| 1 | package converter |
| 2 | |
| 3 | import ( |
| 4 | "bytes" |
| 5 | "fmt" |
| 6 | htmlpkg "html" |
| 7 | "strings" |
| 8 | |
| 9 | "golang.org/x/net/html" |
| 10 | ) |
| 11 | |
| 12 | // ConfluenceToMarkdown converts Confluence storage format XML to Markdown. |
| 13 | func ConfluenceToMarkdown(source string) (string, error) { |
| 14 | // Preprocess: extract CDATA content and replace with escaped text, |
| 15 | // because x/net/html doesn't handle CDATA sections. |
| 16 | preprocessed := preprocessCDATA(source) |
| 17 | |
| 18 | // Wrap in a root element so the HTML parser handles it correctly. |
| 19 | wrapped := "<div>" + preprocessed + "</div>" |
| 20 | doc, err := html.Parse(strings.NewReader(wrapped)) |
| 21 | if err != nil { |
| 22 | return "", fmt.Errorf("parsing confluence xml: %w", err) |
| 23 | } |
| 24 | |
| 25 | var buf bytes.Buffer |
| 26 | c := &xmlConverter{buf: &buf} |
| 27 | |
| 28 | // Navigate to the wrapper div: html > head > body > div |
| 29 | body := findNode(doc, "body") |
| 30 | if body == nil { |
| 31 | return "", fmt.Errorf("unexpected parse structure") |
| 32 | } |
| 33 | wrapper := body.FirstChild |
| 34 | if wrapper != nil { |
| 35 | c.walkChildren(wrapper, 0) |
| 36 | } |
| 37 | |
| 38 | result := buf.String() |
| 39 | // Clean up excessive blank lines |
| 40 | for strings.Contains(result, "\n\n\n") { |
| 41 | result = strings.ReplaceAll(result, "\n\n\n", "\n\n") |
| 42 | } |
| 43 | return strings.TrimSpace(result) + "\n", nil |
| 44 | } |
| 45 | |
| 46 | // preprocessCDATA replaces <![CDATA[...]]> with the content as a data attribute |
| 47 | // on the parent element, since x/net/html doesn't parse CDATA. |
| 48 | func preprocessCDATA(s string) string { |
| 49 | var result strings.Builder |
| 50 | for { |
| 51 | idx := strings.Index(s, "<![CDATA[") |
| 52 | if idx == -1 { |
| 53 | result.WriteString(s) |
| 54 | break |
| 55 | } |
| 56 | result.WriteString(s[:idx]) |
| 57 | s = s[idx+len("<![CDATA["):] |
| 58 | endIdx := strings.Index(s, "]]>") |
| 59 | if endIdx == -1 { |
| 60 | result.WriteString(s) |
| 61 | break |
| 62 | } |
| 63 | // Write CDATA content as a special element that we can detect |
| 64 | content := s[:endIdx] |
| 65 | result.WriteString("<cdatacontent>") |
| 66 | result.WriteString(htmlpkg.EscapeString(content)) |
| 67 | result.WriteString("</cdatacontent>") |
| 68 | s = s[endIdx+len("]]>"):] |
| 69 | } |
| 70 | return result.String() |
| 71 | } |
| 72 | |
| 73 | type xmlConverter struct { |
| 74 | buf *bytes.Buffer |
| 75 | listDepth int |
| 76 | inListItem bool |
| 77 | } |
| 78 | |
| 79 | func (c *xmlConverter) walkChildren(n *html.Node, depth int) { |
| 80 | for child := n.FirstChild; child != nil; child = child.NextSibling { |
| 81 | c.walk(child, depth) |
| 82 | } |
| 83 | } |
| 84 | |
| 85 | func (c *xmlConverter) walk(n *html.Node, depth int) { |
| 86 | if n.Type == html.TextNode { |
| 87 | text := n.Data |
| 88 | // Skip whitespace-only text nodes inside lists |
| 89 | if c.listDepth > 0 && strings.TrimSpace(text) == "" { |
| 90 | return |
| 91 | } |
| 92 | // Collapse whitespace in text nodes (XML indentation artifacts) |
| 93 | if strings.TrimSpace(text) != "" { |
| 94 | // Replace sequences of whitespace (including newlines) with single space, |
| 95 | // but preserve the trimmed content |
| 96 | text = collapseWhitespace(text) |
| 97 | } |
| 98 | c.buf.WriteString(text) |
| 99 | return |
| 100 | } |
| 101 | |
| 102 | if n.Type != html.ElementNode { |
| 103 | c.walkChildren(n, depth) |
| 104 | return |
| 105 | } |
| 106 | |
| 107 | tag := strings.ToLower(n.Data) |
| 108 | |
| 109 | switch { |
| 110 | // Headings |
| 111 | case tag == "h1": |
| 112 | c.buf.WriteString("\n# ") |
| 113 | c.walkChildren(n, depth) |
| 114 | c.buf.WriteString("\n\n") |
| 115 | case tag == "h2": |
| 116 | c.buf.WriteString("\n## ") |
| 117 | c.walkChildren(n, depth) |
| 118 | c.buf.WriteString("\n\n") |
| 119 | case tag == "h3": |
| 120 | c.buf.WriteString("\n### ") |
| 121 | c.walkChildren(n, depth) |
| 122 | c.buf.WriteString("\n\n") |
| 123 | case tag == "h4": |
| 124 | c.buf.WriteString("\n#### ") |
| 125 | c.walkChildren(n, depth) |
| 126 | c.buf.WriteString("\n\n") |
| 127 | case tag == "h5": |
| 128 | c.buf.WriteString("\n##### ") |
| 129 | c.walkChildren(n, depth) |
