Add lax mode to csv reader.

docs/Csv/Options.md

@@ -83,6 +83,20 @@ __Culture__
 The `CultureInfo` used when parsing primitive values. Defaults to
 `InvariantCulture`.
 
+__Style__
+
+Specifies the parsing mode to be used when reading a CSV file.
+
+*Standard*: This mode uses slightly modified RFC4180 parsing, that allows non-comma delimiters to be used. 
+Valid RFC 4180 files should parse as expected in this mode.
+
+*Escaped*: This mode uses escaping instead of quoting fields. 
+Any field delimiter, record delimiter (newline) or escape character in a field value will be escaped by a preceeding escape character.
+
+*Lax*: This mode uses a more lenient parsing mode that will parse malformed fields and avoid throwing an exception.
+This mode starts by parsing using the `Standard` style, and upon finding a closing quote will parse the remainder of the field
+as if it were unquoted.
+
 __OwnsReader__
 
 Indicates if the `CsvDataReader` owns the TextReader and should dispose it when complete. Defaults to true.

docs/Csv/Sylvan.Data.Csv.Releases.md

@@ -1,5 +1,9 @@
 # Sylvan.Data.Csv Release Notes
 
+_1.3.6_
+- Adds `CsvStyle.Lax` which allows parsing CVS files with invalid fields. In this mode, the parser will not produce exceptions, but will
+ do a "best effort" to parse invalid fields.
+
 _1.3.5_
 - Fixes a bug where fields could be incorrectly read when the final character was escaped when reading with `CsvStyle.Escaped`.
 

source/Sylvan.Data.Csv.Tests/CsvDataReaderTests.cs

@@ -5,6 +5,7 @@
 using System.Globalization;
 using System.IO;
 using System.Linq;
+using System.Numerics;
 using System.Text;
 using System.Threading.Tasks;
 using Xunit;
@@ -949,7 +950,7 @@ public void Binary2()
 	[InlineData("N,V\na\\\nb,c\n", "a\nb", "c")]
 	[InlineData("N,V\na\\\r\nb\n", "a\r\nb", "")]
 	[InlineData("N,V\na\\\r\nb", "a\r\nb", "")]
-	public void ImpliedQuote(string input, string a, string b)
+	public void EscapedStyle(string input, string a, string b)
 	{
 		using var reader = new StringReader(input);
 		var options =
@@ -1884,14 +1885,14 @@ public void FinalCharInCellIsEscaped()
 	public void EscapeEOF()
 	{
 		using var reader = new StringReader("\\");
-		
+
 		using var csv = CsvDataReader.Create(reader, new CsvDataReaderOptions
 		{
 			CsvStyle = CsvStyle.Escaped,
 			HasHeaders = false,
 			Escape = '\\',
 		});
-		Assert.Throws<CsvFormatException>(() => csv.Read());		
+		Assert.Throws<CsvFormatException>(() => csv.Read());
 	}
 
 	[Fact]
@@ -1921,7 +1922,7 @@ public void FinalCharInCellIsEscapeError()
 	[InlineData("\"a\"\"a\"\"a\"", true, "a\"a\"a")]
 	[InlineData("a\"a\"a", true, "a\"a\"a")]
 	[InlineData("a\"\"\"a", true, "a\"\"\"a")]
-	
+
 	[InlineData("\"a\"\"\"a\"", false, null)]
 	[InlineData("\"a\"a", false, null)]
 	[InlineData("\"a\"a\"a\"", false, null)]
@@ -1935,18 +1936,80 @@ public void Quotes(string data, bool valid, string expected)
 		var r = new StringReader("a,b,c\n" + data);
 		var csv = CsvDataReader.Create(r);
 
-		if (valid) {
+		if (valid)
+		{
 			csv.Read();
 			var value = csv.GetString(0);
 			Assert.Equal(expected, value);
-		} 
+		}
 		else
 		{
 			var ex = Assert.Throws<CsvFormatException>(() => csv.Read());
 			Assert.Equal(1, ex.RowNumber);
 		}
 	}
 
+	[Theory]
+	// these are valid, and parse the same as the non-lax test
+	[InlineData("a", "a")]
+	[InlineData("\"\"", "")]
+	[InlineData("\"\"\"\"", "\"")]
+	[InlineData("\"\"\"\"\"\"", "\"\"")]
+	[InlineData("\"a\"", "a")]
+	[InlineData("\"a\"\"a\"", "a\"a")]
+	[InlineData("\"a\"\"a\"\"a\"", "a\"a\"a")]
+	[InlineData("a\"a\"a", "a\"a\"a")]
+	[InlineData("a\"\"\"a", "a\"\"\"a")]
+	// these are invalid, but will still produce a string in lax mode.
+	[InlineData("\"a\"\"\"a\"", "a\"a\"")]
+	[InlineData("\"a\"a", "aa")]
+	[InlineData("\"a\"a\"a\"", "aa\"a\"")]
+	[InlineData("\"\"a", "a")]
+	[InlineData("\"\"a\"", "a\"")]
+	//[InlineData("\"\"\"", "\"")]
+	[InlineData("\"\"\"\"\"", "\"\"")]
+	// test when invalid fields exist at the end of a file.
+	public void LaxQuotesEnd(string data, string expected)
+	{
+		var r = new StringReader("a,b,c\n1,2,3\n" + data);
+		var opts = new CsvDataReaderOptions { CsvStyle = CsvStyle.Lax };
+		var csv = CsvDataReader.Create(r, opts);
+		csv.Read(); // skip the 1,2,3
+		csv.Read();
+		var value = csv.GetString(0);
+		Assert.Equal(expected, value);
+	}
+
+	[Theory]
+	// these are valid, and parse the same as the non-lax test
+	[InlineData("a", "a")]
+	[InlineData("\"\"", "")]
+	[InlineData("\"\"\"\"", "\"")]
+	[InlineData("\"\"\"\"\"\"", "\"\"")]
+	[InlineData("\"a\"", "a")]
+	[InlineData("\"a\"\"a\"", "a\"a")]
+	[InlineData("\"a\"\"a\"\"a\"", "a\"a\"a")]
+	[InlineData("a\"a\"a", "a\"a\"a")]
+	[InlineData("a\"\"\"a", "a\"\"\"a")]
+	// these are invalid, but will still produce a string in lax mode.
+	[InlineData("\"a\"\"\"a\"", "a\"a\"")]
+	[InlineData("\"a\"a", "aa")]
+	[InlineData("\"a\"a\"a\"", "aa\"a\"")]
+	[InlineData("\"\"a", "a")]
+	[InlineData("\"\"a\"", "a\"")]
+	[InlineData("\"\"\"", "\"\n4,5,6\n")]
+	[InlineData("\"\"\"\"\"", "\"\"\n4,5,6\n")]
+	// test when invalid fields exist in the middle of a file.
+	public void LaxQuotesMid(string data, string expected)
+	{
+		var r = new StringReader("a,b,c\n1,2,3\n" + data + "\n4,5,6\n");
+		var opts = new CsvDataReaderOptions { CsvStyle = CsvStyle.Lax };
+		var csv = CsvDataReader.Create(r, opts);
+		csv.Read(); // skip the 1,2,3
+		csv.Read();
+		var value = csv.GetString(0);
+		Assert.Equal(expected, value);
+	}
 
 #if NET6_0_OR_GREATER