Sep - the World's Fastest .NET CSV Parser

Modern, minimal, fast, zero allocation, reading and writing of separated values (csv, tsv etc.). Cross-platform, trimmable and AOT/NativeAOT compatible. Featuring an opinionated API design and pragmatic implementation targeted at machine learning use cases.

⭐ Please star this project if you like it. ⭐

🌃 Modern - utilizes features such as Span<T>, Generic Math (ISpanParsable<T>/ ISpanFormattable), ref struct, ArrayPool<T> and similar from .NET 7+ and C# 11+ for a modern and highly efficient implementation.

🔎 Minimal - a succinct yet expressive API with few options and no hidden changes to input or output. What you read/write is what you get. E.g. by default there is no "automatic" escaping/unescaping of quotes or trimming of spaces. To enable this see SepReaderOptions and Unescaping and Trimming. See SepWriterOptions for Escaping.

🚀 Fast - blazing fast with both architecture specific and cross-platform SIMD vectorized parsing incl. 64/128/256/512-bit paths e.g. AVX2, AVX-512 (.NET 8.0+), NEON. Uses csFastFloat for fast parsing of floating points. See detailed benchmarks for cross-platform results.

🌪️ Multi-threaded - unparalleled speed with highly efficient parallel CSV parsing that is up to 35x faster than CsvHelper, see ParallelEnumerate and benchmarks.

🌀 Async support - efficient ValueTask based async/await support. Requires C# 13.0+ and for .NET 9.0+ includes SepReader implementing IAsyncEnumerable<>. See Async Support for details.

🗑️ Zero allocation - intelligent and efficient memory management allowing for zero allocations after warmup incl. supporting use cases of reading or writing arrays of values (e.g. features) easily without repeated allocations.

✅ Thorough tests - great code coverage and focus on edge case testing incl. randomized fuzz testing.

🌐 Cross-platform - works on any platform, any architecture supported by NET. 100% managed and written in beautiful modern C#.

✂️ Trimmable and AOT/NativeAOT compatible - no problematic reflection or dynamic code generation. Hence, fully trimmable and Ahead-of-Time compatible. With a simple console tester program executable possible in just a few MBs. 💾

🗣️ Opinionated and pragmatic - conforms to the essentials of RFC-4180, but takes an opinionated and pragmatic approach towards this especially with regards to quoting and line ends. See section RFC-4180.

Example | Naming and Terminology | API | Limitations and Constraints | Comparison Benchmarks | Example Catalogue | RFC-4180 | FAQ | Public API Reference

Example

var text = """
           A;B;C;D;E;F
           Sep;🚀;1;1.2;0.1;0.5
           CSV;✅;2;2.2;0.2;1.5
           """;

using var reader = Sep.Reader().FromText(text);   // Infers separator 'Sep' (`;`) from header
using var writer = reader.Spec.Writer().ToText(); // Writer defined from reader 'Spec'
                                                  // Use .FromFile(...)/ToFile(...) for files
var idx = reader.Header.IndexOf("B");
var nms = new[] { "E", "F" };

foreach (var readRow in reader)           // Read one row at a time
{
    var a = readRow["A"].Span;            // Column as ReadOnlySpan<char>
    var b = readRow[idx].ToString();      // Column to string (might be pooled)
    var c = readRow["C"].Parse<int>();    // Parse any T : ISpanParsable<T>
    var d = readRow["D"].Parse<float>();  // Parse float/double fast via csFastFloat
    var s = readRow[nms].Parse<double>(); // Parse multiple columns as Span<T>
                                          // - Sep handles array allocation and reuse
    foreach (ref var v in s) { v *= 10; }

    using var writeRow = writer.NewRow(); // Start new row. Row written on Dispose.
    writeRow["A"].Set(a);                 // Set by ReadOnlySpan<char>
    writeRow["B"].Set(b);                 // Set by string
    writeRow["C"].Set($"{c * 2}");        // Set via InterpolatedStringHandler, no allocs
    writeRow["D"].Format(d / 2);          // Format any T : ISpanFormattable
    writeRow[nms].Format(s);              // Format multiple columns directly
    // Columns are added on first access as ordered, header written when first row written
}

var expected = """
               A;B;C;D;E;F
               Sep;🚀;2;0.6;1;5
               CSV;✅;4;1.1;2;15
               
               """;                       // Empty line at end is for line ending,
                                          // which is always written.
Assert.AreEqual(expected, writer.ToString());

// Above example code is for demonstration purposes only.
// Short names and repeated constants are only for demonstration.

For more examples, incl. how to write and read objects (e.g. records) with escape/unescape support, see Example Catalogue.

Naming and Terminology

Sep uses naming and terminology that is not based on RFC-4180, but is more tailored to usage in machine learning or similar. Additionally, Sep takes a pragmatic approach towards names by using short names and abbreviations where it makes sense and there should be no ambiguity given the context. That is, using Sep for Separator and Col for Column to keep code succinct.

Term	Description
Sep	Short for separator, also called delimiter. E.g. comma (,) is the separator for the separated values in a csv-file.
Header	Optional first row defining names of columns.
Row	A row is a collection of col(umn)s, which may span multiple lines. Also called record.
Col	Short for column, also called field.
Line	Horizontal set of characters until a line ending; \r\n, \r, \n.
Index	0-based that is RowIndex will be 0 for first row (or the header if present).
Number	1-based that is LineNumber will be 1 for the first line (as in notepad). Given a row may span multiple lines a row can have a From line number and a ToExcl line number matching the C# range indexing syntax [LineNumberFrom..LineNumberToExcl].

Application Programming Interface (API)

Besides being the succinct name of the library, Sep is both the main entry point to using the library and the container for a validated separator. That is, Sep is basically defined as:

public readonly record struct Sep(char Separator);

The separator char is validated upon construction and is guaranteed to be within a limited range and not being a char like " (quote) or similar. This can be seen in src/Sep/Sep.cs. The separator is constrained also for internal optimizations, so you cannot use any char as a separator.

⚠ Note that all types are within the namespace nietras.SeparatedValues and not Sep since it is problematic to have a type and a namespace with the same name.

To get started you can use Sep as the static entry point to building either a reader or writer. That is, for SepReader:

using var reader = Sep.Reader().FromFile("titanic.csv");

where .Reader() is a convenience method corresponding to:

using var reader = Sep.Auto.Reader().FromFile("titanic.csv");

where Sep? Auto => null; is a static property that returns null for a nullable Sep to signify that the separator should be inferred from the first row, which might be a header. If the first row does not contain any of the by default supported separators or there are no rows, the default separator will be used.

⚠ Note Sep uses ; as the default separator, since this is what was used in an internal proprietary library which Sep was built to replace. This is also to avoid issues with comma , being used as a decimal separator in some locales. Without having to resort to quoting.