| 130 | c.buf.WriteString("\n\n") |
| 131 | case tag == "h6": |
| 132 | c.buf.WriteString("\n###### ") |
| 133 | c.walkChildren(n, depth) |
| 134 | c.buf.WriteString("\n\n") |
| 135 | |
| 136 | // Paragraphs |
| 137 | case tag == "p": |
| 138 | c.walkChildren(n, depth) |
| 139 | if !c.inListItem { |
| 140 | c.buf.WriteString("\n\n") |
| 141 | } |
| 142 | |
| 143 | // Inline formatting |
| 144 | case tag == "strong", tag == "b": |
| 145 | c.buf.WriteString("**") |
| 146 | c.walkChildren(n, depth) |
| 147 | c.buf.WriteString("**") |
| 148 | case tag == "em", tag == "i": |
| 149 | c.buf.WriteString("*") |
| 150 | c.walkChildren(n, depth) |
| 151 | c.buf.WriteString("*") |
| 152 | case tag == "del", tag == "s": |
| 153 | c.buf.WriteString("~~") |
| 154 | c.walkChildren(n, depth) |
| 155 | c.buf.WriteString("~~") |
| 156 | case tag == "code": |
| 157 | if !isPrevSiblingCode(n) { |
| 158 | c.buf.WriteString("`") |
| 159 | } |
| 160 | c.walkChildren(n, depth) |
| 161 | if !isNextSiblingCode(n) { |
| 162 | c.buf.WriteString("`") |
| 163 | } |
| 164 | |
| 165 | // Links |
| 166 | case tag == "a": |
| 167 | href := getAttr(n, "href") |
| 168 | c.buf.WriteString("[") |
| 169 | c.walkChildren(n, depth) |
| 170 | c.buf.WriteString("](") |
| 171 | c.buf.WriteString(href) |
| 172 | c.buf.WriteString(")") |
| 173 | |
| 174 | // Line break |
| 175 | case tag == "br": |
| 176 | c.buf.WriteString(" \n") |
| 177 | |
| 178 | // Horizontal rule |
| 179 | case tag == "hr": |
| 180 | c.buf.WriteString("\n---\n\n") |
| 181 | |
| 182 | // Lists |
| 183 | case tag == "ul": |
| 184 | c.listDepth++ |
| 185 | if c.listDepth == 1 { |
| 186 | c.buf.WriteString("\n") |
| 187 | } |
| 188 | c.walkChildren(n, depth) |
| 189 | c.listDepth-- |
| 190 | if c.listDepth == 0 { |
| 191 | c.buf.WriteString("\n") |
| 192 | } |
| 193 | case tag == "ol": |
| 194 | c.listDepth++ |
| 195 | if c.listDepth == 1 { |
| 196 | c.buf.WriteString("\n") |
| 197 | } |
| 198 | c.walkOL(n, depth) |
| 199 | c.listDepth-- |
| 200 | if c.listDepth == 0 { |
| 201 | c.buf.WriteString("\n") |
| 202 | } |
| 203 | case tag == "li": |
| 204 | prev := c.inListItem |
| 205 | c.inListItem = true |
| 206 | // Check if this list item contains a task checkbox |
| 207 | if hasTaskStatus(n) { |
| 208 | // Task status handler will write the prefix, walkChildrenInline for text |
| 209 | c.walkChildrenInline(n, depth) |
| 210 | c.buf.WriteString("\n") |
| 211 | } else { |
| 212 | indent := strings.Repeat(" ", max(0, c.listDepth-1)) |
| 213 | c.buf.WriteString(indent) |
| 214 | c.buf.WriteString("- ") |
| 215 | c.walkChildrenInline(n, depth) |
| 216 | c.buf.WriteString("\n") |
| 217 | } |
| 218 | c.inListItem = prev |
| 219 | |
| 220 | // Tables - convert to GFM table |
| 221 | case tag == "table": |
| 222 | c.renderTable(n, depth) |
| 223 | |
| 224 | // Confluence macros - handled via ac:* namespace (parsed as ac-*) |
| 225 | // The HTML parser lowercases and handles colons differently. |
| 226 | // We need to handle both ac:structured-macro and the parsed form. |
| 227 | |
| 228 | // Skip layout/structural elements, pass through children |
| 229 | case tag == "div", tag == "span", tag == "tbody", tag == "thead", |
| 230 | tag == "colgroup", tag == "col", tag == "content-wrapper": |
| 231 | c.walkChildren(n, depth) |
| 232 | |
| 233 | // Handle Confluence-specific elements |
| 234 | default: |
| 235 | c.handleConfluenceElement(n, tag, depth) |
| 236 | } |
| 237 | } |
| 238 | |
| 239 | func (c *xmlConverter) handleConfluenceElement(n *html.Node, tag string, depth int) { |
| 240 | switch { |
| 241 | // Layout elements — preserve as HTML comments for round-trip |
| 242 | case strings.Contains(tag, "ac:layout-section") || strings.Contains(tag, "layout-section"): |
| 243 | sectionType := getAttr(n, "ac:type") |
| 244 | if sectionType == "" { |
| 245 | sectionType = getAttr(n, "type") |
| 246 | } |
| 247 | fmt.Fprintf(c.buf, "<!-- ac:layout-section type=%q -->\n", sectionType) |
| 248 | c.walkChildren(n, depth) |
| 249 | c.buf.WriteString("<!-- /ac:layout-section -->\n") |
| 250 | |
| 251 | case strings.Contains(tag, "ac:layout-cell") || strings.Contains(tag, "layout-cell"): |
| 252 | c.buf.WriteString("<!-- ac:layout-cell -->\n") |
| 253 | c.walkChildren(n, depth) |
| 254 | c.buf.WriteString("<!-- /ac:layout-cell -->\n") |
| 255 | |
