feat(doccontent): extract shared markdown canonicalization and SHA

docgov saves content hashes through the new leaf module so it no longer
imports docrevision. Existing docrevision exports stay as thin wrappers.

Co-authored-by: Cursor <cursoragent@cursor.com>

.doc/system-design/module-boundaries.yaml

@@ -114,12 +114,14 @@ imports:
     - docrevision
   docgov:
     - app/uow
+    - doccontent
     - docduty
-    - docrevision
     - fsutil
   docrevision:
     - app/uow
+    - doccontent
     - fsutil
+  doccontent: []
   freedoc:
     - app/uow
     - ctxmgr

internal/doccontent/canonicalize.go

@@ -0,0 +1,45 @@
+package doccontent
+
+import (
+	"crypto/sha256"
+	"encoding/hex"
+	"strings"
+
+	"golang.org/x/text/unicode/norm"
+)
+
+// CanonicalizeMarkdown applies LF, UTF-8 NFC, and a single trailing newline.
+// It must not trim trailing spaces or tabs: two trailing spaces are Markdown hard breaks.
+func CanonicalizeMarkdown(raw string) string {
+	s := strings.ReplaceAll(raw, "\r\n", "\n")
+	s = strings.ReplaceAll(s, "\r", "\n")
+	s = norm.NFC.String(s)
+	if s != "" && !strings.HasSuffix(s, "\n") {
+		s += "\n"
+	}
+	return s
+}
+
+// LegacyDestructiveCanonicalizeMarkdown reproduces the pre-fix canonicalizer that
+// stripped per-line trailing spaces/tabs. Used only to classify historical drift.
+func LegacyDestructiveCanonicalizeMarkdown(raw string) string {
+	s := strings.ReplaceAll(raw, "\r\n", "\n")
+	s = strings.ReplaceAll(s, "\r", "\n")
+	s = norm.NFC.String(s)
+	lines := strings.Split(s, "\n")
+	for i, line := range lines {
+		lines[i] = strings.TrimRightFunc(line, func(r rune) bool {
+			return r == ' ' || r == '\t'
+		})
+	}
+	s = strings.Join(lines, "\n")
+	if s != "" && !strings.HasSuffix(s, "\n") {
+		s += "\n"
+	}
+	return s
+}
+
+func SHA256Hex(content string) string {
+	sum := sha256.Sum256([]byte(content))
+	return hex.EncodeToString(sum[:])
+}

internal/doccontent/canonicalize_golden_test.go

@@ -0,0 +1,70 @@
+package doccontent_test
+
+import (
+	"strings"
+	"testing"
+
+	"chat-agent-system/internal/doccontent"
+)
+
+func TestCanonicalizeMarkdown_PreservesHardBreakDoubleSpaces(t *testing.T) {
+	raw := "第一行  \n第二行  \n普通行\n"
+	got := doccontent.CanonicalizeMarkdown(raw)
+	if !strings.Contains(got, "第一行  \n") || !strings.Contains(got, "第二行  \n") {
+		t.Fatalf("hard-break double spaces must be kept: %q", got)
+	}
+	if strings.Count(got, "  \n") != 2 {
+		t.Fatalf("want 2 hard-breaks, got %q", got)
+	}
+	if got != raw {
+		t.Fatalf("already-canonical hard-break markdown must be identity: got %q", got)
+	}
+}
+
+func TestCanonicalizeMarkdown_OnlyNormalizesLF_NFC_TrailingNewline(t *testing.T) {
+	in := "a\r\n b \t\n"
+	out := doccontent.CanonicalizeMarkdown(in)
+	if out != "a\n b \t\n" {
+		t.Fatalf("must keep trailing spaces/tabs and only normalize CRLF: got %q", out)
+	}
+}
+
+func TestLegacyDestructiveCanonicalize_TrimsTrailingSpaces(t *testing.T) {
+	raw := "第一行  \n第二行  \n"
+	legacy := doccontent.LegacyDestructiveCanonicalizeMarkdown(raw)
+	if strings.Contains(legacy, "  \n") {
+		t.Fatalf("legacy destructive canonicalizer must trim hard-breaks: %q", legacy)
+	}
+	if legacy != "第一行\n第二行\n" {
+		t.Fatalf("got %q", legacy)
+	}
+	if doccontent.SHA256Hex(legacy) == doccontent.SHA256Hex(doccontent.CanonicalizeMarkdown(raw)) {
+		t.Fatal("legacy and semantic-preserving hashes must differ for hard-break docs")
+	}
+}
+
+func TestCanonicalizeMarkdown_Project143HardBreakGolden(t *testing.T) {
+	var b strings.Builder
+	for i := 0; i < 5; i++ {
+		b.WriteString("条目")
+		b.WriteByte('0' + byte(i))
+		b.WriteString("  \n")
+	}
+	b.WriteString("结尾段落。\n")
+	raw := b.String()
+	canon := doccontent.CanonicalizeMarkdown(raw)
+	if canon != raw {
+		t.Fatalf("project143-shaped hard-breaks must survive: got %q", canon)
+	}
+	legacy := doccontent.LegacyDestructiveCanonicalizeMarkdown(raw)
+	if len(raw)-len(legacy) != 10 {
+		t.Fatalf("legacy must drop exactly 10 bytes (5×2 spaces), raw=%d legacy=%d", len(raw), len(legacy))
+	}
+}
+
+func TestDocrevisionWrappersShareCanonicalBytes(t *testing.T) {
+	raw := "第一行  \r\n第二行  \n"
+	if doccontent.CanonicalizeMarkdown(raw) == "" {
+		t.Fatal("empty canon")
+	}
+}