If you want to specify the separator you can write:

using var reader = Sep.New(',').Reader().FromFile("titanic.csv");

or

var sep = new Sep(',');
using var reader = sep.Reader().FromFile("titanic.csv");

Similarly, for SepWriter:

using var writer = Sep.Writer().ToFile("titanic.csv");

or

using var writer = Sep.New(',').Writer().ToFile("titanic.csv");

where you have to specify a valid separator, since it cannot be inferred. To facilitate easy flow of the separator and CultureInfo both SepReader and SepWriter expose a Spec property of type SepSpec that simply defines those two. This means you can write:

using var reader = Sep.Reader().FromFile("titanic.csv");
using var writer = reader.Spec.Writer().ToFile("titanic-survivors.csv");

where the writer then will use the separator inferred by the reader, for example.

API Pattern

In general, both reading and writing follow a similar pattern:

Sep/Spec => SepReaderOptions => SepReader => Row => Col(s) => Span/ToString/Parse
Sep/Spec => SepWriterOptions => SepWriter => Row => Col(s) => Set/Format

where each continuation flows fluently from the preceding type. For example, Reader() is an extension method to Sep or SepSpec that returns a SepReaderOptions. Similarly, Writer() is an extension method to Sep or SepSpec that returns a SepWriterOptions.

SepReaderOptions and SepWriterOptions are optionally configurable. That and the APIs for reader and writer is covered in the following sections.

For a complete example, see the example above or the ReadMeTest.cs.

⚠ Note that it is important to understand that Sep Row/Col/Cols are ref structs (please follow the ref struct link and understand how this limits the usage of those). This is due to these types being simple facades or indirections to the underlying reader or writer. That means you cannot use LINQ or create an array of all rows like reader.ToArray(). While for .NET9+ the reader is now IEnumerable<> since ref structs can now be used in interfaces that have where T: allows ref struct this still does not mean it is LINQ compatible. Hence, if you need store per row state or similar you need to parse or copy to different types instead. The same applies to Col/Cols which point to internal state that is also reused. This is to avoid repeated allocations for each row and get the best possible performance, while still defining a well structured and straightforward API that guides users to relevant functionality. See Why SepReader Was Not IEnumerable Until .NET 9 and Is Not LINQ Compatible for more.

⚠ For a full overview of public types and methods see Public API Reference.

SepReader API

SepReader API has the following structure (in pseudo-C# code):

using var reader = Sep.Reader(o => o).FromFile/FromText/From...;
var header = reader.Header;
var _ = header.IndexOf/IndicesOf/NamesStartingWith...;
foreach (var row in reader)
{
    var _ = row[colName/colNames].Span/ToString/Parse<T>...;
    var _ = row[colIndex/colIndices].Span/ToString/Parse<T>...;
}

That is, to use SepReader follow the points below:

1. Optionally define Sep or use default automatically inferred separator.
2. Specify reader with optional configuration of SepReaderOptions. For example, if a csv-file does not have a header this can be configured via:

   Sep.Reader(o => o with { HasHeader = false })
   

   For all options see SepReaderOptions.
3. Specify source e.g. file, text (string), TextWriter, etc. via From extension methods.
4. Optionally access the header. For example, to get all columns starting with GT_ use:

   var colNames = header.NamesStartingWith("GT_");
   var colIndices = header.IndicesOf(colNames);
   

5. Enumerate rows. One row at a time.
6. Access a column by name or index. Or access multiple columns with names and indices. Sep internally handles pooled allocation and reuse of arrays for multiple columns.
7. Use Span to access the column directly as a ReadOnlySpan<char>. Or use ToString to convert to a string. Or use Parse<T> where T : ISpanParsable<T> to parse the column chars to a specific type.

SepReaderOptions

The following options are available:

/// <summary>
/// Specifies the separator used, if `null` then automatic detection 
/// is used based on first row in source.
/// </summary>
public Sep? Sep { get; init; } = null;
/// <summary>
/// Specifies initial internal `char` buffer length.
/// </summary>
/// <remarks>
/// The length will likely be rounded up to the nearest power of 2. A
/// smaller buffer may end up being used if the underlying source for <see
/// cref="System.IO.TextReader"/> is known to be smaller. Prefer to keep the
/// default length as that has been tuned for performance and cache sizes.
/// Avoid making this unnecessarily large as that will likely not improve
/// performance and may waste memory.
/// </remarks>
public int InitialBufferLength { get; init; } = SepDefaults.InitialBufferLength;
/// <summary>
/// Specifies the culture used for parsing. 
/// May be `null` for default culture.
/// </summary>
public CultureInfo? CultureInfo { get; init; } = SepDefaults.CultureInfo;
/// <summary>
/// Indicates whether the first row is a header row.
/// </summary>
public bool HasHeader { get; init; } = true;
/// <summary>
/// Specifies <see cref="IEqualityComparer{T}" /> to use 
/// for comparing header column names and looking up index.
/// </summary>
public IEqualityComparer<string> ColNameComparer { get; init; } = SepDefaults.ColNameComparer;
/// <summary>
/// Specifies the method factory used to convert a column span 
/// of `char`s to a `string`.
/// </summary>
public SepCreateToString CreateToString { get; init; } = SepToString.Direct;
/// <summary>
/// Disables using [csFastFloat](https://github.com/CarlVerret/csFastFloat)
/// for parsing `float` and `double`.
/// </summary>
public bool DisableFastFloat { get; init; } = false;
/// <summary>
/// Disables checking if column count is the same for all rows.
/// </summary>
public bool DisableColCountCheck { get; init; } = false;
/// <summary>
/// Disables detecting and parsing quotes.
/// </summary>
public bool DisableQuotesParsing { get; init; } = false;
/// <summary>
/// Unescape quotes on column access.
/// </summary>
/// <remarks>
/// When true, if a column starts with a quote then the two outermost quotes
/// are removed and every second inner quote is removed. Note that
/// unquote/unescape happens in-place, which means the <see
/// cref="SepReader.Row.Span" /> (<b>row</b> span) will be modified and
/// contain "garbage" state after unescaped cols before next col. This is
/// for efficiency to avoid allocating secondary memory for unescaped
/// columns. That is, the column span <see cref="SepReader.Col.Span" /> will
/// be correct with unescaped text, but the row span <see
/// cref="SepReader.Row.Span" /> may have "garbage" state between columns if
/// any column had quotes removed. Header columns/names will also be
/// unescaped. Requires <see cref="DisableQuotesParsing"/> to be false.
/// </remarks>
public bool Unescape { get; init; } = false;
/// <summary>
/// Option for trimming spaces (` ` - ASCII 32) on column access.
/// </summary>
/// <remarks>
/// By default no trimming is done. See <see cref="SepTrim"/> for options.
/// Note that trimming may happen in-place e.g. if also unescaping, which
/// means the <see cref="SepReader.Row.Span" /> will be modified and contain
/// "garbage" state for trimmed/unescaped cols. This is for efficiency to
/// avoid allocating secondary memory for trimmed/unescaped columns. That
/// is, the column span <see cref="SepReader.Col.Span" /> will be correct
/// with trimmed text, but the row span <see cref="SepReader.Row.Span" />
/// may have "garbage" state between columns if any column had spaces
/// removed. Header columns/names will also be trimmed. Note that only the
/// space ` ` (ASCII 32) character is trimmed, not any whitespace character.
/// </remarks>
public SepTrim Trim { get; init; } = SepTrim.None;
/// <summary>
/// Forwarded to <see
/// cref="System.Threading.Tasks.ValueTask.ConfigureAwait(bool)"/> or
/// similar when async methods are called.
/// </summary>
public bool AsyncContinueOnCapturedContext { get; init; } = false;

Unescaping

While great care has been taken to ensure Sep unescaping of quotes is both correct and fast, there is always the question of how does one respond to invalid input.