| 256 | case tag == "ac:layout" || strings.Contains(tag, "layout") && !strings.Contains(tag, "layout-"): |
| 257 | c.buf.WriteString("<!-- ac:layout -->\n") |
| 258 | c.walkChildren(n, depth) |
| 259 | c.buf.WriteString("<!-- /ac:layout -->\n") |
| 260 | // Confluence structured macros (code blocks, panels, etc.) |
| 261 | case strings.Contains(tag, "structured-macro") || strings.Contains(tag, "ac:structured-macro"): |
| 262 | macroName := getAttr(n, "ac:name") |
| 263 | if macroName == "" { |
| 264 | macroName = getAttr(n, "name") |
| 265 | } |
| 266 | macroID := getAttr(n, "ac:macro-id") |
| 267 | if macroID == "" { |
| 268 | macroID = getAttr(n, "macro-id") |
| 269 | } |
| 270 | switch macroName { |
| 271 | case "code": |
| 272 | c.renderCodeMacro(n, macroID) |
| 273 | case "info": |
| 274 | c.renderPanelAsBlockquote(n, depth) |
| 275 | case "note": |
| 276 | c.renderPanelAsBlockquote(n, depth) |
| 277 | case "warning": |
| 278 | c.renderPanelAsBlockquote(n, depth) |
| 279 | case "toc": |
| 280 | // Preserve TOC macro as HTML comment |
| 281 | if macroID != "" { |
| 282 | fmt.Fprintf(c.buf, "<!-- ac:toc macro-id=%q -->\n", macroID) |
| 283 | } else { |
| 284 | c.buf.WriteString("<!-- ac:toc -->\n") |
| 285 | } |
| 286 | default: |
| 287 | c.walkChildren(n, depth) |
| 288 | } |
| 289 | |
| 290 | // Confluence images |
| 291 | case strings.Contains(tag, "image") || strings.Contains(tag, "ac:image"): |
| 292 | alt := getAttr(n, "ac:alt") |
| 293 | if alt == "" { |
| 294 | alt = getAttr(n, "alt") |
| 295 | } |
| 296 | imgRef := c.findImageRef(n) |
| 297 | if imgRef.isAttachment { |
| 298 | // Preserve attachment reference as round-trippable HTML |
| 299 | fmt.Fprintf(c.buf, `<span data-attachment="%s"`, imgRef.filename) |
| 300 | if alt != "" { |
| 301 | fmt.Fprintf(c.buf, ` data-alt="%s"`, alt) |
| 302 | } |
| 303 | c.buf.WriteString("/>") |
| 304 | } else { |
| 305 | c.buf.WriteString("![") |
| 306 | c.buf.WriteString(alt) |
| 307 | c.buf.WriteString("](") |
| 308 | c.buf.WriteString(imgRef.url) |
| 309 | c.buf.WriteString(")") |
| 310 | } |
| 311 | |
| 312 | // Confluence links (user mentions, page links) |
| 313 | case strings.Contains(tag, "ac:link"): |
| 314 | if c.hasUserChild(n) { |
| 315 | c.walkChildren(n, depth) |
| 316 | } else { |
| 317 | c.walkChildren(n, depth) |
| 318 | } |
| 319 | |
| 320 | // Confluence emoticons |
| 321 | case strings.Contains(tag, "emoticon") || strings.Contains(tag, "ac:emoticon"): |
| 322 | name := getAttr(n, "ac:name") |
| 323 | if name == "" { |
| 324 | name = getAttr(n, "name") |
| 325 | } |
| 326 | switch name { |
| 327 | case "plus": |
| 328 | c.buf.WriteString("(+)") |
| 329 | case "minus": |
| 330 | c.buf.WriteString("(-)") |
| 331 | case "question": |
| 332 | c.buf.WriteString("(?)") |
| 333 | case "tick": |
| 334 | c.buf.WriteString("(v)") |
| 335 | case "cross": |
| 336 | c.buf.WriteString("(x)") |
| 337 | } |
| 338 | |
| 339 | // Confluence task lists |
| 340 | case strings.Contains(tag, "task-list"): |
| 341 | c.listDepth++ |
| 342 | c.walkChildren(n, depth) |
| 343 | c.listDepth-- |
| 344 | case strings.Contains(tag, "task-body"): |
| 345 | c.walkChildren(n, depth) |
| 346 | c.buf.WriteString("\n") |
| 347 | case strings.Contains(tag, "task-status"): |
| 348 | status := strings.TrimSpace(getTextContent(n)) |
| 349 | indent := strings.Repeat(" ", max(0, c.listDepth-1)) |
| 350 | if status == "complete" { |
| 351 | c.buf.WriteString(indent + "- [x] ") |
| 352 | } else { |
| 353 | c.buf.WriteString(indent + "- [ ] ") |
| 354 | } |
| 355 | case strings.Contains(tag, "task-id"): |
| 356 | // Skip task IDs |
| 357 | case strings.Contains(tag, "task") && !strings.Contains(tag, "task-"): |
| 358 | c.walkChildren(n, depth) |
| 359 | |
| 360 | // Confluence inline comment markers — preserve as span with data attribute |
| 361 | case strings.Contains(tag, "inline-comment-marker"): |
| 362 | ref := getAttr(n, "ac:ref") |
| 363 | if ref == "" { |
| 364 | ref = getAttr(n, "ref") |
| 365 | } |
| 366 | if ref != "" { |
| 367 | fmt.Fprintf(c.buf, `<span data-inline-comment="%s">`, ref) |
| 368 | c.walkChildren(n, depth) |
| 369 | c.buf.WriteString("</span>") |
| 370 | } else { |
| 371 | c.walkChildren(n, depth) |
| 372 | } |
| 373 | |
| 374 | // User references — preserve as round-trippable HTML span |
