fix(doc): preserve round-trip formatting in fetch output (larksuite#469)

* fix(doc): preserve round-trip formatting in fetch output

- trim leading spaces inside bold and italic emphasis exported by docs +fetch

- normalize nested list indentation to avoid flattening and literal text on re-import

- add regression tests for emphasis spacing and nested list indentation

* fix(doc): avoid false positives in markdown spacing fixes

- keep literal * x * and ** x ** text unchanged

- only normalize indented nested list markers when a parent list item exists

- add regression coverage for both CodeRabbit findings

* fix(doc): 修正嵌套列表缩进的空行误判

- 遇到空行时停止向上查找父级列表项，避免把 loose list sibling 误改成嵌套列表
- 避免把列表项中的四空格缩进代码块误改成 tab 缩进列表项
- 补充两个回归测试，并更新 fixBoldSpacing 注释使其与当前实现一致

* fix(doc): 修复 Markdown emphasis 空格回写

- 将 fixBoldSpacingLine 改为按星号 run 扫描，修复 ** hello **、* hello * 和同一行多个 italic span 的空格清理
- 保留 inline code、heading 和 *** hello** 这类近邻字面量，避免误改 emphasis nesting

Author: GaoSSR

shortcuts/doc/markdown_fix.go

@@ -6,6 +6,8 @@ package doc
 import (
 	"regexp"
 	"strings"
+	"unicode"
+	"unicode/utf8"
 )
 
 // fixExportedMarkdown applies post-processing to Lark-exported Markdown to
@@ -15,24 +17,29 @@ import (
 //     and strips redundant ** from ATX headings. Applied only outside fenced
 //     code blocks, and skips inline code spans.
 //
-//  2. fixSetextAmbiguity: inserts a blank line before any "---" that immediately
+//  2. normalizeNestedListIndentation: rewrites space-pair-indented nested list
+//     markers to tab-indented markers. This avoids nested ordered list items
+//     being flattened or interpreted as plain text/code on re-import.
+//
+//  3. fixSetextAmbiguity: inserts a blank line before any "---" that immediately
 //     follows a non-empty line, preventing it from being parsed as a Setext H2.
 //     Applied only outside fenced code blocks.
 //
-//  3. fixBlockquoteHardBreaks: inserts a blank blockquote line (">") between
+//  4. fixBlockquoteHardBreaks: inserts a blank blockquote line (">") between
 //     consecutive blockquote content lines so create-doc preserves line breaks.
 //     Applied only outside fenced code blocks.
 //
-//  4. fixTopLevelSoftbreaks: inserts a blank line between adjacent non-empty
+//  5. fixTopLevelSoftbreaks: inserts a blank line between adjacent non-empty
 //     lines at the top level and inside content containers (callout,
 //     quote-container, lark-td). Code fences are left untouched, and
 //     consecutive list items / continuations are not separated.
 //
-//  5. fixCalloutEmoji: replaces named emoji aliases (e.g. emoji="warning") with
+//  6. fixCalloutEmoji: replaces named emoji aliases (e.g. emoji="warning") with
 //     actual Unicode emoji characters that create-doc understands. Applied only
 //     outside fenced code blocks.
 func fixExportedMarkdown(md string) string {
 	md = applyOutsideCodeFences(md, fixBoldSpacing)
+	md = applyOutsideCodeFences(md, normalizeNestedListIndentation)
 	md = applyOutsideCodeFences(md, fixSetextAmbiguity)
 	md = applyOutsideCodeFences(md, fixBlockquoteHardBreaks)
 	md = fixTopLevelSoftbreaks(md)
@@ -106,20 +113,21 @@ func fixBlockquoteHardBreaks(md string) string {
 	return strings.Join(out, "\n")
 }
 
