Add max_token_length to chargroup tokenizer (#4911) (#4926)

Relates: elastic/elasticsearch#56860

Co-authored-by: Russ Cam <[email protected]>

Author: github-actions[bot]

src/Nest/Analysis/Tokenizers/CharGroupTokenizer.cs

@@ -1,9 +1,10 @@
-// Licensed to Elasticsearch B.V under one or more agreements.
+// Licensed to Elasticsearch B.V under one or more agreements.
 // Elasticsearch B.V licenses this file to you under the Apache 2.0 License.
 // See the LICENSE file in the project root for more information
 
-using System.Collections.Generic;
+using System.Collections.Generic;
 using System.Runtime.Serialization;
+using Elasticsearch.Net.Utf8Json;
 
 namespace Nest
 {
@@ -20,6 +21,16 @@ public interface ICharGroupTokenizer : ITokenizer
 		/// </summary>
 		[DataMember(Name ="tokenize_on_chars")]
 		IEnumerable<string> TokenizeOnCharacters { get; set; }
+
+		/// <summary>
+		/// The maximum token length. If a token is seen that exceeds this length then
+		/// it is split at <see cref="MaxTokenLength"/> intervals. Defaults to `255`.
+		/// <para />
+		/// Valid in Elasticsearch 7.9.0+
+		/// </summary>
+		[DataMember(Name = "max_token_length")]
+		[JsonFormatter(typeof(NullableStringIntFormatter))]
+		int? MaxTokenLength { get; set; }
 	}
 
 	/// <inheritdoc cref="ICharGroupTokenizer" />
@@ -31,6 +42,9 @@ public class CharGroupTokenizer : TokenizerBase, ICharGroupTokenizer
 
 		/// <inheritdoc cref="ICharGroupTokenizer.TokenizeOnCharacters" />
 		public IEnumerable<string> TokenizeOnCharacters { get; set; }
+
+		/// <inheritdoc cref="ICharGroupTokenizer.MaxTokenLength" />
+		public int? MaxTokenLength { get; set; }
 	}
 
 	/// <inheritdoc cref="ICharGroupTokenizer" />
@@ -40,6 +54,7 @@ public class CharGroupTokenizerDescriptor
 		protected override string Type => CharGroupTokenizer.TokenizerType;
 
 		IEnumerable<string> ICharGroupTokenizer.TokenizeOnCharacters { get; set; }
+		int? ICharGroupTokenizer.MaxTokenLength { get; set; }
 
 		/// <inheritdoc cref="ICharGroupTokenizer.TokenizeOnCharacters" />
 		public CharGroupTokenizerDescriptor TokenizeOnCharacters(params string[] characters) =>
@@ -48,5 +63,9 @@ public CharGroupTokenizerDescriptor TokenizeOnCharacters(params string[] charact
 		/// <inheritdoc cref="ICharGroupTokenizer.TokenizeOnCharacters" />
 		public CharGroupTokenizerDescriptor TokenizeOnCharacters(IEnumerable<string> characters) =>
 			Assign(characters, (a, v) => a.TokenizeOnCharacters = v);
+
+		/// <inheritdoc cref="ICharGroupTokenizer.MaxTokenLength" />
+		public CharGroupTokenizerDescriptor MaxTokenLength(int? maxTokenLength) =>
+			Assign(maxTokenLength, (a, v) => a.MaxTokenLength = v);
 	}
 }

tests/Tests/Analysis/Tokenizers/TokenizerTests.cs

@@ -337,6 +337,32 @@ public class CharGroupTests : TokenizerAssertionBase<CharGroupTests>
 			public override string Name => "char_group";
 		}
 
+		[SkipVersion("<7.9.0", "max_token_length introduced in 7.9.0")]
+		public class CharGroupMaxTokenLengthTests : TokenizerAssertionBase<CharGroupMaxTokenLengthTests>
+		{
+			private readonly string[] _chars = { "whitespace", "-", "\n" };
+
+			public override FuncTokenizer Fluent => (n, t) => t.CharGroup(n, e => e
+				.TokenizeOnCharacters(_chars)
+				.MaxTokenLength(255)
+			);
+
+			public override ITokenizer Initializer => new CharGroupTokenizer
+			{
+				TokenizeOnCharacters = _chars,
+				MaxTokenLength = 255
+			};
+
+			public override object Json => new
+			{
+				tokenize_on_chars = _chars,
+				type = "char_group",
+				max_token_length = 255
+			};
+
+			public override string Name => "char_group_max_token_length";
+		}
+
 		[SkipVersion("<7.7.0", "discard_punctuation introduced in 7.7.0")]
 		public class DiscardPunctuationTests : TokenizerAssertionBase<DiscardPunctuationTests>
 		{