| 375 | case strings.Contains(tag, "ri:user"): |
| 376 | userKey := getAttr(n, "ri:userkey") |
| 377 | if userKey == "" { |
| 378 | userKey = getAttr(n, "userkey") |
| 379 | } |
| 380 | if userKey != "" { |
| 381 | fmt.Fprintf(c.buf, `<span data-user-key="%s"/>`, userKey) |
| 382 | } |
| 383 | |
| 384 | // Time elements |
| 385 | case tag == "time": |
| 386 | datetime := getAttr(n, "datetime") |
| 387 | if datetime != "" { |
| 388 | c.buf.WriteString(datetime) |
| 389 | } |
| 390 | |
| 391 | // Fallback: just walk children |
| 392 | default: |
| 393 | c.walkChildren(n, depth) |
| 394 | } |
| 395 | } |
| 396 | |
| 397 | func (c *xmlConverter) renderCodeMacro(n *html.Node, macroID string) { |
| 398 | language := "" |
| 399 | code := "" |
| 400 | |
| 401 | // Walk children to find parameters and body |
| 402 | var walkMacro func(*html.Node) |
| 403 | walkMacro = func(node *html.Node) { |
| 404 | if node.Type == html.ElementNode { |
| 405 | tag := strings.ToLower(node.Data) |
| 406 | if strings.Contains(tag, "parameter") || strings.Contains(tag, "ac:parameter") { |
| 407 | name := getAttr(node, "ac:name") |
| 408 | if name == "" { |
| 409 | name = getAttr(node, "name") |
| 410 | } |
| 411 | if name == "language" { |
| 412 | language = getTextContent(node) |
| 413 | } |
| 414 | } |
| 415 | if strings.Contains(tag, "plain-text-body") || strings.Contains(tag, "ac:plain-text-body") { |
| 416 | code = getCDATAContent(node) |
| 417 | } |
| 418 | } |
| 419 | for child := node.FirstChild; child != nil; child = child.NextSibling { |
| 420 | walkMacro(child) |
| 421 | } |
| 422 | } |
| 423 | walkMacro(n) |
| 424 | |
| 425 | // Extract original attribute order for round-trip fidelity |
| 426 | attrOrder := extractAttrOrder(n) |
| 427 | |
| 428 | if macroID != "" { |
| 429 | if attrOrder != "" { |
| 430 | fmt.Fprintf(c.buf, "\n<!-- ac:code macro-id=%q attr-order=%q -->\n", macroID, attrOrder) |
| 431 | } else { |
| 432 | fmt.Fprintf(c.buf, "\n<!-- ac:code macro-id=%q -->\n", macroID) |
| 433 | } |
| 434 | } else { |
| 435 | c.buf.WriteString("\n") |
| 436 | } |
| 437 | c.buf.WriteString("```") |
| 438 | c.buf.WriteString(language) |
| 439 | c.buf.WriteString("\n") |
| 440 | c.buf.WriteString(code) |
| 441 | if !strings.HasSuffix(code, "\n") { |
| 442 | c.buf.WriteString("\n") |
| 443 | } |
| 444 | c.buf.WriteString("```\n\n") |
| 445 | } |
| 446 | |
| 447 | func (c *xmlConverter) renderPanelAsBlockquote(n *html.Node, depth int) { |
| 448 | // Collect panel body content |
| 449 | var bodyBuf bytes.Buffer |
| 450 | origBuf := c.buf |
| 451 | c.buf = &bodyBuf |
| 452 | |
| 453 | // Find rich-text-body and walk it |
| 454 | var findBody func(*html.Node) |
| 455 | findBody = func(node *html.Node) { |
| 456 | if node.Type == html.ElementNode { |
| 457 | tag := strings.ToLower(node.Data) |
| 458 | if strings.Contains(tag, "rich-text-body") { |
| 459 | c.walkChildren(node, depth) |
| 460 | return |
| 461 | } |
| 462 | } |
| 463 | for child := node.FirstChild; child != nil; child = child.NextSibling { |
| 464 | findBody(child) |
| 465 | } |
| 466 | } |
| 467 | findBody(n) |
| 468 | |
| 469 | c.buf = origBuf |
| 470 | text := strings.TrimSpace(bodyBuf.String()) |
| 471 | lines := strings.Split(text, "\n") |
| 472 | for _, line := range lines { |
| 473 | c.buf.WriteString("> ") |
| 474 | c.buf.WriteString(line) |
| 475 | c.buf.WriteString("\n") |
| 476 | } |
| 477 | c.buf.WriteString("\n") |
| 478 | } |
| 479 | |
| 480 | func (c *xmlConverter) renderTable(n *html.Node, depth int) { |
| 481 | rows := collectTableRows(n) |
| 482 | if len(rows) == 0 { |
| 483 | return |
| 484 | } |
| 485 | |
| 486 | // Determine column count |
| 487 | cols := 0 |
| 488 | for _, row := range rows { |
| 489 | if len(row.cells) > cols { |
| 490 | cols = len(row.cells) |
| 491 | } |
| 492 | } |
| 493 | if cols == 0 { |
| 494 | return |
| 495 | } |
| 496 | |
| 497 | // Preserve table attributes and colgroup as HTML comment |
| 498 | tableAttrs := extractTableAttrs(n) |
| 499 | if tableAttrs != "" { |
| 500 | fmt.Fprintf(c.buf, "\n<!-- table-attrs: %s -->\n", tableAttrs) |
| 501 | } else { |
| 502 | c.buf.WriteString("\n") |
| 503 | } |
| 504 | |
| 505 | // If first row is a header |
| 506 | isFirstRowHeader := len(rows) > 0 && rows[0].isHeader |
| 507 | startIdx := 0 |