-// fixBoldSpacing fixes two issues with bold markers exported by Lark:
+// fixBoldSpacing normalizes emphasis markers exported by Lark while preserving
+// inline code spans:
+//
+//  1. Removes leading whitespace after opening ** and * delimiters:
+//     "** text**" → "**text**", "* text*" → "*text*"
 //
-//  1. Trailing whitespace before closing **: "**text **" → "**text**"
-//     CommonMark requires no space before a closing delimiter; otherwise the
-//     ** is rendered as literal text.
+//  2. Removes trailing whitespace before closing ** and * delimiters:
+//     "**text **" → "**text**", "*text *" → "*text*"
 //
-//  2. Redundant bold in ATX headings: "# **text**" → "# text"
-//     Headings are already bold, so the inner ** is visually redundant and
-//     some renderers display the markers literally.
+//  3. Removes redundant bold around an entire ATX heading:
+//     "# **text**" → "# text"
 //
-// Both fixes skip inline code spans to avoid modifying literal code content.
+// The bold and italic spacing fixes only run on non-code segments so literal
+// code content is left unchanged.
 var (
-	boldTrailingSpaceRe   = regexp.MustCompile(`(\*\*\S[^*]*?)\s+(\*\*)`)
-	italicTrailingSpaceRe = regexp.MustCompile(`(\*\S[^*]*?)\s+(\*)`)
 	// headingBoldRe uses [^*]+ (no asterisks) to avoid mismatching headings
 	// that contain multiple disjoint bold spans such as "# **foo** and **bar**".
 	headingBoldRe = regexp.MustCompile(`(?m)^(#{1,6})\s+\*\*([^*]+)\*\*\s*$`)
@@ -182,38 +190,116 @@ func scanInlineCodeSpans(line string) [][2]int {
 // fixBoldSpacingLine applies bold/italic trailing-space fixes to a single line,
 // skipping content inside inline code spans to avoid corrupting literal code.
 // ATX heading lines are also skipped here because headingBoldRe in fixBoldSpacing
-// handles them separately and boldTrailingSpaceRe can misfire on headings with
-// multiple disjoint bold spans (e.g. "# **foo** and **bar**").
+// handles them separately, keeping heading-only normalization isolated from the
+// inline emphasis spacing scanner below.
 func fixBoldSpacingLine(line string) string {
 	if atxHeadingRe.MatchString(line) {
 		return line
 	}
 	spans := scanInlineCodeSpans(line)
 	if len(spans) == 0 {
-		line = boldTrailingSpaceRe.ReplaceAllString(line, "$1$2")
-		line = italicTrailingSpaceRe.ReplaceAllString(line, "$1$2")
-		return line
+		return fixEmphasisSpacingSegment(line)
 	}
 	var sb strings.Builder
 	pos := 0
 	for _, loc := range spans {
 		// Process the non-code segment before this inline code span.
 		seg := line[pos:loc[0]]
-		seg = boldTrailingSpaceRe.ReplaceAllString(seg, "$1$2")
-		seg = italicTrailingSpaceRe.ReplaceAllString(seg, "$1$2")
-		sb.WriteString(seg)
+		sb.WriteString(fixEmphasisSpacingSegment(seg))
 		// Preserve inline code span as-is.
 		sb.WriteString(line[loc[0]:loc[1]])
 		pos = loc[1]
 	}
 	// Remaining non-code segment after the last code span.
-	seg := line[pos:]
-	seg = boldTrailingSpaceRe.ReplaceAllString(seg, "$1$2")
-	seg = italicTrailingSpaceRe.ReplaceAllString(seg, "$1$2")
-	sb.WriteString(seg)
+	sb.WriteString(fixEmphasisSpacingSegment(line[pos:]))
 	return sb.String()
 }
 