The below table tries to summarize the behavior of Sep vs CsvHelper and Sylvan. Note that all do the same for valid input. There are differences for how invalid input is handled. For Sep the design choice has been based on not wanting to throw exceptions and to use a principle that is both reasonably fast and simple.

Input	Valid	CsvHelper	CsvHelper¹	Sylvan	Sep²
a	True	a	a	a	a
""	True
""""	True	"	"	"	"
""""""	True	""	""	""	""
"a"	True	a	a	a	a
"a""a"	True	a"a	a"a	a"a	a"a
"a""a""a"	True	a"a"a	a"a"a	a"a"a	a"a"a
a""a	False	EXCEPTION	a""a	a""a	a""a
a"a"a	False	EXCEPTION	a"a"a	a"a"a	a"a"a
·""·	False	EXCEPTION	·""·	·""·	·""·
·"a"·	False	EXCEPTION	·"a"·	·"a"·	·"a"·
·""	False	EXCEPTION	·""	·""	·""
·"a"	False	EXCEPTION	·"a"	·"a"	·"a"
a"""a	False	EXCEPTION	a"""a	a"""a	a"""a
"a"a"a"	False	EXCEPTION	aa"a"	a"a"a	aa"a
""·	False	EXCEPTION	·	"	·
"a"·	False	EXCEPTION	a·	a"	a·
"a"""a	False	EXCEPTION	aa	EXCEPTION	a"a
"a"""a"	False	EXCEPTION	aa"	a"a<NULL>	a"a"
""a"	False	EXCEPTION	a"	"a	a"
"a"a"	False	EXCEPTION	aa"	a"a	aa"
""a"a""	False	EXCEPTION	a"a""	"a"a"	a"a"
"""	False			EXCEPTION	"
"""""	False	"	"	EXCEPTION	""

· (middle dot) is whitespace to make this visible

¹ CsvHelper with BadDataFound = null

² Sep with Unescape = true in SepReaderOptions

Trimming

Sep supports trimming by the SepTrim flags enum, which has two options as documented there. Below the result of both trimming and unescaping is shown in comparison to CsvHelper. Note unescaping is enabled for all results shown. It is possible to trim without unescaping too, of course.

As can be seen Sep supports a simple principle of trimming before and after unescaping with trimming before unescaping being important for unescaping if there is a starting quote after spaces.

Input	CsvHelper Trim	CsvHelper InsideQuotes	CsvHelper All¹	Sep Outer	Sep AfterUnescape	Sep All²
a	a	a	a	a	a	a
·a	a	·a	a	a	a	a
a·	a	a·	a	a	a	a
·a·	a	·a·	a	a	a	a
·a·a·	a·a	·a·a·	a·a	a·a	a·a	a·a
"a"	a	a	a	a	a	a
"·a"	·a	a	a	·a	a	a
"a·"	a·	a	a	a·	a	a
"·a·"	·a·	a	a	·a·	a	a
"·a·a·"	·a·a·	a·a	a·a	·a·a·	a·a	a·a
·"a"·	a	·"a"·	a	a	"a"	a
·"·a"·	·a	·"·a"·	a	·a	"·a"	a
·"a·"·	a·	·"a·"·	a	a·	"a·"	a
·"·a·"·	·a·	·"·a·"·	a	·a·	"·a·"	a
·"·a·a·"·	·a·a·	·"·a·a·"·	a·a	·a·a·	"·a·a·"	a·a

· (middle dot) is whitespace to make this visible

¹ CsvHelper with TrimOptions.Trim | TrimOptions.InsideQuotes

² Sep with SepTrim.All = SepTrim.Outer | SepTrim.AfterUnescape in SepReaderOptions

SepReader Debuggability

Debuggability is an important part of any library and in Sep. SepReader has a unique feature when looking at it and its rows or cols in a debug context. Given the following example code:

var text = """
           Key;Value
           A;"1
           2
           3"
           B;"Apple
           Banana
           Orange
           Pear"
           """;
using var reader = Sep.Reader().FromText(text);
foreach (var row in reader)
{
    // Hover over reader, row or col when breaking here
    var col = row[1];
    if (Debugger.IsAttached && row.RowIndex == 2) { Debugger.Break(); }
    Debug.WriteLine(col.ToString());
}

and you are hovering over reader when the break is triggered then this will show something like:

String Length=55

That is, it will show information of the source for the reader, in this case a string of length 55.

SepReader.Row Debuggability

If you are hovering over row then this will show something like:

  2:[5..9] = "B;\"Apple\r\nBanana\r\nOrange\r\nPear\""

This has the format shown below.

<ROWINDEX>:[<LINENUMBERRANGE>] = "<ROW>"

Note how this shows line number range [FromIncl..ToExcl], as in C# range expression, so that one can easily find the row in question in notepad or similar. This means Sep has to track line endings inside quotes and is an example of a feature that makes Sep a bit slower but which is a price considered worth paying.

GitHub doesn't show line numbers in code blocks so consider copying the example text to notepad or similar to see the effect.

Additionally, if you expand the row in the debugger (e.g. via the small triangle) you will see each column of the row similar to below.

00:'Key'   = "B"
01:'Value' = "\"Apple\r\nBanana\r\nOrange\r\nPear\""

SepReader.Col Debuggability

If you hover over col you should see:

"\"Apple\r\nBanana\r\nOrange\r\nPear\""

Why SepReader Was Not IEnumerable Until .NET 9 and Is Not LINQ Compatible

As mentioned earlier Sep only allows enumeration and access to one row at a time and SepReader.Row is just a simple facade or indirection to the underlying reader. This is why it is defined as a ref struct. In fact, the following code:

using var reader = Sep.Reader().FromText(text);
foreach (var row in reader)
{ }

can also be rewritten as:

using var reader = Sep.Reader().FromText(text);
while (reader.MoveNext())
{
    var row = reader.Current;
}

where row is just a facade for exposing row specific functionality. That is, row is still basically the reader underneath. Hence, let's look at using LINQ with SepReader implementing IEnumerable<SepReader.Row> and the Row not being a ref struct. Then, you would be able to write something like below:

using var reader = Sep.Reader().FromText(text);
SepReader.Row[] rows = reader.ToArray();

Given Row is just a facade for the reader, this would be equivalent to writing:

using var reader = Sep.Reader().FromText(text);
SepReader[] rows = reader.ToArray();

which hopefully makes it clear why this is not a good thing. The array would effectively be the reader repeated several times. If this would have to be supported one would have to allocate memory for each row always, which would basically be no different than a ReadLine approach as benchmarked in Comparison Benchmarks.