| 508 | |
| 509 | if isFirstRowHeader { |
| 510 | c.writeTableRow(rows[0].cells, cols) |
| 511 | c.writeTableSep(cols) |
| 512 | startIdx = 1 |
| 513 | } else { |
| 514 | // Write empty header and separator |
| 515 | empty := make([]string, cols) |
| 516 | c.writeTableRow(empty, cols) |
| 517 | c.writeTableSep(cols) |
| 518 | } |
| 519 | |
| 520 | for i := startIdx; i < len(rows); i++ { |
| 521 | c.writeTableRow(rows[i].cells, cols) |
| 522 | } |
| 523 | c.buf.WriteString("\n") |
| 524 | } |
| 525 | |
| 526 | func (c *xmlConverter) writeTableRow(cells []string, cols int) { |
| 527 | c.buf.WriteString("|") |
| 528 | for i := range cols { |
| 529 | cell := "" |
| 530 | if i < len(cells) { |
| 531 | cell = cells[i] |
| 532 | } |
| 533 | c.buf.WriteString(" ") |
| 534 | c.buf.WriteString(cell) |
| 535 | c.buf.WriteString(" |") |
| 536 | } |
| 537 | c.buf.WriteString("\n") |
| 538 | } |
| 539 | |
| 540 | func (c *xmlConverter) writeTableSep(cols int) { |
| 541 | c.buf.WriteString("|") |
| 542 | for range cols { |
| 543 | c.buf.WriteString("---|") |
| 544 | } |
| 545 | c.buf.WriteString("\n") |
| 546 | } |
| 547 | |
| 548 | func (c *xmlConverter) walkOL(n *html.Node, depth int) { |
| 549 | idx := 1 |
| 550 | for child := n.FirstChild; child != nil; child = child.NextSibling { |
| 551 | if child.Type != html.ElementNode { |
| 552 | continue |
| 553 | } |
| 554 | tag := strings.ToLower(child.Data) |
| 555 | if tag == "li" { |
| 556 | indent := strings.Repeat(" ", max(0, c.listDepth-1)) |
| 557 | c.buf.WriteString(indent) |
| 558 | fmt.Fprintf(c.buf, "%d. ", idx) |
| 559 | c.walkChildrenInline(child, depth) |
| 560 | c.buf.WriteString("\n") |
| 561 | idx++ |
| 562 | } |
| 563 | } |
| 564 | } |
| 565 | |
| 566 | func (c *xmlConverter) walkChildrenInline(n *html.Node, depth int) { |
| 567 | for child := n.FirstChild; child != nil; child = child.NextSibling { |
| 568 | if child.Type == html.TextNode { |
| 569 | // Collapse whitespace but preserve a single space between inline elements |
| 570 | text := collapseWhitespace(child.Data) |
| 571 | // Only trim leading space if this is the very first child |
| 572 | if child == n.FirstChild { |
| 573 | text = strings.TrimLeft(text, " ") |
| 574 | } |
| 575 | // Only trim trailing space if this is the very last child |
| 576 | if child.NextSibling == nil { |
| 577 | text = strings.TrimRight(text, " ") |
| 578 | } |
| 579 | if text != "" { |
| 580 | c.buf.WriteString(text) |
| 581 | } |
| 582 | continue |
| 583 | } |
| 584 | if child.Type == html.ElementNode { |
| 585 | tag := strings.ToLower(child.Data) |
| 586 | switch { |
| 587 | case tag == "p": |
| 588 | c.walkChildrenInline(child, depth) |
| 589 | case tag == "ul", tag == "ol": |
| 590 | c.buf.WriteString("\n") |
| 591 | c.walk(child, depth) |
| 592 | default: |
| 593 | c.walk(child, depth) |
| 594 | } |
| 595 | } |
| 596 | } |
| 597 | } |
| 598 | |
| 599 | // extractTableAttrs extracts class, style, and colgroup info as a JSON-like string for preservation. |
| 600 | func extractTableAttrs(table *html.Node) string { |
| 601 | var parts []string |
| 602 | |
| 603 | // Table class and style |
| 604 | cls := getAttr(table, "class") |
| 605 | style := getAttr(table, "style") |
| 606 | if cls != "" { |
| 607 | parts = append(parts, fmt.Sprintf("class=%q", cls)) |
| 608 | } |
| 609 | if style != "" { |
| 610 | parts = append(parts, fmt.Sprintf("style=%q", style)) |
| 611 | } |
| 612 | |
| 613 | // Colgroup |
| 614 | var colWidths []string |
| 615 | for child := table.FirstChild; child != nil; child = child.NextSibling { |
| 616 | if child.Type == html.ElementNode && strings.ToLower(child.Data) == "colgroup" { |
| 617 | for col := child.FirstChild; col != nil; col = col.NextSibling { |
| 618 | if col.Type == html.ElementNode && strings.ToLower(col.Data) == "col" { |
| 619 | colStyle := getAttr(col, "style") |
| 620 | if colStyle != "" { |
| 621 | colWidths = append(colWidths, colStyle) |
| 622 | } |
| 623 | } |
| 624 | } |
| 625 | } |
| 626 | } |
| 627 | if len(colWidths) > 0 { |
| 628 | parts = append(parts, fmt.Sprintf("cols=[%s]", strings.Join(colWidths, "|"))) |
| 629 | } |
| 630 | |
| 631 | return strings.Join(parts, " ") |
| 632 | } |
| 633 | |
| 634 | type tableRow struct { |
| 635 | isHeader bool |
| 636 | cells []string |
| 637 | } |
| 638 | |