+// fixEmphasisSpacingSegment trims only the whitespace immediately inside simple
+// *...* and **...** spans. It deliberately ignores runs of 3+ asterisks and
+// any candidate whose payload contains another asterisk so nested emphasis-like
+// text remains untouched. When both inner sides contain whitespace, single-rune
+// payloads are preserved as literal text (for example "* x *" and "** x **").
+func fixEmphasisSpacingSegment(seg string) string {
+	if !strings.Contains(seg, "*") {
+		return seg
+	}
+
+	var sb strings.Builder
+	pos := 0
+	for pos < len(seg) {
+		openStart, openEnd, ok := nextAsteriskRun(seg, pos)
+		if !ok {
+			sb.WriteString(seg[pos:])
+			break
+		}
+
+		sb.WriteString(seg[pos:openStart])
+
+		markerLen := openEnd - openStart
+		if markerLen != 1 && markerLen != 2 {
+			sb.WriteString(seg[openStart:openEnd])
+			pos = openEnd
+			continue
+		}
+
+		closeStart, closeEnd, ok := nextAsteriskRun(seg, openEnd)
+		if !ok || closeEnd-closeStart != markerLen {
+			sb.WriteString(seg[openStart:openEnd])
+			pos = openEnd
+			continue
+		}
+
+		payload := seg[openEnd:closeStart]
+		normalized, shouldNormalize := normalizeEmphasisPayload(payload)
+		if !shouldNormalize {
+			sb.WriteString(seg[openStart:closeEnd])
+			pos = closeEnd
+			continue
+		}
+
+		marker := seg[openStart:openEnd]
+		sb.WriteString(marker)
+		sb.WriteString(normalized)
+		sb.WriteString(marker)
+		pos = closeEnd
+	}
+	return sb.String()
+}
+
+func nextAsteriskRun(s string, start int) (runStart, runEnd int, ok bool) {
+	for i := start; i < len(s); i++ {
+		if s[i] != '*' {
+			continue
+		}
+		j := i
+		for j < len(s) && s[j] == '*' {
+			j++
+		}
+		return i, j, true
+	}
+	return 0, 0, false
+}
+
+func normalizeEmphasisPayload(payload string) (string, bool) {
+	trimmedLeft := strings.TrimLeftFunc(payload, unicode.IsSpace)
+	trimmed := strings.TrimRightFunc(trimmedLeft, unicode.IsSpace)
+	if trimmed == "" {
+		return payload, false
+	}
+
+	hasLeadingSpace := len(trimmedLeft) != len(payload)
+	hasTrailingSpace := len(trimmed) != len(trimmedLeft)
+	if !hasLeadingSpace && !hasTrailingSpace {
+		return payload, true
+	}
+
+	if hasLeadingSpace && hasTrailingSpace && utf8.RuneCountInString(trimmed) == 1 {
+		return payload, false
+	}
+	return trimmed, true
+}
+
 var setextRe = regexp.MustCompile(`(?m)^([^\n]+)\n(-{3,}\s*$)`)
 
 func fixSetextAmbiguity(md string) string {
@@ -291,6 +377,44 @@ var contentContainers = [][2]string{
 // indented (nested) items.
 var listItemRe = regexp.MustCompile(`^[ \t]*([-*+]|\d+[.)]) `)
 
+// nestedListIndentRe matches nested list item markers indented with pairs of
+// spaces. We rewrite those space pairs to tabs because some downstream
+// round-trip paths treat multi-space indented ordered items as flat items or
+// literal text, while tab indentation remains nested and avoids 4-space code
+// block ambiguity.
+var nestedListIndentRe = regexp.MustCompile(`^( {2,})([-*+]|\d+[.)]) `)
+
+func normalizeNestedListIndentation(md string) string {
+	lines := strings.Split(md, "\n")
+	for i, line := range lines {
+		matches := nestedListIndentRe.FindStringSubmatch(line)
+		if len(matches) != 3 {
+			continue
+		}
+		if !hasPreviousNonBlankListItem(lines, i) {
+			continue
+		}
+		indent := matches[1]
+		if len(indent)%2 != 0 {
+			continue
+		}
+		tabs := strings.Repeat("\t", len(indent)/2)
+		lines[i] = tabs + line[len(indent):]
+	}
+	return strings.Join(lines, "\n")
+}
+
+func hasPreviousNonBlankListItem(lines []string, index int) bool {
+	for i := index - 1; i >= 0; i-- {
+		trimmed := strings.TrimSpace(lines[i])
+		if trimmed == "" {
+			return false
+		}
+		return listItemRe.MatchString(lines[i])
+	}
+	return false
+}
+
 // isListItemOrContinuation returns true for lines that are part of a list:
 // either a list item marker line or an indented continuation of a list item.
 // This is used to prevent blank lines being inserted between tight list lines,

