fix sub emote indices

TPP.Core/Chat/TwitchEventSubChat.cs

@@ -3,6 +3,7 @@
 using System.Collections.Immutable;
 using System.Diagnostics;
 using System.Linq;
+using System.Text;
 using System.Text.Json;
 using System.Text.RegularExpressions;
 using System.Threading;
@@ -525,6 +526,22 @@ await _overlayConnection.Send(new NewSubscriber
         }, CancellationToken.None);
     }
 
+    private static readonly Encoding Utf8NoBom = new UTF8Encoding(encoderShouldEmitUTF8Identifier: false);
+
+    /// Parsing the emotes of the message sadly is not trivial, see <see cref="ChannelSubscriptionMessage.Emote"/>.
+    /// Of course, we currently don't even need the indices and could just look up an emote's code by its ID using the
+    /// Twitch API (maybe we don't even need the emote code in the webapp?), but this saves us roundtrips to Twitch.
+    public static ImmutableList<EmoteOccurrence> ParseEmotes(ChannelSubscriptionMessage.Emote[] emotes, string text)
+    {
+        byte[] messageBytes = Utf8NoBom.GetBytes(text);
+        return emotes.Select(emote =>
+        {
+            int endExclusive = emote.End + 1;
+            string emoteCode = Utf8NoBom.GetString(messageBytes[emote.Begin..endExclusive]);
+            return new EmoteOccurrence(emote.Id, emoteCode);
+        }).ToImmutableList();
+    }
+
     private async Task ChannelSubscriptionMessageReceived(ChannelSubscriptionMessage channelSubscriptionMessage)
     {
         ChannelSubscriptionMessage.Event evt = channelSubscriptionMessage.Payload.Event;
@@ -544,9 +561,9 @@ private async Task ChannelSubscriptionMessageReceived(ChannelSubscriptionMessage
             SubscriptionAt: channelSubscriptionMessage.Metadata.MessageTimestamp,
             IsGift: false, // Resubscriptions are never gifts. Gifts are always "new" subscriptions, not continuations
             Message: evt.Message?.Text,
-            Emotes: (evt.Message?.Emotes ?? []).Select(e => new EmoteOccurrence(
-                    e.Id, evt.Message!.Text!.Substring(e.Begin, e.End - e.Begin), e.Begin, e.End))
-                .ToImmutableList()
+            Emotes: evt.Message?.Emotes is { Length: > 0 }
+                ? ParseEmotes(evt.Message.Emotes, evt.Message.Text!)
+                : []
         );
         ISubscriptionProcessor.SubResult subResult = await _subscriptionProcessor.ProcessSubscription(
             subscriptionInfo);

TPP.Core/Subscriptions.cs

@@ -9,10 +9,13 @@
 using TPP.Common;
 using TPP.Model;
 using TPP.Persistence;
+using TPP.Twitch.EventSub.Notifications;
 
 namespace TPP.Core
 {
-    public record EmoteOccurrence(string Id, string Code, int StartIndex, int EndIndex);
+    /// This used to also include <c>int StartIndex</c> and <c>int EndIndex</c>, but since those were unused and how
+    /// to count was ambiguous (see <see cref="ChannelSubscriptionMessage.Emote"/>), they were removed.
+    public record EmoteOccurrence(string Id, string Code);
 
     /// <summary>
     /// Information on a user subscription directly (not via a gift).

TPP.Twitch.EventSub/Notifications/ChannelSubscriptionMessage.cs

@@ -27,6 +27,19 @@ public record Condition(string BroadcasterUserId) : EventSub.Condition;
     /// <param name="Begin">The index of where the Emote starts in the text.</param>
     /// <param name="End">The index of where the Emote ends in the text.</param>
     /// <param name="Id">The emote ID.</param>
+    /// Some additional gotchas that are not mentioned in the official documentation:
+    /// - The positions are counted in what appears to be UTF-8 bytes
+    /// - The begin is 0-based and inclusive (unsurprising), but the end is also inclusive (kinda surprising).
+    /// An experiment showed that this string:
+    /// <code>test Kappa 🌿 BabyRage 🐍 PogChamp 🎅🏿 RaccAttack ♥ PraiseIt 12345678901234567890</code>
+    /// resulted in these emote indices:
+    /// <code>
+    /// {"begin": 5, "end": 9, "id": "25"},
+    /// {"begin": 16, "end": 23, "id": "22639"},
+    /// {"begin": 30, "end": 37, "id": "305954156"},
+    /// {"begin": 48, "end": 57, "id": "114870"},
+    /// {"begin": 63, "end": 70, "id": "38586"}
+    /// </code>
     public record Emote(
         int Begin,
         int End,

tests/TPP.Core.Tests/Chat/TwitchEventSubChatTest.cs

@@ -0,0 +1,38 @@
+using System.Collections.Immutable;
+using NUnit.Framework;
+using TPP.Core.Chat;
+using TPP.Twitch.EventSub.Notifications;
+
+namespace TPP.Core.Tests.Chat;
+
+[TestFixture]
+[TestOf(typeof(TwitchEventSubChat))]
+public class TwitchEventSubChatTest
+{
+    /// <summary>
+    /// The data comes from an actual subscription that was observed.
+    /// See also <see cref="ChannelSubscriptionMessage.Emote"/>
+    /// </summary>
+    [Test]
+    public void ParseSubscriptionMessageEmotes()
+    {
+        const string message = "test Kappa 🌿 BabyRage 🐍 PogChamp 🎅🏿 RaccAttack ♥ PraiseIt 12345678901234567890";
+        ChannelSubscriptionMessage.Emote[] emotes =
+        [
+            new(5, 9, "25"),
+            new(16, 23, "22639"),
+            new(30, 37, "305954156"),
+            new(48, 57, "114870"),
+            new(63, 70, "38586")
+        ];
+        ImmutableList<EmoteOccurrence> occurrences = TwitchEventSubChat.ParseEmotes(emotes, message);
+        ImmutableList<EmoteOccurrence> expected = [
+            new("25", "Kappa"),
+            new("22639", "BabyRage"),
+            new("305954156", "PogChamp"),
+            new("114870", "RaccAttack"),
+            new("38586", "PraiseIt")
+        ];
+        Assert.That(occurrences, Is.EquivalentTo(expected));
+    }
+}