| 639 | func collectTableRows(table *html.Node) []tableRow { |
| 640 | var rows []tableRow |
| 641 | var walk func(*html.Node, bool) |
| 642 | walk = func(n *html.Node, inHeader bool) { |
| 643 | if n.Type == html.ElementNode { |
| 644 | tag := strings.ToLower(n.Data) |
| 645 | switch tag { |
| 646 | case "thead": |
| 647 | for child := n.FirstChild; child != nil; child = child.NextSibling { |
| 648 | walk(child, true) |
| 649 | } |
| 650 | return |
| 651 | case "tbody": |
| 652 | for child := n.FirstChild; child != nil; child = child.NextSibling { |
| 653 | walk(child, false) |
| 654 | } |
| 655 | return |
| 656 | case "tr": |
| 657 | row := tableRow{isHeader: inHeader} |
| 658 | for child := n.FirstChild; child != nil; child = child.NextSibling { |
| 659 | if child.Type == html.ElementNode { |
| 660 | cellTag := strings.ToLower(child.Data) |
| 661 | if cellTag == "th" { |
| 662 | row.isHeader = true |
| 663 | row.cells = append(row.cells, strings.TrimSpace(renderCellMarkdown(child))) |
| 664 | } else if cellTag == "td" { |
| 665 | row.cells = append(row.cells, strings.TrimSpace(renderCellMarkdown(child))) |
| 666 | } |
| 667 | } |
| 668 | } |
| 669 | rows = append(rows, row) |
| 670 | return |
| 671 | } |
| 672 | } |
| 673 | for child := n.FirstChild; child != nil; child = child.NextSibling { |
| 674 | walk(child, inHeader) |
| 675 | } |
| 676 | } |
| 677 | walk(table, false) |
| 678 | return rows |
| 679 | } |
| 680 | |
| 681 | // renderCellMarkdown renders cell content to inline markdown, preserving |
| 682 | // formatting like bold, italic, code, links, br, and user references. |
| 683 | func renderCellMarkdown(cell *html.Node) string { |
| 684 | var buf bytes.Buffer |
| 685 | renderCellNode(&buf, cell) |
| 686 | return buf.String() |
| 687 | } |
| 688 | |
| 689 | func renderCellNode(buf *bytes.Buffer, n *html.Node) { |
| 690 | for child := n.FirstChild; child != nil; child = child.NextSibling { |
| 691 | switch child.Type { |
| 692 | case html.TextNode: |
| 693 | text := collapseWhitespace(child.Data) |
| 694 | buf.WriteString(text) |
| 695 | case html.ElementNode: |
| 696 | tag := strings.ToLower(child.Data) |
| 697 | switch { |
| 698 | case tag == "strong" || tag == "b": |
| 699 | buf.WriteString("**") |
| 700 | renderCellNode(buf, child) |
| 701 | buf.WriteString("**") |
| 702 | case tag == "em" || tag == "i": |
| 703 | buf.WriteString("*") |
| 704 | renderCellNode(buf, child) |
| 705 | buf.WriteString("*") |
| 706 | case tag == "del" || tag == "s": |
| 707 | buf.WriteString("~~") |
| 708 | renderCellNode(buf, child) |
| 709 | buf.WriteString("~~") |
| 710 | case tag == "code": |
| 711 | buf.WriteString("`") |
| 712 | buf.WriteString(getTextContent(child)) |
| 713 | buf.WriteString("`") |
| 714 | case tag == "a": |
| 715 | href := getAttr(child, "href") |
| 716 | buf.WriteString("[") |
| 717 | renderCellNode(buf, child) |
| 718 | buf.WriteString("](") |
| 719 | buf.WriteString(href) |
| 720 | buf.WriteString(")") |
| 721 | case tag == "br": |
| 722 | buf.WriteString("<br/>") |
| 723 | case tag == "p": |
| 724 | // Unwrap <p> inside cells |
| 725 | renderCellNode(buf, child) |
| 726 | case tag == "div": |
| 727 | renderCellNode(buf, child) |
| 728 | case strings.Contains(tag, "user"): |
| 729 | userKey := getAttr(child, "ri:userkey") |
| 730 | if userKey == "" { |
| 731 | userKey = getAttr(child, "userkey") |
| 732 | } |
| 733 | if userKey != "" { |
| 734 | fmt.Fprintf(buf, `<span data-user-key="%s"/>`, userKey) |
| 735 | } |
| 736 | case strings.Contains(tag, "ac:link"): |
| 737 | renderCellNode(buf, child) |
| 738 | case strings.Contains(tag, "image"): |
| 739 | // Handle images in cells |
| 740 | alt := getAttr(child, "ac:alt") |
| 741 | if alt == "" { |
| 742 | alt = getAttr(child, "alt") |
| 743 | } |
| 744 | var imgBuf bytes.Buffer |
| 745 | c := &xmlConverter{buf: &imgBuf} |
| 746 | ref := c.findImageRef(child) |
| 747 | if ref.isAttachment { |
| 748 | fmt.Fprintf(buf, `<span data-attachment="%s"`, ref.filename) |
| 749 | if alt != "" { |
| 750 | fmt.Fprintf(buf, ` data-alt="%s"`, alt) |
| 751 | } |
| 752 | buf.WriteString("/>") |
| 753 | } else if ref.url != "" { |
| 754 | buf.WriteString("![") |
| 755 | buf.WriteString(alt) |
| 756 | buf.WriteString("](") |
| 757 | buf.WriteString(ref.url) |
| 758 | buf.WriteString(")") |