shortcuts/doc/markdown_fix_test.go

@@ -14,6 +14,56 @@ func TestFixBoldSpacing(t *testing.T) {
 		input string
 		want  string
 	}{
+		{
+			name:  "leading space after opening bold",
+			input: "** hello**",
+			want:  "**hello**",
+		},
+		{
+			name:  "leading space after opening italic",
+			input: "* hello*",
+			want:  "*hello*",
+		},
+		{
+			name:  "leading and trailing spaces inside bold are collapsed",
+			input: "** hello **",
+			want:  "**hello**",
+		},
+		{
+			name:  "leading and trailing spaces inside italic are collapsed",
+			input: "* hello *",
+			want:  "*hello*",
+		},
+		{
+			name:  "multiple spaced italic spans on one line are each collapsed",
+			input: "* a* * b*",
+			want:  "*a* *b*",
+		},
+		{
+			name:  "ambiguous italic span stays literal",
+			input: "2 * x * y",
+			want:  "2 * x * y",
+		},
+		{
+			name:  "ambiguous bold span stays literal",
+			input: "2 ** x ** y",
+			want:  "2 ** x ** y",
+		},
+		{
+			name:  "single-rune italic with spaces on both sides stays literal",
+			input: "* x *",
+			want:  "* x *",
+		},
+		{
+			name:  "single-rune bold with spaces on both sides stays literal",
+			input: "** x **",
+			want:  "** x **",
+		},
+		{
+			name:  "triple-asterisk near miss stays literal",
+			input: "*** hello**",
+			want:  "*** hello**",
+		},
 		{
 			name:  "trailing space before closing bold",
 			input: "**hello **",
@@ -54,6 +104,16 @@ func TestFixBoldSpacing(t *testing.T) {
 			input: "**foo ** and `**bar **`",
 			want:  "**foo** and `**bar **`",
 		},
+		{
+			name:  "inline code with spaced italic stays literal while outside span is fixed",
+			input: "`* hello *` and * hello *",
+			want:  "`* hello *` and *hello*",
+		},
+		{
+			name:  "opening space inside text tag fixed",
+			input: `<text color="red">** Helpful - 有用性：**</text>`,
+			want:  `<text color="red">**Helpful - 有用性：**</text>`,
+		},
 		{
 			name:  "double-backtick inline code not modified",
 			input: "``**hello **`` and **world **",
@@ -222,6 +282,53 @@ func TestFixTopLevelSoftbreaks(t *testing.T) {
 	}
 }
 
+func TestNormalizeNestedListIndentation(t *testing.T) {
+	tests := []struct {
+		name  string
+		input string
+		want  string
+	}{
+		{
+			name:  "nested ordered list uses tabs instead of space pairs",
+			input: "1. parent\n  1. child\n    1. grandchild",
+			want:  "1. parent\n\t1. child\n\t\t1. grandchild",
+		},
+		{
+			name:  "nested mixed list markers use tabs instead of space pairs",
+			input: "- parent\n  - child\n    1. grandchild",
+			want:  "- parent\n\t- child\n\t\t1. grandchild",
+		},
+		{
+			name:  "top-level list unchanged",
+			input: "1. parent\n2. sibling",
+			want:  "1. parent\n2. sibling",
+		},
+		{
+			name:  "indented top-level marker without parent list stays unchanged",
+			input: "paragraph\n\n  1. item",
+			want:  "paragraph\n\n  1. item",
+		},
+		{
+			name:  "blank-line-separated loose-list sibling stays unchanged",
+			input: "1. a\n\n  1. b",
+			want:  "1. a\n\n  1. b",
+		},
+		{
+			name:  "indented code block inside list item stays unchanged",
+			input: "- parent\n\n    1. code",
+			want:  "- parent\n\n    1. code",
+		},
+	}
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got := normalizeNestedListIndentation(tt.input)
+			if got != tt.want {
+				t.Errorf("normalizeNestedListIndentation(%q) = %q, want %q", tt.input, got, tt.want)
+			}
+		})
+	}
+}
+
 func TestFixExportedMarkdown(t *testing.T) {
 	// End-to-end: all fixes applied together
 	input := "# **Title**\nparagraph one\nparagraph two\n**bold **\n> q1\n> q2\nsome text\n---"