From f1bcb6b99deb25a0a5dc3054c2be6312516dfa37 Mon Sep 17 00:00:00 2001
From: Greg Price <greg@zulip.com>
Date: Sat, 7 Dec 2024 01:09:03 -0800
Subject: [PATCH] wip emoji: Recognize word-aligned matches in ranking; TODO
 test

---
 lib/model/emoji.dart | 55 +++++++++++++++++++++-----------------------
 1 file changed, 26 insertions(+), 29 deletions(-)

diff --git a/lib/model/emoji.dart b/lib/model/emoji.dart
index 7aea5ff8e7..0cd03a0b61 100644
--- a/lib/model/emoji.dart
+++ b/lib/model/emoji.dart
@@ -331,7 +331,12 @@ enum EmojiMatchQuality {
   /// The query matches a prefix of the emoji name, but not the whole name.
   prefix,
 
-  /// The query matches somewhere in the emoji name, but not at the start.
+  /// The query matches starting at the start of a word in the emoji name,
+  /// but not the start of the whole name.
+  wordAligned,
+
+  /// The query matches somewhere in the emoji name,
+  /// but not at the start of any word.
   other;
 
   /// The best possible quality of match.
@@ -385,10 +390,8 @@ class EmojiAutocompleteView extends AutocompleteView<EmojiAutocompleteQuery, Emo
     // Compare sort_emojis in Zulip web:shared/src/typeahead.ts .
     //
     // Behavior differences we might copy, TODO:
-    //  * Web ranks matches starting at a word boundary ahead of
-    //    other non-prefix matches; we don't yet.
-    //  * Relatedly, web favors popular emoji only upon a word-aligned match.
     //  * Web ranks each name of a Unicode emoji separately.
+    //  * Web recognizes a word-aligned match starting after [ /-] as well as [_].
     //
     // Behavior differences that web should probably fix, TODO(web):
     //  * Among popular emoji with non-exact matches,
@@ -417,15 +420,15 @@ class EmojiAutocompleteView extends AutocompleteView<EmojiAutocompleteQuery, Emo
       ReactionType.unicodeEmoji => false,
     };
     return switch (matchQuality) {
-      EmojiMatchQuality.exact  => 0,
-      EmojiMatchQuality.prefix => isPopular ? 1 : isCustomEmoji ? 3 : 4,
-      // TODO word-boundary vs. not
-      EmojiMatchQuality.other  => isPopular ? 2 : isCustomEmoji ? 5 : 6,
+      EmojiMatchQuality.exact       => 0,
+      EmojiMatchQuality.prefix      => isPopular ? 1 : isCustomEmoji ? 3 : 5,
+      EmojiMatchQuality.wordAligned => isPopular ? 2 : isCustomEmoji ? 4 : 6,
+      EmojiMatchQuality.other       =>                 isCustomEmoji ? 7 : 8,
     };
   }
 
   /// The number of possible values returned by [_rankResult].
-  static const _numResultRanks = 7;
+  static const _numResultRanks = 9;
 
   static bool _isPopularEmoji(EmojiCandidate candidate) {
     return candidate.emojiType == ReactionType.unicodeEmoji
@@ -473,29 +476,23 @@ class EmojiAutocompleteQuery extends ComposeAutocompleteQuery {
   }
 
   EmojiMatchQuality? _matchName(String emojiName) {
+    // Compare query_matches_string_in_order in Zulip web:shared/src/typeahead.ts .
     // TODO(#1067) this assumes emojiName is already lower-case (and no diacritics)
-    if (emojiName == _adjusted)          return EmojiMatchQuality.exact;
-    if (emojiName.startsWith(_adjusted)) return EmojiMatchQuality.prefix;
-    if (_nameMatches(emojiName))         return EmojiMatchQuality.other;
-    return null;
-  }
-
-  // Compare query_matches_string_in_order in Zulip web:shared/src/typeahead.ts .
-  bool _nameMatches(String emojiName) {
-    const String separator = '_';
-
-    if (!_adjusted.contains(separator)) {
+    if (emojiName == _adjusted)              return EmojiMatchQuality.exact;
+    if (emojiName.startsWith(_adjusted))     return EmojiMatchQuality.prefix;
+    const String sep = '_';
+    if (emojiName.contains(sep + _adjusted)) return EmojiMatchQuality.wordAligned;
+    if (!_adjusted.contains(sep)) {
       // If the query is a single token (doesn't contain a separator),
-      // the match can be anywhere in the string.
-      return emojiName.contains(_adjusted);
+      // allow a match anywhere in the string, too.
+      if (emojiName.contains(_adjusted))     return EmojiMatchQuality.other;
+    } else {
+      // Otherwise, when there is a separator in the query,
+      // require the match to start at the start of a token.
+      // (E.g. for 'ab_cd_ef', query could be 'ab_c' or 'cd_ef',
+      // but not 'b_cd_ef'.)
     }
-
-    // If there is a separator in the query, then we
-    // require the match to start at the start of a token.
-    // (E.g. for 'ab_cd_ef', query could be 'ab_c' or 'cd_ef',
-    // but not 'b_cd_ef'.)
-    assert(!emojiName.startsWith(_adjusted)); // checked before calling this method
-    return emojiName.contains(separator + _adjusted);
+    return null;
   }
 
   @override