| 759 | } |
| 760 | case strings.Contains(tag, "task-list"): |
| 761 | renderCellTaskList(buf, child) |
| 762 | case strings.Contains(tag, "emoticon"): |
| 763 | name := getAttr(child, "ac:name") |
| 764 | if name == "" { |
| 765 | name = getAttr(child, "name") |
| 766 | } |
| 767 | switch name { |
| 768 | case "plus": |
| 769 | buf.WriteString("(+)") |
| 770 | case "minus": |
| 771 | buf.WriteString("(-)") |
| 772 | case "question": |
| 773 | buf.WriteString("(?)") |
| 774 | case "tick": |
| 775 | buf.WriteString("(v)") |
| 776 | case "cross": |
| 777 | buf.WriteString("(x)") |
| 778 | } |
| 779 | case strings.Contains(tag, "inline-comment-marker"): |
| 780 | ref := getAttr(child, "ac:ref") |
| 781 | if ref == "" { |
| 782 | ref = getAttr(child, "ref") |
| 783 | } |
| 784 | if ref != "" { |
| 785 | fmt.Fprintf(buf, `<span data-inline-comment="%s">`, ref) |
| 786 | renderCellNode(buf, child) |
| 787 | buf.WriteString("</span>") |
| 788 | } else { |
| 789 | renderCellNode(buf, child) |
| 790 | } |
| 791 | default: |
| 792 | renderCellNode(buf, child) |
| 793 | } |
| 794 | } |
| 795 | } |
| 796 | } |
| 797 | |
| 798 | // renderCellTaskList renders a task list inside a table cell as inline markdown. |
| 799 | func renderCellTaskList(buf *bytes.Buffer, n *html.Node) { |
| 800 | for child := n.FirstChild; child != nil; child = child.NextSibling { |
| 801 | if child.Type != html.ElementNode { |
| 802 | continue |
| 803 | } |
| 804 | tag := strings.ToLower(child.Data) |
| 805 | if !strings.Contains(tag, "task") || strings.Contains(tag, "task-list") { |
| 806 | continue |
| 807 | } |
| 808 | // This is an ac:task element |
| 809 | status := "" |
| 810 | var bodyContent string |
| 811 | for tc := child.FirstChild; tc != nil; tc = tc.NextSibling { |
| 812 | if tc.Type != html.ElementNode { |
| 813 | continue |
| 814 | } |
| 815 | tcTag := strings.ToLower(tc.Data) |
| 816 | if strings.Contains(tcTag, "task-status") { |
| 817 | status = strings.TrimSpace(getTextContent(tc)) |
| 818 | } else if strings.Contains(tcTag, "task-body") { |
| 819 | bodyContent = strings.TrimSpace(renderCellMarkdown(tc)) |
| 820 | } |
| 821 | } |
| 822 | check := "[ ]" |
| 823 | if status == "complete" { |
| 824 | check = "[x]" |
| 825 | } |
| 826 | fmt.Fprintf(buf, "- %s %s<br/>", check, bodyContent) |
| 827 | } |
| 828 | } |
| 829 | |
| 830 | type imageRef struct { |
| 831 | url string |
| 832 | filename string |
| 833 | isAttachment bool |
| 834 | } |
| 835 | |
| 836 | func (c *xmlConverter) findImageRef(n *html.Node) imageRef { |
| 837 | var ref imageRef |
| 838 | var walk func(*html.Node) |
| 839 | walk = func(node *html.Node) { |
| 840 | if node.Type == html.ElementNode { |
| 841 | tag := strings.ToLower(node.Data) |
| 842 | // <ri:url ri:value="..."/> |
| 843 | if strings.Contains(tag, "url") { |
| 844 | v := getAttr(node, "ri:value") |
| 845 | if v == "" { |
| 846 | v = getAttr(node, "value") |
| 847 | } |
| 848 | if v != "" { |
| 849 | ref.url = v |
| 850 | return |
| 851 | } |
| 852 | } |
| 853 | // <ri:attachment ri:filename="..."/> |
| 854 | if strings.Contains(tag, "attachment") { |
| 855 | f := getAttr(node, "ri:filename") |
| 856 | if f == "" { |
| 857 | f = getAttr(node, "filename") |
| 858 | } |
| 859 | if f != "" { |
| 860 | ref.filename = f |
| 861 | ref.isAttachment = true |
| 862 | return |
| 863 | } |
| 864 | } |
| 865 | } |
| 866 | for child := node.FirstChild; child != nil; child = child.NextSibling { |
| 867 | walk(child) |
| 868 | } |
| 869 | } |
| 870 | walk(n) |
| 871 | return ref |
| 872 | } |
| 873 | |
| 874 | func (c *xmlConverter) hasUserChild(n *html.Node) bool { |
| 875 | for child := n.FirstChild; child != nil; child = child.NextSibling { |
| 876 | if child.Type == html.ElementNode { |
| 877 | tag := strings.ToLower(child.Data) |
| 878 | if strings.Contains(tag, "user") { |
| 879 | return true |
| 880 | } |
| 881 | } |
| 882 | } |
| 883 | return false |
| 884 | } |
| 885 | |
| 886 | // Helper functions |
| 887 | |
| 888 | func findNode(n *html.Node, tag string) *html.Node { |
| 889 | if n.Type == html.ElementNode && n.Data == tag { |
| 890 | return n |
| 891 | } |
| 892 | for child := n.FirstChild; child != nil; child = child.NextSibling { |
| 893 | if found := findNode(child, tag); found != nil { |
| 894 | return found |
