < 0, there is no limit on the number of replacements.
func Replace(s, old, new string, n int) string {
if old == new || n == 0 {
return s // avoid allocation
}
// Compute number of replacements.
if m := Count(s, old); m == 0 {
return s // avoid allocation
} else if n < 0 || m < n {
n = m
}
// Apply replacements to buffer.
t := make([]byte, len(s)+n*(len(new)-len(old)))
w := 0
start := 0
for i := 0; i < n; i++ {
j := start
if len(old) == 0 {
if i > 0 {
_, wid := utf8.DecodeRuneInString(s[start:])
j += wid
}
} else {
j += Index(s[start:], old)
}
w += copy(t[w:], s[start:j])
w += copy(t[w:], new)
start = j + len(old)
}
w += copy(t[w:], s[start:])
return string(t[0:w])
}
Based on Rob Pike's load balancer demo, I implemented my own priority queue, but my Pop method is not right, can anyone tell me what's wrong?
package main
import (
"fmt"
"container/heap"
)
type ClassRecord struct {
name string
grade int
}
type RecordHeap []*ClassRecord
func (p RecordHeap) Len() int { return len(p) }
func (p RecordHeap) Less(i, j int) bool {
return p[i].grade < p[j].grade
}
func (p *RecordHeap) Swap(i, j int) {
a := *p
a[i], a[j] = a[j], a[i]
}
func (p *RecordHeap) Push(x interface{}) {
a := *p
n := len(a)
a = a[0 : n+1]
r := x.(*ClassRecord)
a[n] = r
*p = a
}
func (p *RecordHeap) Pop() interface{} {
a := *p
*p = a[0 : len(a)-1]
r := a[len(a)-1]
return r
}
func main() {
a := make([]ClassRecord, 6)
a[0] = ClassRecord{"John", 80}
a[1] = ClassRecord{"Dan", 85}
a[2] = ClassRecord{"Aron", 90}
a[3] = ClassRecord{"Mark", 65}
a[4] = ClassRecord{"Rob", 99}
a[5] = ClassRecord{"Brian", 78}
h := make(RecordHeap, 0, 100)
for _, c := range a {
fmt.Println(c)
heap.Push(&h, &c)
fmt.Println("Push: heap has", h.Len(), "items")
}
for i, x := 0, heap.Pop(&h).(*ClassRecord); i < 10 && x != nil; i++ {
fmt.Println("Pop: heap has", h.Len(), "items")
fmt.Println(*x)
}
}
EDIT: besides the way cthom06 pointed out, another way to fix this is to create a pointer array as follows,
a := make([]*ClassRecord, 6)
a[0] = &ClassRecord{"John", 80}
a[1] = &ClassRecord{"Dan", 85}
......
EDIT:
Oh I should've seen this right away.
heap.Push(&h, &c)
You push the address of c, which gets reused on each iteration of range. Every record in the heap is a pointer to the same area in memory, which ends up being Brian. I'm not sure if this is intended behavior or a compiler bug, but
t := c
heap.Push(&h, &t)
works around it.
Also: Your for loop is wrong.
for h.Len() > 0 {
x := heap.Pop(&h...
should fix it.