internal/docgov/lease.go

@@ -11,7 +11,7 @@ import (
 	"strings"
 	"time"
 
-	"chat-agent-system/internal/docrevision"
+	"chat-agent-system/internal/doccontent"
 	"chat-agent-system/internal/fsutil"
 )
 
@@ -382,8 +382,8 @@ func (e *GovernanceEngine) SaveDocumentContent(ctx context.Context, req ContentS
 	}
 
 	now := time.Now()
-	canon := docrevision.CanonicalizeMarkdown(req.MarkdownBody)
-	shaStr := docrevision.SHA256Hex(canon)
+	canon := doccontent.CanonicalizeMarkdown(req.MarkdownBody)
+	shaStr := doccontent.SHA256Hex(canon)
 	doc.ContentSHA256 = shaStr
 	doc.IsTemplateOnly = false
 	doc.AuthorClaimed = true

internal/docrevision/canonicalize.go

@@ -1,49 +1,23 @@
 package docrevision
 
 import (
-	"crypto/sha256"
-	"encoding/hex"
 	"strings"
 	"unicode"
 	"unicode/utf8"
 
-	"golang.org/x/text/unicode/norm"
+	"chat-agent-system/internal/doccontent"
 )
 
-// CanonicalizeMarkdown applies LF, UTF-8 NFC, and a single trailing newline.
-// It must not trim trailing spaces or tabs: two trailing spaces are Markdown hard breaks.
 func CanonicalizeMarkdown(raw string) string {
-	s := strings.ReplaceAll(raw, "\r\n", "\n")
-	s = strings.ReplaceAll(s, "\r", "\n")
-	s = norm.NFC.String(s)
-	if s != "" && !strings.HasSuffix(s, "\n") {
-		s += "\n"
-	}
-	return s
+	return doccontent.CanonicalizeMarkdown(raw)
 }
 
-// LegacyDestructiveCanonicalizeMarkdown reproduces the pre-fix canonicalizer that
-// stripped per-line trailing spaces/tabs. Used only to classify historical drift.
 func LegacyDestructiveCanonicalizeMarkdown(raw string) string {
-	s := strings.ReplaceAll(raw, "\r\n", "\n")
-	s = strings.ReplaceAll(s, "\r", "\n")
-	s = norm.NFC.String(s)
-	lines := strings.Split(s, "\n")
-	for i, line := range lines {
-		lines[i] = strings.TrimRightFunc(line, func(r rune) bool {
-			return r == ' ' || r == '\t'
-		})
-	}
-	s = strings.Join(lines, "\n")
-	if s != "" && !strings.HasSuffix(s, "\n") {
-		s += "\n"
-	}
-	return s
+	return doccontent.LegacyDestructiveCanonicalizeMarkdown(raw)
 }
 
 func SHA256Hex(content string) string {
-	sum := sha256.Sum256([]byte(content))
-	return hex.EncodeToString(sum[:])
+	return doccontent.SHA256Hex(content)
 }
 
 func ContentFingerprint(s string) string {

internal/docrevision/canonicalize_golden_test.go

@@ -4,6 +4,7 @@ import (
 	"strings"
 	"testing"
 
+	"chat-agent-system/internal/doccontent"
 	"chat-agent-system/internal/docrevision"
 )
 
@@ -62,3 +63,16 @@ func TestCanonicalizeMarkdown_Project143HardBreakGolden(t *testing.T) {
 		t.Fatalf("legacy must drop exactly 10 bytes (5×2 spaces), raw=%d legacy=%d", len(raw), len(legacy))
 	}
 }
+
+func TestCanonicalizeWrappersMatchDoccontent(t *testing.T) {
+	raw := "第一行  \r\n第二行  \n"
+	if docrevision.CanonicalizeMarkdown(raw) != doccontent.CanonicalizeMarkdown(raw) {
+		t.Fatal("CanonicalizeMarkdown wrapper drifted")
+	}
+	if docrevision.LegacyDestructiveCanonicalizeMarkdown(raw) != doccontent.LegacyDestructiveCanonicalizeMarkdown(raw) {
+		t.Fatal("legacy wrapper drifted")
+	}
+	if docrevision.SHA256Hex(raw) != doccontent.SHA256Hex(raw) {
+		t.Fatal("SHA256Hex wrapper drifted")
+	}
+}