| 895 | } |
| 896 | } |
| 897 | return nil |
| 898 | } |
| 899 | |
| 900 | func getAttr(n *html.Node, key string) string { |
| 901 | for _, attr := range n.Attr { |
| 902 | attrKey := attr.Key |
| 903 | if attr.Namespace != "" { |
| 904 | attrKey = attr.Namespace + ":" + attr.Key |
| 905 | } |
| 906 | if attrKey == key { |
| 907 | return attr.Val |
| 908 | } |
| 909 | } |
| 910 | return "" |
| 911 | } |
| 912 | |
| 913 | // collapseWhitespace replaces runs of whitespace with a single space, |
| 914 | // preserving leading/trailing single space if original had whitespace there. |
| 915 | func collapseWhitespace(s string) string { |
| 916 | var buf strings.Builder |
| 917 | inWS := false |
| 918 | for _, r := range s { |
| 919 | if r == ' ' || r == '\t' || r == '\n' || r == '\r' { |
| 920 | if !inWS { |
| 921 | buf.WriteByte(' ') |
| 922 | inWS = true |
| 923 | } |
| 924 | } else { |
| 925 | buf.WriteRune(r) |
| 926 | inWS = false |
| 927 | } |
| 928 | } |
| 929 | return buf.String() |
| 930 | } |
| 931 | |
| 932 | // hasTaskStatus checks if a node contains a task-status element. |
| 933 | func hasTaskStatus(n *html.Node) bool { |
| 934 | for child := n.FirstChild; child != nil; child = child.NextSibling { |
| 935 | if child.Type == html.ElementNode { |
| 936 | tag := strings.ToLower(child.Data) |
| 937 | if strings.Contains(tag, "task-status") { |
| 938 | return true |
| 939 | } |
| 940 | } |
| 941 | } |
| 942 | return false |
| 943 | } |
| 944 | |
| 945 | // getCDATAContent retrieves content from preprocessed CDATA sections. |
| 946 | // It looks for <cdatacontent> elements and unescapes their text. |
| 947 | func getCDATAContent(n *html.Node) string { |
| 948 | var buf bytes.Buffer |
| 949 | var walk func(*html.Node) |
| 950 | walk = func(node *html.Node) { |
| 951 | if node.Type == html.ElementNode && node.Data == "cdatacontent" { |
| 952 | text := getTextContent(node) |
| 953 | buf.WriteString(htmlpkg.UnescapeString(text)) |
| 954 | return |
| 955 | } |
| 956 | if node.Type == html.TextNode { |
| 957 | buf.WriteString(node.Data) |
| 958 | } |
| 959 | for child := node.FirstChild; child != nil; child = child.NextSibling { |
| 960 | walk(child) |
| 961 | } |
| 962 | } |
| 963 | walk(n) |
| 964 | return buf.String() |
| 965 | } |
| 966 | |
| 967 | // extractAttrOrder returns a comma-separated list of short attribute names |
| 968 | // (e.g. "name,schema-version,macro-id") preserving the original order from the HTML node. |
| 969 | // The "ac:" prefix is stripped for brevity. |
| 970 | func extractAttrOrder(n *html.Node) string { |
| 971 | var names []string |
| 972 | for _, attr := range n.Attr { |
| 973 | key := attr.Key |
| 974 | if attr.Namespace != "" { |
| 975 | key = attr.Namespace + ":" + attr.Key |
| 976 | } |
| 977 | short := strings.TrimPrefix(key, "ac:") |
| 978 | names = append(names, short) |
| 979 | } |
| 980 | return strings.Join(names, ",") |
| 981 | } |
| 982 | |
| 983 | // isNextSiblingCode checks if the next non-whitespace sibling is a <code> element. |
| 984 | func isNextSiblingCode(n *html.Node) bool { |
| 985 | for s := n.NextSibling; s != nil; s = s.NextSibling { |
| 986 | if s.Type == html.TextNode && strings.TrimSpace(s.Data) == "" { |
| 987 | continue |
| 988 | } |
| 989 | return s.Type == html.ElementNode && strings.ToLower(s.Data) == "code" |
| 990 | } |
| 991 | return false |
| 992 | } |
| 993 | |
| 994 | // isPrevSiblingCode checks if the previous non-whitespace sibling is a <code> element. |
| 995 | func isPrevSiblingCode(n *html.Node) bool { |
| 996 | for s := n.PrevSibling; s != nil; s = s.PrevSibling { |
| 997 | if s.Type == html.TextNode && strings.TrimSpace(s.Data) == "" { |
| 998 | continue |
| 999 | } |
| 1000 | return s.Type == html.ElementNode && strings.ToLower(s.Data) == "code" |
| 1001 | } |
| 1002 | return false |
| 1003 | } |
| 1004 | |
| 1005 | func getTextContent(n *html.Node) string { |
| 1006 | var buf bytes.Buffer |
| 1007 | var walk func(*html.Node) |
| 1008 | walk = func(node *html.Node) { |
| 1009 | if node.Type == html.TextNode { |
| 1010 | buf.WriteString(node.Data) |
| 1011 | } |
| 1012 | for child := node.FirstChild; child != nil; child = child.NextSibling { |
| 1013 | walk(child) |
| 1014 | } |
| 1015 | } |
| 1016 | walk(n) |
| 1017 | return buf.String() |
| 1018 | } |
| 1019 | |