Expose GetFieldSpan on IParser

Author: Rob-Hague

src/CsvHelper/CsvParser.cs

@@ -14,7 +14,6 @@
 using System.Globalization;
 using System.Runtime.CompilerServices;
 using System.Threading;
-using System.Configuration;
 
 namespace CsvHelper
 {

src/CsvHelper/CsvReader.cs

@@ -7,6 +7,7 @@
 using System.Diagnostics;
 using System.Linq;
 using System.Runtime.CompilerServices;
+using System.Runtime.InteropServices;
 using System.Text;
 using System.Threading.Tasks;
 
@@ -34,26 +35,26 @@ public FieldCache(int initialSize = 128, int maxFieldSize = 128)
 			entries = new Entry[size];
 		}
 
-		public string GetField(char[] buffer, int start, int length)
+		public string GetField(ReadOnlySpan<char> buffer)
 		{
-			if (length == 0)
+			if (buffer.IsEmpty)
 			{
 				return string.Empty;
 			}
 
-			if (length > maxFieldSize)
+			if (buffer.Length > maxFieldSize)
 			{
-				return new string(buffer, start, length);
+				return buffer.ToString();
 			}
 
-			var hashCode = GetHashCode(buffer, start, length);
+			var hashCode = GetHashCode(buffer);
 			ref var bucket = ref GetBucket(hashCode);
 			int i = bucket - 1;
 			while ((uint)i < (uint)entries.Length)
 			{
 				ref var entry = ref entries[i];
 
-				if (entry.HashCode == hashCode && entry.Value.AsSpan().SequenceEqual(new Span<char>(buffer, start, length)))
+				if (entry.HashCode == hashCode && entry.Value.AsSpan().SequenceEqual(buffer))
 				{
 					return entry.Value;
 				}
@@ -70,25 +71,31 @@ public string GetField(char[] buffer, int start, int length)
 			ref var reference = ref entries[count];
 			reference.HashCode = hashCode;
 			reference.Next = bucket - 1;
-			reference.Value = new string(buffer, start, length);
+			reference.Value = buffer.ToString();
 			bucket = count + 1;
 			count++;
 
 			return reference.Value;
 		}
 
 		[MethodImpl(MethodImplOptions.AggressiveInlining)]
-		private uint GetHashCode(char[] buffer, int start, int length)
+		private static uint GetHashCode(ReadOnlySpan<char> buffer)
 		{
 			unchecked
 			{
+#if NET6_0_OR_GREATER
+				HashCode hash = new();
+				hash.AddBytes(MemoryMarshal.AsBytes(buffer));
+				return (uint)hash.ToHashCode();
+#else
 				uint hash = 17;
-				for (var i = start; i < start + length; i++)
+				foreach (char c in buffer)
 				{
-					hash = hash * 31 + buffer[i];
+					hash = hash * 31 + c;
 				}
 
 				return hash;
+#endif
 			}
 		}
 

src/CsvHelper/FieldCache.cs

@@ -33,10 +33,21 @@ public interface IParser : IDisposable
 		/// <summary>
 		/// Gets the field at the specified index for the current row.
 		/// </summary>
-		/// <param name="index">The index.</param>
-		/// <returns>The field.</returns>
+		/// <param name="index">The index of the field in the current row.</param>
+		/// <returns>A <see cref="string"/> representing the field at the specified index in the current row.</returns>
 		string this[int index] { get; }
 
+		/// <summary>
+		/// Gets a span over the field at the specified index in the current row.
+		/// </summary>
+		/// <param name="index">The index of the field in the current row.</param>
+		/// <returns>A span representing the field at the specified index in the current row.</returns>
+		ReadOnlySpan<char> GetFieldSpan(int index)
+#if NET
+			=> this[index]
+#endif
+			;
+
 		/// <summary>
 		/// Gets the record for the current row. Note:
 		/// It is much more efficient to only get the fields you need. If
@@ -49,6 +60,16 @@ public interface IParser : IDisposable
 		/// </summary>
 		string RawRecord { get; }
 