This is perhaps also the reason why no other efficient .NET CSV parser (known to author) implements an API pattern like Sep, but instead let the reader define all functionality directly and hence only lets you access the current row and cols on that. This API, however, is in this author's opinion not ideal and can be a bit confusing, which is why Sep is designed like it is. The downside is the above caveat.

The main culprit above is that for example ToArray() would store a ref struct in a heap allocated array, the actual enumeration is not a problem and hence implementing IEnumerable<SepReader.Row> is not the problem as such. The problem was that prior to .NET 9 it was not possible to implement this interface with T being a ref struct, but with C# 13 allows ref struct and .NET 9 having annotated such interfaces it is now possible and you can assign SepReader to IEnumerable, but most if not all of LINQ will still not work as shown below.

var text = """
           Key;Value
           A;1.1
           B;2.2
           """;
using var reader = Sep.Reader().FromText(text);
IEnumerable<SepReader.Row> enumerable = reader;
// Currently, most LINQ methods do not work for ref types. See below.
//
// The type 'SepReader.Row' may not be a ref struct or a type parameter
// allowing ref structs in order to use it as parameter 'TSource' in the
// generic type or method 'Enumerable.Select<TSource,
// TResult>(IEnumerable<TSource>, Func<TSource, TResult>)'
//
// enumerable.Select(row => row["Key"].ToString()).ToArray();

Calling Select should in principle be possible if this was annotated with allows ref struct, but it isn't currently.

If you want to use LINQ or similar you have to first parse or transform the rows into some other type and enumerate it. This is easy to do and instead of counting lines you should focus on how such enumeration can be easily expressed using C# iterators (aka yield return). With local functions this can be done inside a method like:

var text = """
           Key;Value
           A;1.1
           B;2.2
           """;
var expected = new (string Key, double Value)[] {
    ("A", 1.1),
    ("B", 2.2),
};

using var reader = Sep.Reader().FromText(text);
var actual = Enumerate(reader).ToArray();

CollectionAssert.AreEqual(expected, actual);

static IEnumerable<(string Key, double Value)> Enumerate(SepReader reader)
{
    foreach (var row in reader)
    {
        yield return (row["Key"].ToString(), row["Value"].Parse<double>());
    }
}

Now if instead refactoring this to something LINQ-compatible by defining a common Enumerate or similar method it could be:

var text = """
           Key;Value
           A;1.1
           B;2.2
           """;
var expected = new (string Key, double Value)[] {
    ("A", 1.1),
    ("B", 2.2),
};

using var reader = Sep.Reader().FromText(text);
var actual = Enumerate(reader,
    row => (row["Key"].ToString(), row["Value"].Parse<double>()))
    .ToArray();

CollectionAssert.AreEqual(expected, actual);

static IEnumerable<T> Enumerate<T>(SepReader reader, SepReader.RowFunc<T> select)
{
    foreach (var row in reader)
    {
        yield return select(row);
    }
}

In fact, Sep provides such a convenience extension method. And, discounting the Enumerate method, this does have less boilerplate, but not really more effective lines of code. The issue here is that this tends to favor factoring code in a way that can become very inefficient quickly. Consider if one wanted to only enumerate rows matching a predicate on Key which meant only 1% of rows were to be enumerated e.g.:

var text = """
           Key;Value
           A;1.1
           B;2.2
           """;
var expected = new (string Key, double Value)[] {
    ("B", 2.2),
};

using var reader = Sep.Reader().FromText(text);
var actual = reader.Enumerate(
    row => (row["Key"].ToString(), row["Value"].Parse<double>()))
    .Where(kv => kv.Item1.StartsWith('B'))
    .ToArray();

CollectionAssert.AreEqual(expected, actual);

This means you are still parsing the double (which is magnitudes slower than getting just the key) for all rows. Imagine if this was an array of floating points or similar. Not only would you then be parsing a lot of values you would also be allocated 99x arrays that aren't used after filtering with Where.

Instead, you should focus on how to express the enumeration in a way that is both efficient and easy to read. For example, the above could be rewritten as:

var text = """
           Key;Value
           A;1.1
           B;2.2
           """;
var expected = new (string Key, double Value)[] {
    ("B", 2.2),
};

using var reader = Sep.Reader().FromText(text);
var actual = Enumerate(reader).ToArray();

CollectionAssert.AreEqual(expected, actual);

static IEnumerable<(string Key, double Value)> Enumerate(SepReader reader)
{
    foreach (var row in reader)
    {
        var keyCol = row["Key"];
        if (keyCol.Span.StartsWith("B"))
        {
            yield return (keyCol.ToString(), row["Value"].Parse<double>());
        }
    }
}

To accommodate this Sep provides an overload for Enumerate that is similar to:

static IEnumerable<T> Enumerate<T>(this SepReader reader, SepReader.RowTryFunc<T> trySelect)
{
    foreach (var row in reader)
    {
        if (trySelect(row, out var value))
        {
            yield return value;
        }
    }
}

With this the above custom Enumerate can be replaced with:

var text = """
           Key;Value
           A;1.1
           B;2.2
           """;
var expected = new (string Key, double Value)[] {
    ("B", 2.2),
};

using var reader = Sep.Reader().FromText(text);
var actual = reader.Enumerate((SepReader.Row row, out (string Key, double Value) kv) =>
{
    var keyCol = row["Key"];
    if (keyCol.Span.StartsWith("B"))
    {
        kv = (keyCol.ToString(), row["Value"].Parse<double>());
        return true;
    }
    kv = default;
    return false;
}).ToArray();

CollectionAssert.AreEqual(expected, actual);

Note how this is pretty much the same length as the previous custom Enumerate. Also worse due to how C# requires specifying types for out parameters which then requires all parameter types for the lambda to be specified. Hence, in this case the custom Enumerate does not take significantly longer to write and is a lot more efficient than using LINQ .Where (also avoids allocating a string for key for each row) and is easier to debug and perhaps even read. All examples above can be seen in ReadMeTest.cs.

There is a strong case for having an enumerate API though and that is for parallelized enumeration, which will be discussed next.

ParallelEnumerate and Enumerate

As discussed in the previous section Sep provides Enumerate convenience extension methods, that should be used carefully. Alongside these there are ParallelEnumerate extension methods that provide very efficient multi-threaded enumeration. See benchmarks for numbers and Public API Reference.

