autocomplete: Identify when the user intends a channel link autocomplete

For this commit we temporarily intercept the query at the
AutocompleteField widget, to avoid invoking the widgets that are
still unimplemented. That lets us defer those widgets' logic to a
separate later commit.

Author: sm-sayedi

lib/model/autocomplete.dart

@@ -18,6 +18,18 @@ import 'narrow.dart';
 import 'store.dart';
 
 extension ComposeContentAutocomplete on ComposeContentController {
+  // To avoid spending a lot of time searching for autocomplete intents
+  // in long messages, we bound how far back we look for the intent's start.
+  int get _maxLookbackForAutocompleteIntent {
+    return 1 // intent character, e.g. "#"
+      + 2 // some optional characters e.g., "_" for silent mention or "**"
+
+      // Per the API doc, maxChannelNameLength is in Unicode code points.
+      // We walk the string by UTF-16 code units, and there might be one or two
+      // of those encoding each Unicode code point.
+      + 2 * store.maxChannelNameLength;
+  }
+
   AutocompleteIntent<ComposeAutocompleteQuery>? autocompleteIntent() {
     if (!selection.isValid || !selection.isNormalized) {
       // We don't require [isCollapsed] to be true because we've seen that
@@ -28,9 +40,7 @@ extension ComposeContentAutocomplete on ComposeContentController {
       return null;
     }
 
-    // To avoid spending a lot of time searching for autocomplete intents
-    // in long messages, we bound how far back we look for the intent's start.
-    final earliest = max(0, selection.end - 30);
+    final earliest = max(0, selection.end - _maxLookbackForAutocompleteIntent);
 
     if (selection.start < earliest) {
       // The selection extends to before any position we'd consider
@@ -48,6 +58,9 @@ extension ComposeContentAutocomplete on ComposeContentController {
       } else if (charAtPos == ':') {
         final match = _emojiIntentRegex.matchAsPrefix(textUntilCursor, pos);
         if (match == null) continue;
+      } else if (charAtPos == '#') {
+        final match = _channelLinkIntentRegex.matchAsPrefix(textUntilCursor, pos);
+        if (match == null) continue;
       } else {
         continue;
       }
@@ -66,6 +79,10 @@ extension ComposeContentAutocomplete on ComposeContentController {
         final match = _emojiIntentRegex.matchAsPrefix(textUntilCursor, pos);
         if (match == null) continue;
         query = EmojiAutocompleteQuery(match[1]!);
+      } else if (charAtPos == '#') {
+        final match = _channelLinkIntentRegex.matchAsPrefix(textUntilCursor, pos);
+        if (match == null) continue;
+        query = ChannelLinkAutocompleteQuery(match[1] ?? match[2]!);
       } else {
         continue;
       }
@@ -165,6 +182,61 @@ final RegExp _emojiIntentRegex = (() {
     + r')$');
 })();
 
+final RegExp _channelLinkIntentRegex = () {
+  // What's likely to come just before #channel syntax: the start of the string,
+  // whitespace, or punctuation. Letters are unlikely; in that case a GitHub-
+  // style "zulip/zulip-flutter#124" link might be intended (as on CZO where
+  // there's a custom linkifier for that).
+  //
+  // Only some punctuation, like "(", is actually likely here. We don't
+  // currently try to be specific about that, except we exclude "#" and "@"
+  // in order to allow typing "##channel" for the channel query "#channel"
+  // and "@#user" for the mention query "#user". See discussion:
+  //   https://chat.zulip.org/#narrow/channel/243-mobile-team/topic/channel.20autocomplete.3A.20channels.20with.20.22.23.22.20in.20name/near/2288883
+  const before = r'(?<=^|\s|\p{Punctuation})(?<![#@])';
+  // TODO(dart-future): Regexps in ES 2024 have a /v aka unicodeSets flag;
+  //   if Dart matches that, we could combine into one character class
+  //   meaning "whitespace and punctuation, except not `#` or `@`":
+  //     r'(?<=^|[[\s\p{Punctuation}]--[#@]])'
+
+  // For a channel name, server excludes a wide range of characters and
+  // code points in `\p{C}` major category, namely the minor catories `\p{Cc}`,
+  // `\p{Cs}`, and part of `\p{Cn}`.
+  //   - https://github.com/zulip/zulip/blob/1c28e96d0/zerver/lib/string_validation.py#L38-L55
+  //   - https://github.com/zulip/zulip/blob/1c28e96d0/zerver/lib/string_validation.py#L8-L26
+  //
+  // TODO: match the server constraints
+  const nameCharExclusions = r'\r\n';
+
+  // TODO(upstream): maybe use duplicate-named capture groups for better readability?
+  //   https://github.com/dart-lang/sdk/issues/61337
+  return RegExp(unicode: true,
+    before
+    + r'#'
+    // As Web, match both '#channel' and '#**channel'. In both cases, the raw
+    // query is going to be 'channel'. Matching the second case ('#**channel')
+    // is useful when the user selects a channel from the autocomplete list, but
+    // then starts pressing "backspace" to edit the query and choose another
+    // option, instead of clearing the entire query and starting from scratch.
+    + r'(?:'
+      // Case '#channel': right after '#', reject whitespace as well as '**'.
+      + r'(?!\s|\*\*)([^' + nameCharExclusions + r']*)'
+      + r'|'
+      // Case '#**channel': right after '#**', reject whitespace.
+      // Also, make sure that the remaining query doesn't contain '**',
+      // otherwise '#**channel**' (which is a completed channel link syntax) and
+      // any text followed by that will always match.
+      + r'\*\*(?!\s)'
+      + r'((?:'
+        + r'[^*' + nameCharExclusions + r']'
+        + r'|'
+        + r'\*[^*' + nameCharExclusions + r']'
+        + r'|'
+        + r'\*$'
+      + r')*)'
+    + r')$');
+}();
+
 /// The text controller's recognition that the user might want autocomplete UI.
 class AutocompleteIntent<QueryT extends AutocompleteQuery> {
   AutocompleteIntent({

