From 123f5b717c9223520da75d022d30476a3348f57f Mon Sep 17 00:00:00 2001
From: =?UTF-8?q?Tadeusz=20So=C5=9Bnierz?= <tadeusz@sosnierz.com>
Date: Wed, 21 Jul 2021 16:19:45 +0200
Subject: [PATCH] Try not to break words when quoting original message during
 replies

---
 changelog.d/1428.feature     |  1 +
 spec/unit/TrimString.spec.js | 16 ++++++++++++++++
 src/util/TrimString.ts       | 35 ++++++++++++++++++++++++-----------
 3 files changed, 41 insertions(+), 11 deletions(-)
 create mode 100644 changelog.d/1428.feature

diff --git a/changelog.d/1428.feature b/changelog.d/1428.feature
new file mode 100644
index 000000000..fca637986
--- /dev/null
+++ b/changelog.d/1428.feature
@@ -0,0 +1 @@
+Truncate original messages more gently when replying
diff --git a/spec/unit/TrimString.spec.js b/spec/unit/TrimString.spec.js
index 72822c20a..ad35a81f0 100644
--- a/spec/unit/TrimString.spec.js
+++ b/spec/unit/TrimString.spec.js
@@ -17,4 +17,20 @@ describe("trimString", function() {
 
         done();
     });
+
+    it('should stop trimming at the word boundary if reasonable', (done) => {
+        const input = "this sentence is waaaaay too long";
+        const result = trimString(input, 20);
+        expect(result).toEqual('this sentence is');
+
+        done();
+    });
+
+    it('should give up looking for a word boundary if result would become too short', (done) => {
+        const input = "we're in Llanfairpwllgwyngyll";
+        const result = trimString(input, 24);
+        expect(result).toContain("we're in Llan");
+
+        done();
+    });
 });
diff --git a/src/util/TrimString.ts b/src/util/TrimString.ts
index 606b22aaa..58b87de27 100644
--- a/src/util/TrimString.ts
+++ b/src/util/TrimString.ts
@@ -14,20 +14,33 @@ See the License for the specific language governing permissions and
 limitations under the License.
 */
 
-// a unicode-aware substring(0, $x)
+// since trimRight()/trimEnd() may break unicode characters
+function trimTrailingWhitespace(input: string): string {
+    return input.replace(/\s*$/u, '');
+}
+
+// a unicode-aware substring(0, $x) that tries to not break words if possible
 export function trimString(input: string, maxLength: number): string {
-    const re = new RegExp(`^([\\s\\S]{0,${maxLength}})`, 'u');
+    const re = new RegExp(`^([\\s\\S]{0,${maxLength}})(\\p{L}?)`, 'u');
     const match = input.match(re);
 
-    let trimmed: string;
-    if (match) {
-        trimmed = match[1];
-    }
-    else {
+    if (!match) {
         // fallback to a dumb substring() if the regex failed for any reason
-        trimmed = input.substring(0, maxLength);
+        return trimTrailingWhitespace(input.substring(0, maxLength));
+    }
+
+    const trimmed = trimTrailingWhitespace(match[1]);
+
+    if (match[2]) {
+        // find as much as you can that is followed by a word boundary,
+        // shorter than what we have now, but at least 75% of the desired length
+        const smallerMatch = trimmed.match(/^([\s\S]*\S)\b[\s\S]/u);
+        const minLength = maxLength * 0.75;
+
+        if (smallerMatch && smallerMatch[1].length >= minLength) {
+            return smallerMatch[1];
+        }
     }
-    //
-    // trimRight()/trimEnd() may break unicode characters
-    return trimmed.replace(/\s*$/u, '');
+
+    return trimmed;
 }