+		/// <summary>
+		/// Gets a span over the raw record for the current row.
+		/// </summary>
+		ReadOnlySpan<char> RawRecordSpan
+#if NET
+			=> RawRecord;
+#else
+			{ get; }
+#endif
+
 		/// <summary>
 		/// Gets the CSV row the parser is currently on.
 		/// </summary>

src/CsvHelper/IParser.cs

@@ -1366,5 +1366,18 @@ public void RawRowCountWithSingleLineAndNoLineEndingTest()
 				Assert.Equal(1, parser.RawRow);
 			}
 		}
+
+		[Theory]
+		[InlineData(-1)]
+		[InlineData(2)]
+		public void Parser_IndexOutOfRangeException(int index)
+		{
+			using (var reader = new StringReader("1,2\r\n"))
+			using (var parser = new CsvParser(reader, CultureInfo.InvariantCulture))
+			{
+				Assert.True(parser.Read());
+				Assert.Throws<IndexOutOfRangeException>(() => parser[index]);
+			}
+		}
 	}
 }

tests/CsvHelper.Tests/CsvParserTests.cs

@@ -29,6 +29,8 @@ public class ParserMock : IParser, IEnumerable<string[]>
 
 		public string RawRecord => string.Empty;
 
+		public ReadOnlySpan<char> RawRecordSpan => ReadOnlySpan<char>.Empty;
+
 		public int Row => row;
 
 		public int RawRow => row;
@@ -41,6 +43,8 @@ public class ParserMock : IParser, IEnumerable<string[]>
 
 		public string this[int index] => record[index];
 
+		public ReadOnlySpan<char> GetFieldSpan(int index) => record[index];
+
 		public ParserMock() : this(new CsvConfiguration(CultureInfo.InvariantCulture)) { }
 
 		public ParserMock(CsvConfiguration configuration)

tests/CsvHelper.Tests/Mocks/ParserMock.cs

@@ -149,5 +149,36 @@ public void Read_AccessingParserRecordInBadDataFound_ThrowsParserException()
 
 			Assert.Throws<ParserException>(() => parser[1]);
 		}
+
+		[Fact]
+		public void ConsecutiveBadDataTest()
+		{
+			int badDataCount = 0;
+			var config = new CsvConfiguration(CultureInfo.InvariantCulture)
+			{
+				BadDataFound = _ => badDataCount++,
+				CacheFields = false,
+				ProcessFieldBufferSize = 4
+			};
+			// These 3 fields each use the processFieldBuffer.
+			// The test is to ensure consistency of the fields during a read,
+			// i.e. the memory that each field points to is not overwritten
+			// during the processing of the other fields in the same row.
+			string csv = "\"\"\"\",\"two\" \"2,\"three\" \"3\r\n"; // """","two" "2,"three" "3
+			using (var reader = new StringReader(csv)) 
+			using (var parser = new CsvParser(reader, config))
+			{
+				Assert.True(parser.Read());
+
+				Assert.Equal(3, parser.Count);
+				Assert.Equal("\"", parser.GetFieldSpan(0).ToString());
+				Assert.Equal("two \"2", parser.GetFieldSpan(1).ToString());
+				Assert.Equal("three \"3", parser.GetFieldSpan(2).ToString());
+				Assert.Equal("two \"2", parser.GetFieldSpan(1).ToString());
+				Assert.Equal("\"", parser.GetFieldSpan(0).ToString());
+
+				Assert.False(parser.Read());
+			}
+		}
 	}
 }

tests/CsvHelper.Tests/Parsing/BadDataTests.cs

@@ -64,30 +64,23 @@ public void Read_WithFieldCacheDisabled_ReturnsDifferentFieldInstance()
 		[Fact]
 		public void Test1()
 		{
-			// "542008", "27721116", "98000820" have hash code 3769566006
-
 			var value1 = "542008";
 			var value2 = "27721116";
 			var value3 = "98000820";
 			var value4 = "542008";
 
 			var cache = new FieldCache(1);
 
-			var field1 = cache.GetField(value1.ToCharArray(), 0, value1.Length);
-			var field2 = cache.GetField(value2.ToCharArray(), 0, value2.Length);
-			var field3 = cache.GetField(value3.ToCharArray(), 0, value3.Length);
-			var field4 = cache.GetField(value4.ToCharArray(), 0, value4.Length);
+			var field1 = cache.GetField(value1);
+			var field2 = cache.GetField(value2);
+			var field3 = cache.GetField(value3);
+			var field4 = cache.GetField(value4);
 
 			Assert.Equal(value1, field1);
 			Assert.Equal(value2, field2);
 			Assert.Equal(value3, field3);
 			Assert.Equal(value4, field4);
 
-			Assert.NotSame(value1, field1);
-			Assert.NotSame(value2, field2);
-			Assert.NotSame(value3, field3);
-			Assert.NotSame(value4, field4);
-
 			Assert.Same(field1, field4);
 		}
 	}