From db2bcfe31f8412d9b61e855a4bbcab7cd2a20e81 Mon Sep 17 00:00:00 2001
From: dualfroz <me@dualfroz.com>
Date: Fri, 2 Oct 2026 16:50:12 +0200
Subject: [PATCH] Keep a closing bracket in the word when it closes one inside
 it

---
 .../tokenize/methods/03-whitespace/tokenize.js | 18 ++++++++++++++++++
 tests/one/tokenize/punctuation.test.js         | 16 ++++++++++++++++
 2 files changed, 34 insertions(+)

diff --git a/src/1-one/tokenize/methods/03-whitespace/tokenize.js b/src/1-one/tokenize/methods/03-whitespace/tokenize.js
index 576cf10..d6a541a 100644
--- a/src/1-one/tokenize/methods/03-whitespace/tokenize.js
+++ b/src/1-one/tokenize/methods/03-whitespace/tokenize.js
@@ -6,6 +6,21 @@ const isNumber = /[\p{Number}\p{Currency_Symbol}]/u
 const hasAcronym = /^[a-z]\.([a-z]\.)+/i
 const chillin = /[sn]['’]$/
 const isFullNumber = /^[(+\-]?\d+(th|st|nd|rd)?[)+\-]?$/
+const openers = { ')': '(', ']': '[', '}': '{' }
+
+// does this closing bracket close one inside the word, like 'foo(bar)'
+const closesInside = function (chars, start, end) {
+  const close = chars[end - 1]
+  let depth = 0
+  for (let i = start; i < end - 1; i += 1) {
+    if (chars[i] === openers[close]) {
+      depth += 1
+    } else if (chars[i] === close) {
+      depth -= 1
+    }
+  }
+  return depth > 0
+}
 
 const normalizePunctuation = function (str, model) {
   // quick lookup for allowed pre/post punctuation
@@ -68,6 +83,9 @@ const normalizePunctuation = function (str, model) {
     if ((c === '+' || c === ')') && isFullNumber.test(str.trim())) {
       break//done
     }
+    if (Object.hasOwn(openers, c) && closesInside(chars, start, end)) {
+      break//keep it
+    }
     // punctuation
     end--
   }
diff --git a/tests/one/tokenize/punctuation.test.js b/tests/one/tokenize/punctuation.test.js
index eaab3e5..ad84528 100644
--- a/tests/one/tokenize/punctuation.test.js
+++ b/tests/one/tokenize/punctuation.test.js
@@ -29,6 +29,22 @@ test('term punctuation', function (t) {
   t.end()
 })
 
+test('closing bracket inside a word', function (t) {
+  const arr = [
+    ['foo(bar) foo', 'foo(bar)', ' '],
+    ['foo[bar] foo', 'foo[bar]', ' '],
+    ['foo{bar} foo', 'foo{bar}', ' '],
+    ['f(x), y', 'f(x)', ', '],
+    ['(foo) bar', 'foo', ') '],
+  ]
+  arr.forEach(([str, text, post]) => {
+    const term = nlp(str).docs[0][0]
+    t.equal(term.text, text, here + `text "${str}"`)
+    t.equal(term.post, post, here + `post "${str}"`)
+  })
+  t.end()
+})
+
 test('modify existing punctuation', function (t) {
   const world = nlp.world()