htmlToText renders HTML to plain text: drops script/style, inserts newlines at block boundaries, and decodes entities (via the tokenizer).
(h string)
| 213 | // htmlToText renders HTML to plain text: drops script/style, inserts newlines |
| 214 | // at block boundaries, and decodes entities (via the tokenizer). |
| 215 | func htmlToText(h string) string { |
| 216 | z := html.NewTokenizer(strings.NewReader(h)) |
| 217 | var sb strings.Builder |
| 218 | skip := 0 // depth inside script/style |
| 219 | for { |
| 220 | tt := z.Next() |
| 221 | switch tt { |
| 222 | case html.ErrorToken: |
| 223 | return sb.String() |
| 224 | case html.StartTagToken, html.SelfClosingTagToken: |
| 225 | name, _ := z.TagName() |
| 226 | tag := string(name) |
| 227 | switch tag { |
| 228 | case "script", "style", "head": |
| 229 | if tt == html.StartTagToken { |
| 230 | skip++ |
| 231 | } |
| 232 | case "br", "p", "div", "tr", "li", "h1", "h2", "h3", "h4", "h5", "h6", "blockquote": |
| 233 | sb.WriteByte('\n') |
| 234 | } |
| 235 | case html.EndTagToken: |
| 236 | name, _ := z.TagName() |
| 237 | switch string(name) { |
| 238 | case "script", "style", "head": |
| 239 | if skip > 0 { |
| 240 | skip-- |
| 241 | } |
| 242 | case "p", "div", "tr", "li", "blockquote": |
| 243 | sb.WriteByte('\n') |
| 244 | } |
| 245 | case html.TextToken: |
| 246 | if skip == 0 { |
| 247 | sb.Write(z.Text()) |
| 248 | } |
| 249 | } |
| 250 | } |
| 251 | } |
| 252 | |
| 253 | // quoteMarker matches the start of a quoted reply / forwarded block. Anything |
| 254 | // from the first match onward is dropped. |