ParallelEnumerate is built on top of LINQ AsParallel().AsOrdered() and will return exactly the same as Enumerate but with enumeration parallelized. This will use more memory during execution and as many threads as possible via the .NET thread pool. When using ParallelEnumerate one should, therefore (as always), be certain the provided delegate does not refer to or change any mutable state.

ParallelEnumerate comes with a lot of overhead compared to single-threaded foreach or Enumerate and should be used carefully based on measuring any potential benefit. Sep goes a long way to make this very efficient by using pooled arrays and parsing multiple rows in batches, but if the source only has a few rows then any benefit is unlikely.

Due to ParallelEnumerate being based on batches of rows it is also important not to "abuse" it in-place of LINQ AsParallel. The idea is to use it for parsing rows, not for doing expensive per row operations like loading an image or similar. In that case, you are better off using AsParallel() after ParallelEnumerate or Enumerate similarly to:

using var reader = Sep.Reader().FromFile("very-long.csv");
var results = reader.ParallelEnumerate(ParseRow)
                    .AsParallel().AsOrdered()
                    .Select(LoadData) // Expensive load
                    .ToList();

As a rule of thumb if the time per row exceeds 1 millisecond consider moving the expensive work to after ParallelEnumerate/Enumerate,

SepWriter API

SepWriter API has the following structure (in pseudo-C# code):

using var writer = Sep.Writer(o => o).ToFile/ToText/To...;
foreach (var data in EnumerateData())
{
    using var row = writer.NewRow();
    var _ = row[colName/colNames].Set/Format<T>...;
    var _ = row[colIndex/colIndices].Set/Format<T>...;
}

That is, to use SepWriter follow the points below:

1. Optionally define Sep or use default automatically inferred separator.
2. Specify writer with optional configuration of SepWriterOptions. For all options see SepWriterOptions.
3. Specify destination e.g. file, text (string via StringWriter), TextWriter, etc. via To extension methods.
4. SepWriter does allow you to define the header up front via SepWriter.Header.Add(...). However, header is usually defined by the order in which column names are accessed/created when defining the row.
5. Define new rows with NewRow. ⚠ Be sure to dispose any new rows before starting the next! For convenience Sep provides an overload for NewRow that takes a SepReader.Row and copies the columns from that row to the new row:

   using var reader = Sep.Reader().FromText(text);
   using var writer = reader.Spec.Writer().ToText();
   foreach (var readRow in reader)
   {   using var writeRow = writer.NewRow(readRow); }
   

6. Create a column by selecting by name or index. Or multiple columns via indices and names. Sep internally handles pooled allocation and reuse of arrays for multiple columns.
7. Use Set to set the column value either as a ReadOnlySpan<char>, string or via an interpolated string. Or use Format<T> where T : IFormattable to format T to the column value.
8. Row is written when Dispose is called on the row.

   Note this is to allow a row to be defined flexibly with both column removal, moves and renames in the future. This is not yet supported.

SepWriterOptions

The following options are available:

/// <summary>
/// Specifies the separator used.
/// </summary>
public Sep Sep { get; init; }
/// <summary>
/// Specifies the culture used for formatting. 
/// May be `null` for default culture.
/// </summary>
public CultureInfo? CultureInfo { get; init; }
/// <summary>
/// Specifies whether to write a header row 
/// before data rows. Requires all columns 
/// to have a name. Otherwise, columns can be
/// added by indexing alone.
/// </summary>
public bool WriteHeader { get; init; } = true;
/// <summary>
/// Disables checking if column count is the 
/// same for all rows.
/// </summary>
/// <remarks>
/// When true, the <see cref="ColNotSetOption"/>
/// will define how columns that are not set
/// are handled. For example, whether to skip
/// or write an empty column if a column has
/// not been set for a given row.
/// <para>
/// If any columns are skipped, then columns of
/// a row may, therefore, be out of sync with
/// column names if <see cref="WriteHeader"/>
/// is true.
/// </para>
/// As such, any number of columns can be
/// written as long as done sequentially.
/// </remarks>
public bool DisableColCountCheck { get; init; } = false;
/// <summary>
/// Specifies how to handle columns that are 
/// not set.
/// </summary>
public SepColNotSetOption ColNotSetOption { get; init; } = SepColNotSetOption.Throw;
/// <summary>
/// Specifies whether to escape column names 
/// and values when writing.
/// </summary>
/// <remarks>
/// When true, if a column contains a separator 
/// (e.g. `;`), carriage return (`\r`), line 
/// feed (`\n`) or quote (`"`) then the column 
/// is prefixed and suffixed with quotes `"` 
/// and any quote in the column is escaped by
/// adding an extra quote so it becomes `""`.
/// Note that escape applies to column names 
/// too, but only the written name.
/// </remarks>
public bool Escape { get; init; } = false;
/// <summary>
/// Forwarded to <see
/// cref="System.Threading.Tasks.ValueTask.ConfigureAwait(bool)"/> or
/// similar when async methods are called.
/// </summary>
public bool AsyncContinueOnCapturedContext { get; init; } = false;

Escaping

Escaping is not enabled by default in Sep, but when it is it gives the same results as other popular CSV libraries as shown below. Although, CsvHelper appears to be escaping spaces as well, which is not necessary.

Input	CsvHelper	Sylvan	Sep¹
``
·	"·"	·	·
a	a	a	a
;	";"	";"	";"
,	,	,	,
"	""""	""""	""""
\r	"\r"	"\r"	"\r"
\n	"\n"	"\n"	"\n"
a"aa"aaa	"a""aa""aaa"	"a""aa""aaa"	"a""aa""aaa"
a;aa;aaa	"a;aa;aaa"	"a;aa;aaa"	"a;aa;aaa"

Separator/delimiter is set to semi-colon ; (default for Sep)

· (middle dot) is whitespace to make this visible

\r, \n are carriage return and line feed special characters to make these visible

¹ Sep with Escape = true in SepWriterOptions

Async Support

Sep supports efficient ValueTask based asynchronous reading and writing.

However, given both SepReader.Row and SepWriter.Row are ref structs, as they point to internal state and should only be used one at a time, async/await usage is only supported on C# 13.0+ as this has support for "ref and unsafe in iterators and async methods" as covered in What's new in C# 13. Please consult details in that for limitations and constraints due to this.

Similarly, SepReader only implements IAsyncEnumerable<SepReader.Row> (and IEnumerable<SepReader.Row>) for .NET 9.0+/C# 13.0+ since then the interfaces have been annotated with allows ref struct for T.

Async support is provided on the existing SepReader and SepWriter types similar to how TextReader and TextWriter support both sync and async usage. This means you as a developer are responsible for calling async methods and using await when necessary. See below for a simple example and consult tests on GitHub for more examples.

var text = """
           A;B;C;D;E;F
           Sep;🚀;1;1.2;0.1;0.5
           CSV;✅;2;2.2;0.2;1.5
           
           """; // Empty line at end is for line ending

using var reader = await Sep.Reader().FromTextAsync(text);
await using var writer = reader.Spec.Writer().ToText();
await foreach (var readRow in reader)
{
    await using var writeRow = writer.NewRow(readRow);
}
Assert.AreEqual(text, writer.ToString());

Note how for SepReader the FromTextAsync is suffixed with Async to indicate async creation, this is due to the reader having to read the first row of the source at creation to determine both separator and, if file has a header, column names of the header. The From*Async call then has to be awaited. After that rows can be enumerated asynchronously simply by putting await before foreach. If one forgets to do that the rows will be enumerated synchronously.

