unhtml: HTML unmarshaler
摘要：
最近要用golang
寫一個需要解析HTML
的專案，到網上找了一個庫叫goquery
。雖然它的API
挺不錯，css selector
基本上也全支援了，但寫這種程式碼果然還是有點無聊，於是我就想，為什麼不能跟go
的json
庫和...
最近要用golang
寫一個需要解析HTML
的專案，到網上找了一個庫叫goquery
。雖然它的API
挺不錯，css selector
基本上也全支援了，但寫這種程式碼果然還是有點無聊，於是我就想，為什麼不能跟go
的json
庫和xml
庫一樣，直接Unmarshal(HTML)
呢？
然後我花了兩天時間擼出了unhtml
->ofollow,noindex" target="_blank">Github 傳送門
樣例 & 效能
有個HTML
var AllTypeHTML = []byte(` <!DOCTYPE html> <html lang="en"> <head> <meta charset="UTF-8"> <title>Title</title> </head> <body> <div id="test"> <ul> <li>0</li> <li>1</li> <li>2</li> <li>3</li> </ul> <div> <p>Hexilee</p> <p>20</p> <p>true</p> </div> <p>Hello World!</p> <p>10</p> <p>3.14</p> <p>true</p> </div> </body> </html> `)
如果你想把它解析為一個結構體
package example
type (
PartTypesStruct struct {
Slice[]int
StructTestUser
Stringstring
Intint
Float64 float64
Boolbool
}
TestUser struct {
Namestring
Ageuint
LikeLemon bool
}
)
直接用goquery
要這樣寫
package example
import (
"bytes"
"github.com/PuerkitoBio/goquery"
"strconv"
)
func parsePartTypesLogically() (PartTypesStruct, error) {
doc, err := goquery.NewDocumentFromReader(bytes.NewReader(AllTypeHTML))
partTypes := PartTypesStruct{}
if err == nil {
selection := doc.Find(partTypes.Root())
partTypes.Slice = make([]int, 0)
selection.Find(`ul > li`).Each(func(i int, selection *goquery.Selection) {
Int, parseErr := strconv.Atoi(selection.Text())
if parseErr != nil {
err = parseErr
}
partTypes.Slice = append(partTypes.Slice, Int)
})
if err == nil {
partTypes.Struct.Name = selection.Find(`#test > div > p:nth-child(1)`).Text()
Int, parseErr := strconv.Atoi(selection.Find(`#test > div > p:nth-child(2)`).Text())
if err = parseErr; err == nil {
partTypes.Struct.Age = uint(Int)
Bool, parseErr := strconv.ParseBool(selection.Find(`#test > div > p:nth-child(3)`).Text())
if err = parseErr; err == nil {
partTypes.Struct.LikeLemon = Bool
String := selection.Find(`#test > p:nth-child(3)`).Text()
Int, parseErr := strconv.Atoi(selection.Find(`#test > p:nth-child(4)`).Text())
if err = parseErr; err != nil {
return partTypes, err
}
Float64, parseErr := strconv.ParseFloat(selection.Find(`#test > p:nth-child(5)`).Text(), 0)
if err = parseErr; err != nil {
return partTypes, err
}
Bool, parseErr := strconv.ParseBool(selection.Find(`#test > p:nth-child(6)`).Text())
if err = parseErr; err != nil {
return partTypes, err
}
partTypes.String = String
partTypes.Int = Int
partTypes.Float64 = Float64
partTypes.Bool = Bool
}
}
}
}
return partTypes, err
}
寫得很難受
而現在你只要這麼寫
package main
import (
"encoding/json"
"fmt"
"github.com/Hexilee/unhtml"
"io/ioutil"
)
type (
PartTypesStruct struct {
Slice[]int`html:"ul > li"`
StructTestUser `html:"#test > div"`
Stringstring`html:"#test > p:nth-child(3)"`
Intint`html:"#test > p:nth-child(4)"`
Float64 float64`html:"#test > p:nth-child(5)"`
Boolbool`html:"#test > p:nth-child(6)"`
}
TestUser struct {
Namestring `html:"p:nth-child(1)"`
Ageuint`html:"p:nth-child(2)"`
LikeLemon bool`html:"p:nth-child(3)"`
}
)
func (PartTypesStruct) Root() string {
return "#test"
}
func main() {
allTypes := PartTypesStruct{}
_ := unhtml.Unmarshal(AllTypeHTML, &allTypes)
result, _ := json.Marshal(&allTypes)
fmt.Println(string(result))
}
就能得到結果
{
"Slice": [
0,
1,
2,
3
],
"Struct": {
"Name": "Hexilee",
"Age": 20,
"LikeLemon": true
},
"String": "Hello World!",
"Int": 10,
"Float64": 3.14,
"Bool": true
}
開發效率大大提升！但毫無疑問用了大量反射，讓人擔心它的執行效率。於是我寫了兩個Benchmarks
func BenchmarkUnmarshalPartTypes(b *testing.B) {
assert.NotNil(b, AllTypeHTML)
for i := 0; i < b.N; i++ {
partTypes := PartTypesStruct{}
assert.Nil(b, Unmarshal(AllTypeHTML, &partTypes))
}
}
func BenchmarkParsePartTypesLogically(b *testing.B) {
assert.NotNil(b, AllTypeHTML)
for i := 0; i < b.N; i++ {
_, err := parsePartTypesLogically()
assert.Nil(b, err)
}
}
測試結果：
> go test -bench=. goos: darwin goarch: amd64 pkg: github.com/Hexilee/unhtml BenchmarkUnmarshalPartTypes-43000054096 ns/op BenchmarkParsePartTypesLogically-43000045188 ns/op PASS okgithub.com/Hexilee/unhtml4.098s
執行效率稍微低些，但這只是展示和測試用的HTML
，在解析實際中更復雜的HTML
時兩者的執行效率是十分接近的。
一些注意事項和特性請看README