lib/widgets/autocomplete.dart

@@ -45,7 +45,8 @@ class _AutocompleteFieldState<QueryT extends AutocompleteQuery, ResultT extends
   }
 
   void _handleControllerChange() {
-    final newQuery = widget.autocompleteIntent()?.query;
+    var newQuery = widget.autocompleteIntent()?.query;
+    if (newQuery is ChannelLinkAutocompleteQuery) newQuery = null; // TODO(#124)
     // First, tear down the old view-model if necessary.
     if (_viewModel != null
         && (newQuery == null

test/model/autocomplete_test.dart

@@ -76,12 +76,15 @@ void main() {
     ///
     /// For example, "~@chris^" means the text is "@chris", the selection is
     /// collapsed at index 6, and we expect the syntax to start at index 0.
-    void doTest(String markedText, ComposeAutocompleteQuery? expectedQuery) {
+    void doTest(String markedText, ComposeAutocompleteQuery? expectedQuery, {
+      int? maxChannelName,
+    }) {
       final description = expectedQuery != null
         ? 'in ${jsonEncode(markedText)}, query ${jsonEncode(expectedQuery.raw)}'
         : 'no query in ${jsonEncode(markedText)}';
       test(description, () {
-        final store = eg.store();
+        final store = eg.store(initialSnapshot:
+          eg.initialSnapshot(maxChannelNameLength: maxChannelName));
         final controller = ComposeContentController(store: store);
         final parsed = parseMarkedText(markedText);
         assert((expectedQuery == null) == (parsed.expectedSyntaxStart == null));
@@ -98,6 +101,7 @@ void main() {
 
     MentionAutocompleteQuery mention(String raw) => MentionAutocompleteQuery(raw, silent: false);
     MentionAutocompleteQuery silentMention(String raw) => MentionAutocompleteQuery(raw, silent: true);
+    ChannelLinkAutocompleteQuery channelLink(String raw) => ChannelLinkAutocompleteQuery(raw);
     EmojiAutocompleteQuery emoji(String raw) => EmojiAutocompleteQuery(raw);
 
     doTest('', null);
@@ -179,8 +183,13 @@ void main() {
     doTest('~@_Rodion Romanovich Raskolniko^', silentMention('Rodion Romanovich Raskolniko'));
     doTest('~@Родион Романович Раскольников^', mention('Родион Романович Раскольников'));
     doTest('~@_Родион Романович Раскольнико^', silentMention('Родион Романович Раскольнико'));
-    doTest('If @chris is around, please ask him.^', null); // @ sign is too far away from cursor
-    doTest('If @_chris is around, please ask him.^', null); // @ sign is too far away from cursor
+
+    // "@" sign can be (3 + 2 * maxChannelName) utf-16 code units
+    // away to the left of the cursor.
+    doTest('If ~@chris^ is around, please ask him.', mention('chris'), maxChannelName: 10);
+    doTest('If ~@_chris is^ around, please ask him.', silentMention('chris is'), maxChannelName: 10);
+    doTest('If @chris is around, please ask him.^', null, maxChannelName: 10);
+    doTest('If @_chris is around, please ask him.^', null, maxChannelName: 10);
 
     // Emoji (":smile:").
 
