Autocomplete: Reopen @-mention typeahead when editing inside `**` by MritunjayTiwari14 · Pull Request #2166 · zulip/zulip-flutter

MritunjayTiwari14 · 2026-02-21T12:15:11Z

Fixes #1967.

Changes

Before	After
before.webm	after.webm

MritunjayTiwari14 · 2026-02-27T19:42:10Z

@chrisbobbe Ready for review.

Khader-1

Thanks @MritunjayTiwari14! I think this is close, a few notes bellow.

Khader-1 · 2026-03-07T19:16:44Z

I think this should be bumped to be + 3. The longest prefix after @ is now _** (3 chars, for a silent mention like @_**foo). With a small maxChannelNameLength, the lookback limit can prevent the scanner from reaching the @ and then the intent would be missed.

For example:
@_**ab with maxChannelName: 1, lookback is 1+2+2=5 but the pattern is 6 chars.

Yes, it does clearly make sense. This should be increased to 3 because of _**. Thanks for noticing!

Khader-1 · 2026-03-07T19:18:24Z

-      + r'[^'   + fullNameAndEmailCharExclusions + r']*'
+    + r'(?:'
+      // Reject on whitespace and "**" right after "@" or "@_". Emails can't start
+      // with both, and full_name can't either (it's run through Python's `.strip()`).


(nit):

Suggested change

// with both, and full_name can't either (it's run through Python's `.strip()`).

// with either, and full_name can't either (it's run through Python's `.strip()`).

Khader-1 · 2026-03-07T19:20:44Z

+    doTest('Ask @**Chris"Bobbe^', null);
+
+    // "**" are not allowed inside the query.
+    doTest('Ask @**Chris**Bobbe^', null);


I'd add more edge cases here, maybe:

// Three asterisks: "@***" = prefix "@**" + query "*". doTest('Ask ~@***^', mention('*')); doTest('Ask ~@_***^', silentMention('*')); doTest('~@_**ab^', silentMention('ab'), maxChannelName: 1); doTest('~@**abc^', mention('abc'), maxChannelName: 1); doTest('~@**ab^', mention('ab'), maxChannelName: 1);

Copilot

Pull request overview

This PR fixes the compose-box autocomplete behavior for @-mentions when a user edits inside the @**Full Name** syntax (e.g., after selecting a mention and then backspacing into the bolded name), so the typeahead reopens using the intended query text.

Changes:

Extend @-mention intent parsing to recognize @**…-prefixed mention text and exclude the leading ** from the query.
Adjust autocomplete lookback bounds to account for the longer optional prefix (@_**).
Expand and refine unit tests for @-mention and channel-link lookback/query parsing, including UTF-16 code-unit edge cases.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
`lib/model/autocomplete.dart`	Updates mention-intent regex and lookback length to support reopening autocomplete while editing inside `@**...` mention syntax.
`test/model/autocomplete_test.dart`	Adds comprehensive tests for `@**...` edit/backspace scenarios and updates lookback-boundary tests.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-08T06:34:38Z

+    doTest('Ask @****^', null);
+    doTest('Ask @_****^', null);
+
+    // " ' " (Aspotrophe) and "-" are allowed inside the query.


Spelling: "Aspotrophe" should be "Apostrophe" in this comment.

Suggested change

// " ' " (Aspotrophe) and "-" are allowed inside the query.

// " ' " (Apostrophe) and "-" are allowed inside the query.

MritunjayTiwari14 · 2026-03-31T11:03:08Z

@chrisbobbe Waiting for your review on this...

chrisbobbe

Thanks, and I'm sorry I've let this sit.

Comments below, including one that unlocks some simplification, I think: if we can continue prohibiting stars ("*") in the query, then I think we can neatly express three mutually exclusive branches in the regex:

no stars before or after the query
two stars before the query and no stars after
two stars before the query and one star after (for the case where you've backspaced through one star after finishing an autocomplete)

Would you give that a try?

chrisbobbe · 2026-04-01T06:29:44Z

The reasoning in #2166 (comment) for bumping the 2 to a 3 doesn't quite hold, I think. The max length of a channel name isn't relevant when you're trying to do a silent @-mention, because in that case you're searching for users, not channels.

I think the complexity of trying to maintain an exactly correct number here isn't worth it, and we should just do the simple thing of hard-coding it to 30 or 60 or something. This came up in another context recently: #2137 (comment)

Well, I disagree, maxlookback should be bumped from 2 to 3. I totally agree, channel name isn't relevant. But how the channel name is affected when the value is bumped from 2 to 3 is what I can't understand in your comment.

