108 lines
2.7 KiB
Go
108 lines
2.7 KiB
Go
package common
|
||
|
||
import (
|
||
"encoding/json"
|
||
"sort"
|
||
"strings"
|
||
"unicode"
|
||
|
||
"github.com/mozillazg/go-pinyin"
|
||
|
||
"36wisdom/biz/consts"
|
||
)
|
||
|
||
// AnnotatePinyin 写时一次性标注：词表最长匹配优先（整体转拼音），未命中逐字转
|
||
// （pinyin.Tone 带声调）；返回逐字对齐的 JSON 数组字符串，非汉字对应空串，
|
||
// 下标与原文 rune 一一对应。纯函数，文本未变结果不变（幂等）。
|
||
func AnnotatePinyin(text string) string {
|
||
runes := []rune(text)
|
||
out := make([]string, len(runes))
|
||
if len(runes) == 0 {
|
||
return "[]"
|
||
}
|
||
|
||
words := make([]string, 0, len(consts.PinyinOverrides))
|
||
for w := range consts.PinyinOverrides {
|
||
words = append(words, w)
|
||
}
|
||
sort.Slice(words, func(i, j int) bool { return len(words[i]) > len(words[j]) })
|
||
|
||
// 词表最长匹配切分：命中片段整体标注（词表拼音音节数须与字数一致）
|
||
for i := 0; i < len(runes); {
|
||
matched := false
|
||
for _, w := range words {
|
||
n := len([]rune(w))
|
||
if i+n > len(runes) || string(runes[i:i+n]) != w {
|
||
continue
|
||
}
|
||
pys := strings.Fields(consts.PinyinOverrides[w])
|
||
if len(pys) != n {
|
||
continue
|
||
}
|
||
copy(out[i:i+n], pys)
|
||
i += n
|
||
matched = true
|
||
break
|
||
}
|
||
if !matched {
|
||
i++
|
||
}
|
||
}
|
||
|
||
// 未命中汉字逐字转拼音：词典缺失字（生僻字等）SinglePinyin 返回空串，
|
||
// 该字自然保持 ""，后续字不错位
|
||
args := pinyin.Args{Style: pinyin.Tone}
|
||
i := 0
|
||
for _, r := range text {
|
||
if out[i] == "" && unicode.Is(unicode.Han, r) {
|
||
if pys := pinyin.SinglePinyin(r, args); len(pys) > 0 {
|
||
out[i] = pys[0]
|
||
}
|
||
}
|
||
i++
|
||
}
|
||
|
||
b, _ := json.Marshal(out)
|
||
return string(b)
|
||
}
|
||
|
||
// ToPinyinPlain 文本 → 空格分隔带声调拼音串（词表最长匹配优先，未命中逐字转），
|
||
// 与种子数据 strategy.pinyin 格式一致；用于写时标注 name → pinyin。
|
||
func ToPinyinPlain(text string) string {
|
||
runes := []rune(text)
|
||
if len(runes) == 0 {
|
||
return ""
|
||
}
|
||
words := make([]string, 0, len(consts.PinyinOverrides))
|
||
for w := range consts.PinyinOverrides {
|
||
words = append(words, w)
|
||
}
|
||
sort.Slice(words, func(i, j int) bool { return len(words[i]) > len(words[j]) })
|
||
|
||
parts := make([]string, 0, len(runes))
|
||
args := pinyin.Args{Style: pinyin.Tone}
|
||
for i := 0; i < len(runes); {
|
||
matched := false
|
||
for _, w := range words {
|
||
n := len([]rune(w))
|
||
if i+n > len(runes) || string(runes[i:i+n]) != w {
|
||
continue
|
||
}
|
||
parts = append(parts, strings.Fields(consts.PinyinOverrides[w])...)
|
||
i += n
|
||
matched = true
|
||
break
|
||
}
|
||
if matched {
|
||
continue
|
||
}
|
||
if unicode.Is(unicode.Han, runes[i]) {
|
||
if pys := pinyin.SinglePinyin(runes[i], args); len(pys) > 0 {
|
||
parts = append(parts, pys[0])
|
||
}
|
||
}
|
||
i++
|
||
}
|
||
return strings.Join(parts, " ")
|
||
}
|