For SepWriter the usage is kind of reversed. To* methods have no Async variants, since creation is synchronous. That is, StreamWriter is created by a simple constructor call. Nothing is written until a header or row is defined and Dispose/DisposeAsync is called on the row.

For reader nothing needs to be asynchronously disposed, so using does not require await. However, for SepWriter dispose may have to write/flush data to underlying TextWriter and hence it should be using DisposeAsync, so you must use await using.

To support cancellation many methods have overloads that accept a CancellationToken like the From*Async methods for creating a SepReader or for example NewRow for SepWriter. Consult Public API Reference for full set of available methods.

Additionally, both SepReaderOptions and SepWriterOptions feature the bool AsyncContinueOnCapturedContext option that is forwarded to internal ConfigureAwait calls, see the ConfigureAwait FAQ for details on that.

Limitations and Constraints

Sep is designed to be minimal and fast. As such, it has some limitations and constraints:

• Comments # are not directly supported. You can skip a row by:

  foreach (var row in reader)
  {
       // Skip row if starts with #
       if (!row.Span.StartsWith("#"))
       {
            // ...
       }
  }
  

  This does not allow skipping lines before a header row starting with # though. In Example Catalogue a full example is given detailing how to skip lines before header.

Comparison Benchmarks

To investigate the performance of Sep it is compared to:

• CsvHelper - the most commonly used CSV library with a staggering downloads on NuGet. Fully featured and battle tested.
• Sylvan - is well-known and has previously been shown to be the fastest CSV libraries for parsing (Sep changes that 😉).
• ReadLine/WriteLine - basic naive implementations that read line by line and split on separator. While writing columns, separators and line endings directly. Does not handle quotes or similar correctly.

All benchmarks are run from/to memory either with:

• StringReader or StreamReader + MemoryStream
• StringWriter or StreamWriter + MemoryStream

This to avoid confounding factors from reading from or writing to disk.

When using StringReader/StringWriter each char counts as 2 bytes, when measuring throughput e.g. MB/s. When using StreamReader/StreamWriter content is UTF-8 encoded and each char typically counts as 1 byte, as content usually limited to 1 byte per char in UTF-8. Note that in .NET for TextReader and TextWriter data is converted to/from char, but for reading such conversion can often be just as fast as Memmove.

By default only StringReader/StringWriter results are shown, if a result is based on StreamReader/StreamWriter it will be called out. Usually, results for StreamReader/StreamWriter are in line with StringReader/StringWriter but with half the throughput due to 1 byte vs 2 bytes. For brevity they are not shown here.

For all benchmark results, Sep has been defined as the Baseline in BenchmarkDotNet. This means Ratio will be 1.00 for Sep. For the others Ratio will then show how many times faster Sep is than that. Or how many times more bytes are allocated in Alloc Ratio.

Disclaimer: Any comparison made is based on a number of preconditions and assumptions. Sep is a new library written from the ground up to use the latest and greatest features in .NET. CsvHelper has a long history and has to take into account backwards compatibility and still supporting older runtimes, so may not be able to easily utilize more recent features. Same goes for Sylvan. Additionally, Sep has a different feature set compared to the two. Performance is a feature, but not the only feature. Keep that in mind when evaluating results.

Runtime and Platforms

The following runtime is used for benchmarking:

• NET 9.0.X

NOTE: Garbage Collection DATAS mode is disabled since this severely impacts (e.g. 1.7x slower) performance for some benchmarks due to the bursty accumulated allocations. That is, GarbageCollectionAdaptationMode is set to 0.

The following platforms are used for benchmarking:

• AMD EPYC 7763 (Virtual) X64 Platform Information

  OS=Ubuntu 22.04.5 LTS (Jammy Jellyfish)
  AMD EPYC 7763, 1 CPU, 4 logical and 2 physical cores
  

• AMD Ryzen 7 PRO 7840U (Laptop on battery) X64 Platform Information

  OS=Windows 11 (10.0.22631.4460/23H2/2023Update/SunValley3)
  AMD Ryzen 7 PRO 7840U w/ Radeon 780M Graphics, 1 CPU, 16 logical and 8 physical cores
  

• AMD 5950X (Desktop) X64 Platform Information (no longer available)

  OS=Windows 10 (10.0.19044.2846/21H2/November2021Update)
  AMD Ryzen 9 5950X, 1 CPU, 32 logical and 16 physical cores
  

• AMD 9950X (Desktop) X64 Platform Information

  OS=Windows 10 (10.0.19044.3086/21H2/November2021Update)
  AMD Ryzen 9 9950X, 1 CPU, 32 logical and 16 physical cores
  

• Apple M1 (Virtual) ARM64 Platform Information

  OS=macOS Sonoma 14.7.1 (23H222) [Darwin 23.6.0]
  Apple M1 (Virtual), 1 CPU, 3 logical and 3 physical cores
  

Reader Comparison Benchmarks

The following reader scenarios are benchmarked:

• NCsvPerf from The fastest CSV parser in .NET
• Floats as for example in machine learning.

Details for each can be found in the following. However, for each of these 3 different scopes are benchmarked to better ascertain the low-level performance of each library and approach and what parts of the parsing consume the most time:

• Row - for this scope only the row is enumerated. That is, for Sep all that is done is:

  foreach (var row in reader) { }
  

  this should capture parsing both row and columns but without accessing these. Note that some libraries (like Sylvan) will defer work for columns to when these are accessed.
• Cols - for this scope all rows and all columns are enumerated. If possible columns are accessed as spans, if not as strings, which then might mean a string has to be allocated. That is, for Sep this is:

  foreach (var row in reader)
  {
      for (var i = 0; i < row.ColCount; i++)
      {
          var span = row[i].Span;
      }
  }
  

• XYZ - finally the full scope is performed which is specific to each of the scenarios.

Additionally, as Sep supports multi-threaded parsing via ParallelEnumerate benchmarks results with _MT in the method name are multi-threaded. These show Sep provides unparalleled performance compared to any other CSV parser.

The overhead of Sep async support is also benchmarked and can be seen with _Async in the method name. Note that this is the absolute best case for async given there is no real IO involved and hence no actual asynchronous work or continuations (thus no Task allocations) since benchmarks run from memory only. This is fine as the main purpose of the benchmark is to gauge the overhead of the async code path.