Those two are totally independent things, when we introduced silent mention. The min length would inflate from being **, i.e., 2, to _**, i.e., 3?

chrisbobbe · 2026-04-01T06:32:23Z

@@ -73,7 +73,7 @@ extension ComposeContentAutocomplete on ComposeContentController {
      if (charAtPos == '@') {
        final match = _mentionIntentRegex.matchAsPrefix(textUntilCursor, pos);
        if (match == null) continue;
-        query = MentionAutocompleteQuery(match[2]!, silent: match[1]! == '_');
+        query = MentionAutocompleteQuery(match[2] ?? match[3]!, silent: match[1]! == '_');


If I understand correctly, I think this could be simplified by using duplicate named capture groups? I see an existing TODO mentioning that feature:

// TODO(upstream): maybe use duplicate-named capture groups for better readability? // https://github.com/dart-lang/sdk/issues/61337

and that Dart PR was merged recently! 🎉 Can we try that?

chrisbobbe · 2026-04-01T06:38:40Z

+    doTest('Ask ~@_***^', silentMention('*'));
+    doTest('~@_**ab^', silentMention('ab'), maxChannelName: 1);
+    doTest('~@**abc^', mention('abc'), maxChannelName: 1);
+    doTest('~@**ab^', mention('ab'), maxChannelName: 1);


Is there a reason to allow "*" in the query at all? We have a comment suggesting that no users will ever match if that character (or a handful of others) is in the query:

// Characters that would defeat searches in full_name and emails, since // they're prohibited in both forms. These are all the characters prohibited // in full_name except "@", which appears in emails. (For the form of // full_name, find uses of UserProfile.NAME_INVALID_CHARS in zulip/zulip.) const fullNameAndEmailCharExclusions = r'\*`\\>"\p{Other}';

Fixes zulip#1967. RegEx logic is now divide into 3 mutually exclusive case where 1st case deals with query not starting with "**", while the 2nd and 3rd case handles query starting with "**" and, 1. Not containing any "*" afterwards. 2. Ending with an "*".

…dcoding Recently, dart added support for duplicate-named group. Using this would improve readability. Following this, we should use duplicate-named group instead of access the same query in different cases using hardcoded indices using former provides better readability. This was implemented by typecasting match to RegExpMatch. This seems safe since match was obtained by a regex operation and was required as only RegExpMatch supports `namedGroup('name')` method to access capture group named 'name'. notes: 1. Implemention of named capturing group was not done for capturing group which were accessible by single definite index, since there is only one capture group. more details about namedGroup: https://api.flutter.dev/flutter/dart-core/RegExpMatch/namedGroup.html dart upstream issue which was resolved: dart-lang/sdk#61337

MritunjayTiwari14 · 2026-04-11T06:46:32Z

Thanks, @chrisbobbe, for the review. I have left a comment on your feedback. As well as addressing all of your feedback. PTAL.

Changes:

Used a 3 mutually exclusive case-wise breakdown to identify both normal and silent mentions. Please check the commit description for more details.
Added a post-refactor commit to replace hardcoded RegExp group capture using namedGroup().
I have not updated the maxlookback value since the last review; it's still (4 + 2*maxChannelName).

gnprice added the maintainer review PR ready for review by Zulip maintainers label Feb 27, 2026

chrisbobbe self-assigned this Feb 27, 2026

Khader-1 reviewed Mar 7, 2026

View reviewed changes

MritunjayTiwari14 force-pushed the issue_1967 branch from 424de02 to 6b77919 Compare March 8, 2026 06:31

Copilot AI review requested due to automatic review settings March 8, 2026 06:31

Copilot started reviewing on behalf of MritunjayTiwari14 March 8, 2026 06:32

View session

Copilot AI reviewed Mar 8, 2026

View reviewed changes

MritunjayTiwari14 force-pushed the issue_1967 branch from 6b77919 to 81eda1a Compare March 8, 2026 11:15

chrisbobbe reviewed Apr 1, 2026

View reviewed changes

MritunjayTiwari14 added 2 commits April 11, 2026 12:06

MritunjayTiwari14 force-pushed the issue_1967 branch from 81eda1a to 50240ab Compare April 11, 2026 06:39

MritunjayTiwari14 requested a review from chrisbobbe April 11, 2026 06:47

MritunjayTiwari14 marked this pull request as draft April 14, 2026 06:02

MritunjayTiwari14 marked this pull request as ready for review April 17, 2026 14:05

chrisbobbe mentioned this pull request Apr 30, 2026

Topic link autocomplete #2137

Open

	// with both, and full_name can't either (it's run through Python's `.strip()`).
	// with either, and full_name can't either (it's run through Python's `.strip()`).

	// " ' " (Aspotrophe) and "-" are allowed inside the query.
	// " ' " (Apostrophe) and "-" are allowed inside the query.

Conversation

MritunjayTiwari14 commented Feb 21, 2026

Uh oh!

MritunjayTiwari14 commented Feb 27, 2026

Uh oh!

Khader-1 left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 8, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

MritunjayTiwari14 commented Mar 31, 2026

Uh oh!

chrisbobbe left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

MritunjayTiwari14 commented Apr 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Changes:

Uh oh!

Reviewers

Uh oh!

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

Khader-1 left a comment •

edited

Loading

MritunjayTiwari14 commented Apr 11, 2026 •

edited

Loading