@@ -259,6 +268,90 @@ void main() {
     doTest(',~:^', emoji('')); doTest(',~:a^', emoji('a'));
     doTest('，~:^', emoji('')); doTest('，~:a^', emoji('a'));
     doTest('。~:^', emoji('')); doTest('。~:a^', emoji('a'));
+
+    // #channel links.
+
+    doTest('^#',    null);
+    doTest('^#abc', null);
+    doTest('#abc',  null); // (no cursor)
+
+    // Link syntax can be at the start of a string.
+    doTest('~#^',    channelLink(''));
+    doTest('~##^',   channelLink('#'));
+    doTest('~#abc^', channelLink('abc'));
+
+    // Link syntax can contain multiple words.
+    doTest('~#abc ^',    channelLink('abc '));
+    doTest('~#abc def^', channelLink('abc def'));
+
+    // Link syntax can come after a word or space.
+    doTest('xyz ~#abc^', channelLink('abc'));
+    doTest(' ~#abc^',    channelLink('abc'));
+
+    // Link syntax can come after punctuation…
+    doTest(':~#abc^', channelLink('abc'));
+    doTest('!~#abc^', channelLink('abc'));
+    doTest(',~#abc^', channelLink('abc'));
+    doTest('.~#abc^', channelLink('abc'));
+    doTest('(~#abc^', channelLink('abc')); doTest(')~#abc^', channelLink('abc'));
+    doTest('{~#abc^', channelLink('abc')); doTest('}~#abc^', channelLink('abc'));
+    doTest('[~#abc^', channelLink('abc')); doTest(']~#abc^', channelLink('abc'));
+    doTest('“~#abc^', channelLink('abc')); doTest('”~#abc^', channelLink('abc'));
+    doTest('«~#abc^', channelLink('abc')); doTest('»~#abc^', channelLink('abc'));
+    // … except for '#' and '@', because they start
+    // channel link and mention syntaxes, respectively.
+    doTest('~##abc^', channelLink('#abc'));
+    doTest('~@#abc^', mention('#abc'));
+
+    // Avoid interpreting as queries a URL or a common linkifier syntax.
+    doTest('https://example.com/docs#install^', null);
+    doTest('zulip/zulip-flutter#124^',          null);
+
+    // Query can't start with a space; channel names don't.
+    doTest('# ^',    null);
+    doTest('# abc^', null);
+
+    // Query shouldn't be multiple lines.
+    doTest('#\n^',   null); doTest('#a\n^',   null); doTest('#\na^',   null); doTest('#a\nb^',   null);
+    doTest('#\r^',   null); doTest('#a\r^',   null); doTest('#\ra^',   null); doTest('#a\rb^',   null);
+    doTest('#\r\n^', null); doTest('#a\r\n^', null); doTest('#\r\na^', null); doTest('#a\r\nb^', null);
+
+    // Query can contain a wide range of characters.
+    doTest('~#`^', channelLink('`')); doTest('~#a`b^', channelLink('a`b'));
+    doTest('~#"^', channelLink('"')); doTest('~#a"b^', channelLink('a"b'));
+    doTest('~#>^', channelLink('>')); doTest('~#a>b^', channelLink('a>b'));
+    doTest('~#&^', channelLink('&')); doTest('~#a&b^', channelLink('a&b'));
+    doTest('~#_^', channelLink('_')); doTest('~#a_b^', channelLink('a_b'));
+    doTest('~#*^', channelLink('*')); doTest('~#a*b^', channelLink('a*b'));
+
+    // Avoid interpreting already-entered `#**foo**` syntax as queries.
+    doTest('#**abc**^',     null);
+    doTest('#**abc** ^',    null);
+    doTest('#**abc** def^', null);
+
+    // Accept syntax like "#**foo" (as from the user finishing an autocomplete
+    // and then hitting backspace to edit it), but leave the "**" out of the query.
+    doTest('~#**^',        channelLink(''));
+    doTest('~#**abc^',     channelLink('abc'));
+    doTest('~#**abc ^',    channelLink('abc '));
+    doTest('~#**abc def^', channelLink('abc def'));
+    doTest('~#**ab*c^',    channelLink('ab*c'));
+    doTest('~#**abc*^',    channelLink('abc*'));
+    doTest('#** ^',     null);
+    doTest('#** abc^',  null);
+    doTest('#**a\n^',   null); doTest('#**\na^',   null); doTest('#**a\nb^',   null);
+    doTest('#**a\r^',   null); doTest('#**\ra^',   null); doTest('#**a\rb^',   null);
+    doTest('#**a\r\n^', null); doTest('#**\r\na^', null); doTest('#**a\r\nb^', null);
+
+    // "#" sign can be (3 + 2 * maxChannelName) utf-16 code units
+    // away to the left of the cursor.
+    doTest('check ~#**mobile dev^ team', channelLink('mobile dev'), maxChannelName: 5);
+    doTest('check ~#mobile dev t^eam', channelLink('mobile dev t'), maxChannelName: 5);
+    doTest('check #mobile dev te^am', null, maxChannelName: 5);
+    doTest('check #mobile dev team for more info^', null, maxChannelName: 5);
+    // '🙂' is 2 utf-16 code units.
+    doTest('check ~#**🙂🙂🙂🙂🙂^', channelLink('🙂🙂🙂🙂🙂'), maxChannelName: 5);
+    doTest('check #**🙂🙂🙂🙂🙂🙂^', null, maxChannelName: 5);
   });
 
   test('MentionAutocompleteView misc', () async {