NCsvPerf PackageAssets Reader Comparison Benchmarks

NCsvPerf from The fastest CSV parser in .NET is a benchmark which in Joel Verhagen's own words was defined with:

My goal was to find the fastest low-level CSV parser. Essentially, all I wanted was a library that gave me a string[] for each line where each field in the line was an element in the array.

What is great about this work is it tests a whole of 35 different libraries and approaches to this. Providing a great overview of those and their performance on this specific scenario. Given Sylvan is the fastest of those it is used as the one to beat here, while CsvHelper is used to compare to the most commonly used library.

The source used for this benchmark PackageAssetsBench.cs is a PackageAssets.csv with NuGet package information in 25 columns with rows like:

75fcf875-017d-4579-bfd9-791d3e6767f0,2020-11-28T01:50:41.2449947+00:00,Akinzekeel.BlazorGrid,0.9.1-preview,2020-11-27T22:42:54.3100000+00:00,AvailableAssets,RuntimeAssemblies,,,net5.0,,,,,,lib/net5.0/BlazorGrid.dll,BlazorGrid.dll,.dll,lib,net5.0,.NETCoreApp,5.0.0.0,,,0.0.0.0
75fcf875-017d-4579-bfd9-791d3e6767f0,2020-11-28T01:50:41.2449947+00:00,Akinzekeel.BlazorGrid,0.9.1-preview,2020-11-27T22:42:54.3100000+00:00,AvailableAssets,CompileLibAssemblies,,,net5.0,,,,,,lib/net5.0/BlazorGrid.dll,BlazorGrid.dll,.dll,lib,net5.0,.NETCoreApp,5.0.0.0,,,0.0.0.0
75fcf875-017d-4579-bfd9-791d3e6767f0,2020-11-28T01:50:41.2449947+00:00,Akinzekeel.BlazorGrid,0.9.1-preview,2020-11-27T22:42:54.3100000+00:00,AvailableAssets,ResourceAssemblies,,,net5.0,,,,,,lib/net5.0/de/BlazorGrid.resources.dll,BlazorGrid.resources.dll,.dll,lib,net5.0,.NETCoreApp,5.0.0.0,,,0.0.0.0
75fcf875-017d-4579-bfd9-791d3e6767f0,2020-11-28T01:50:41.2449947+00:00,Akinzekeel.BlazorGrid,0.9.1-preview,2020-11-27T22:42:54.3100000+00:00,AvailableAssets,MSBuildFiles,,,any,,,,,,build/Microsoft.AspNetCore.StaticWebAssets.props,Microsoft.AspNetCore.StaticWebAssets.props,.props,build,any,Any,0.0.0.0,,,0.0.0.0
75fcf875-017d-4579-bfd9-791d3e6767f0,2020-11-28T01:50:41.2449947+00:00,Akinzekeel.BlazorGrid,0.9.1-preview,2020-11-27T22:42:54.3100000+00:00,AvailableAssets,MSBuildFiles,,,any,,,,,,build/Akinzekeel.BlazorGrid.props,Akinzekeel.BlazorGrid.props,.props,build,any,Any,0.0.0.0,,,0.0.0.0

For Scope = Asset the columns are parsed into a PackageAsset class, which consists of 25 properties of which 22 are strings. Each asset is accumulated into a List<PackageAsset>. Each column is accessed as a string regardless.

This means this benchmark is dominated by turning columns into strings for the decently fast parsers. Hence, the fastest libraries in this test employ string pooling. That is, basically a custom dictionary from ReadOnlySpan<char> to string, which avoids allocating a new string for repeated values. And as can be seen in the csv-file there are a lot of repeated values. Both Sylvan and CsvHelper do this in the benchmark. So does Sep and as with Sep this is an optional configuration that has to be explicitly enable. For Sep this means the reader is created with something like:

using var reader = Sep.Reader(o => o with
{
    HasHeader = false,
    CreateToString = SepToString.PoolPerCol(maximumStringLength: 128),
})
.From(CreateReader());

What is unique for Sep is that it allows defining a pool per column e.g. via SepToString.PoolPerCol(...). This is based on the fact that often each column has its own set of values or strings that may be repeated without any overlap to other columns. This also allows one to define per column specific handling of ToString behavior. Whether to pool or not. Or even to use a statically defined pool.

Sep supports unescaping via an option, see SepReaderOptions and Unescaping. Therefore, Sep has two methods being tested. The default Sep without unescaping and Sep_Unescape where unescaping is enabled. Note that only if there are quotes will there be any unescaping, but to support unescape one has to track extra state during parsing which means there is a slight cost no matter what, most notably for the Cols scope. Sep is still the fastest of all (by far in many cases).

PackageAssets Benchmark Results

The results below show Sep is the fastest .NET CSV Parser (for this benchmark on these platforms and machines 😀). While for pure parsing allocating only a fraction of the memory due to extensive use of pooling and the ArrayPool<T>.

This is in many aspects due to Sep having extremely optimized string pooling and optimized hashing of ReadOnlySpan<char>, and thus not really due to the csv-parsing itself, since that is not a big part of the time consumed. At least not for a decently fast csv-parser.

With ParallelEnumerate (MT) Sep is >2x faster than Sylvan and up to 9x faster than CsvHelper.

At the lowest level of enumerating rows only, that is csv parsing only, Sep hits a staggering 21 GB/s on 9950X. Single-threaded.

AMD.EPYC.7763 - PackageAssets Benchmark Results (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	3.593 ms	1.00	29	8094.2	71.9	1.02 KB	1.00
Sep_Async	Row	50000	3.886 ms	1.08	29	7484.8	77.7	1.02 KB	1.00
Sep_Unescape	Row	50000	3.586 ms	1.00	29	8110.7	71.7	1.02 KB	1.00
Sylvan___	Row	50000	4.341 ms	1.21	29	6700.0	86.8	8.46 KB	8.33
ReadLine_	Row	50000	22.018 ms	6.13	29	1321.0	440.4	88608.23 KB	87,245.02
CsvHelper	Row	50000	63.984 ms	17.81	29	454.6	1279.7	19.95 KB	19.64
Sep______	Cols	50000	4.957 ms	1.00	29	5868.2	99.1	1.02 KB	1.00
Sep_Unescape	Cols	50000	5.868 ms	1.18	29	4956.5	117.4	1.02 KB	1.00
Sylvan___	Cols	50000	8.174 ms	1.65	29	3558.5	163.5	8.46 KB	8.32
ReadLine_	Cols	50000	23.316 ms	4.70	29	1247.5	466.3	88608.23 KB	87,161.22
CsvHelper	Cols	50000	103.321 ms	20.85	29	281.5	2066.4	445.67 KB	438.39
Sep______	Asset	50000	39.423 ms	1.00	29	737.8	788.5	13802.24 KB	1.00
Sep_MT___	Asset	50000	28.737 ms	0.73	29	1012.1	574.7	13861.74 KB	1.00
Sylvan___	Asset	50000	48.458 ms	1.23	29	600.2	969.2	13961.98 KB	1.01
ReadLine_	Asset	50000	123.884 ms	3.14	29	234.8	2477.7	102133.97 KB	7.40
CsvHelper	Asset	50000	123.793 ms	3.14	29	235.0	2475.9	13971.04 KB	1.01
Sep______	Asset	1000000	825.400 ms	1.00	581	705.0	825.4	266668.32 KB	1.00
Sep_MT___	Asset	1000000	506.193 ms	0.61	581	1149.5	506.2	275203.41 KB	1.03
Sylvan___	Asset	1000000	998.747 ms	1.21	581	582.6	998.7	266823.98 KB	1.00
ReadLine_	Asset	1000000	2,634.271 ms	3.19	581	220.9	2634.3	2038835.21 KB	7.65
CsvHelper	Asset	1000000	2,494.118 ms	3.02	581	233.3	2494.1	266839.37 KB	1.00

