@createiq/htmldiff 1.2.0-beta.7 → 1.2.0-beta.8

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@createiq/htmldiff",
-  "version": "1.2.0-beta.7",
+  "version": "1.2.0-beta.8",
   "description": "TypeScript port of htmldiff.net",
   "type": "module",
   "author": "Mathew Mannion <mathew.mannion@linklaters.com>",

package/src/HtmlDiff.ts

@@ -1119,6 +1119,31 @@ export default class HtmlDiff {
         continue
       }
 
+      // Never orphan-reject a match whose tokens are ALL HTML tags.
+      // Tag tokens are structural; rejecting `</strong>` / `</em>` as
+      // an orphan match between two content deletions merges the tag
+      // into the deletion, leaving the matching opener unclosed —
+      // browsers then auto-close the opener at the END of the
+      // deletion, producing visually-wrong output (e.g. the body of
+      // a section deletion rendered as bold-italic because the
+      // closing `</strong></em>` ended up after the body deletion
+      // rather than after the heading). The orphan threshold is
+      // designed for stray word matches between heavily-edited spans,
+      // not for formatting boundaries.
+      let allTags = true
+      for (let i = curr.startInNew; i < curr.endInNew; i++) {
+        if (!Utils.isTag(wordsForDiffNew[i])) {
+          allTags = false
+          break
+        }
+      }
+      if (allTags) {
+        yield curr
+        prev = curr
+        curr = next
+        continue
+      }
+
       let oldDistanceInChars = 0
       for (let i = prev.endInOld; i < next.startInOld; i++) {
         oldDistanceInChars += wordsForDiffOld[i].length

package/test/HtmlDiff.threeWay.spec.ts

@@ -260,6 +260,46 @@ describe('HtmlDiff.executeThreeWay (genesis-spine)', () => {
     })
   })
 
+  describe('orphan-match guard for structural tags', () => {
+    // Real regression from the live preview (Additional Condition
+    // Precedent in the 2002 ISDA Schedule): when CP deletes a section
+    // whose answer renders as an empty formatting shell —
+    //   <p data-html="x"><em><strong></strong></em></p>
+    // — the `</strong>` and `</em>` matches sit between two content
+    // deletions ("Heading. " before, body after). At
+    // WORD_ALIGNED_OPTIONS.orphanMatchThreshold=0.25 those structural
+    // matches were rejected as orphans, swallowed into the deletion
+    // span, and the browser auto-closed the openers AT THE END of
+    // the deletion — visually rendering the entire deletion as
+    // bold-italic. The orphan filter now exempts tag-only matches
+    // so structural boundaries always survive.
+
+    it('CP deletes section with em+strong heading + plain body — closers stay between heading and body', () => {
+      const genesis =
+        '<p data-html="x"><em><strong>Additional Condition Precedent. </strong></em>For the purposes of Section 2(a)(iii).</p>'
+      const cp = '<p data-html="x"><em><strong></strong></em></p>'
+      const me = genesis
+
+      const out = HtmlDiff.executeThreeWay(genesis, cp, me, WORD_ALIGNED_OPTIONS)
+
+      // </strong> appears BEFORE the body deletion — meaning the
+      // body sits outside the bold-italic wrap, not inside it.
+      const closeStrongIdx = out.indexOf('</strong>')
+      const bodyDelIdx = out.indexOf('For the purposes')
+      expect(closeStrongIdx).toBeGreaterThan(0)
+      expect(bodyDelIdx).toBeGreaterThan(closeStrongIdx)
+      // No `<strong>…<del>body` substring exists — confirm by exact
+      // shape too. Heading wraps in strong+em, body is a plain del.
+      expect(out).toBe(
+        '<p data-html="x"><em><strong>' +
+          "<del class='diffdel cp' data-author='cp'>Additional Condition Precedent. </del>" +
+          '</strong></em>' +
+          "<del class='diffdel cp' data-author='cp'>For the purposes of Section 2(a)(iii).</del>" +
+          '</p>'
+      )
+    })
+  })
+
   describe('first-turn fallback', () => {
     it('cp == genesis means CP made no changes — Me-only attribution', () => {
       // Common case: this is the first turn where the counterparty hasn't