AMD.Ryzen.7.PRO.7840U.w.Radeon.780M - PackageAssets Benchmark Results (Sep 0.12.0.0, Sylvan 1.4.3.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	3.501 ms	1.00	29	8334.3	70.0	1.01 KB	1.00
Sep_Async	Row	50000	3.748 ms	1.07	29	7785.1	75.0	1.01 KB	1.00
Sep_Unescape	Row	50000	3.489 ms	1.00	29	8364.7	69.8	1.01 KB	1.00
Sylvan___	Row	50000	4.436 ms	1.27	29	6578.6	88.7	7.65 KB	7.59
ReadLine_	Row	50000	18.126 ms	5.18	29	1609.9	362.5	88608.23 KB	87,921.34
CsvHelper	Row	50000	65.367 ms	18.69	29	446.4	1307.3	19.95 KB	19.79
Sep______	Cols	50000	4.800 ms	1.00	29	6079.7	96.0	1.01 KB	1.00
Sep_Unescape	Cols	50000	5.952 ms	1.24	29	4902.5	119.0	1.01 KB	1.00
Sylvan___	Cols	50000	8.572 ms	1.79	29	3404.3	171.4	7.65 KB	7.59
ReadLine_	Cols	50000	19.182 ms	4.00	29	1521.3	383.6	88608.23 KB	87,921.34
CsvHelper	Cols	50000	108.414 ms	22.59	29	269.2	2168.3	445.6 KB	442.15
Sep______	Asset	50000	52.283 ms	1.00	29	558.1	1045.7	13802.65 KB	1.00
Sep_MT___	Asset	50000	34.441 ms	0.66	29	847.3	688.8	13913.24 KB	1.01
Sylvan___	Asset	50000	53.960 ms	1.04	29	540.8	1079.2	13962.29 KB	1.01
ReadLine_	Asset	50000	210.735 ms	4.04	29	138.5	4214.7	102133.96 KB	7.40
CsvHelper	Asset	50000	131.644 ms	2.53	29	221.7	2632.9	13970.83 KB	1.01
Sep______	Asset	1000000	978.280 ms	1.00	583	596.7	978.3	266668.88 KB	1.00
Sep_MT___	Asset	1000000	515.844 ms	0.53	583	1131.7	515.8	267776.55 KB	1.00
Sylvan___	Asset	1000000	1,130.594 ms	1.16	583	516.4	1130.6	266825.13 KB	1.00
ReadLine_	Asset	1000000	3,155.567 ms	3.23	583	185.0	3155.6	2038835.23 KB	7.65
CsvHelper	Asset	1000000	2,669.767 ms	2.73	583	218.7	2669.8	266844.99 KB	1.00

AMD.Ryzen.9.5950X - PackageAssets Benchmark Results (Sep 0.9.0.0, Sylvan 1.3.9.0, CsvHelper 33.0.1.24)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	2.230 ms	1.00	29	13088.4	44.6	1.09 KB	1.00
Sep_Async	Row	50000	2.379 ms	1.07	29	12264.0	47.6	1.02 KB	0.93
Sep_Unescape	Row	50000	2.305 ms	1.03	29	12657.6	46.1	1.02 KB	0.93
Sylvan___	Row	50000	2.993 ms	1.33	29	9750.2	59.9	7.65 KB	7.52
ReadLine_	Row	50000	12.106 ms	5.36	29	2410.5	242.1	88608.25 KB	87,077.59
CsvHelper	Row	50000	43.313 ms	19.19	29	673.7	866.3	20.04 KB	19.69
Sep______	Cols	50000	3.211 ms	1.00	29	9089.3	64.2	1.02 KB	1.00
Sep_Unescape	Cols	50000	3.845 ms	1.20	29	7589.1	76.9	1.02 KB	1.00
Sylvan___	Cols	50000	5.065 ms	1.58	29	5760.9	101.3	7.66 KB	7.52
ReadLine_	Cols	50000	12.850 ms	4.00	29	2270.9	257.0	88608.25 KB	86,910.78
CsvHelper	Cols	50000	68.999 ms	21.49	29	422.9	1380.0	445.85 KB	437.31
Sep______	Asset	50000	33.615 ms	1.00	29	868.1	672.3	13802.47 KB	1.00
Sep_MT___	Asset	50000	20.231 ms	0.60	29	1442.4	404.6	13992.1 KB	1.01
Sylvan___	Asset	50000	34.762 ms	1.03	29	839.5	695.2	13962.2 KB	1.01
ReadLine_	Asset	50000	97.204 ms	2.89	29	300.2	1944.1	102133.9 KB	7.40
CsvHelper	Asset	50000	83.550 ms	2.49	29	349.3	1671.0	13970.66 KB	1.01
Sep______	Asset	1000000	629.552 ms	1.00	583	927.3	629.6	266669.13 KB	1.00
Sep_MT___	Asset	1000000	261.089 ms	0.41	583	2236.0	261.1	267793.45 KB	1.00
Sylvan___	Asset	1000000	761.171 ms	1.21	583	767.0	761.2	266825.09 KB	1.00
ReadLine_	Asset	1000000	1,636.526 ms	2.60	583	356.7	1636.5	2038835.59 KB	7.65
CsvHelper	Asset	1000000	1,754.461 ms	2.79	583	332.7	1754.5	266833.16 KB	1.00

AMD.Ryzen.9.9950X - PackageAssets Benchmark Results (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	1.371 ms	1.00	29	21288.2	27.4	1.02 KB	1.00
Sep_Async	Row	50000	1.511 ms	1.10	29	19318.0	30.2	1.02 KB	1.00
Sep_Unescape	Row	50000	1.399 ms	1.02	29	20858.2	28.0	1.02 KB	1.00
Sylvan___	Row	50000	1.911 ms	1.39	29	15271.3	38.2	8.45 KB	8.32
ReadLine_	Row	50000	7.884 ms	5.75	29	3701.4	157.7	88608.23 KB	87,245.02
CsvHelper	Row	50000	23.786 ms	17.35	29	1226.8	475.7	19.95 KB	19.64
Sep______	Cols	50000	2.232 ms	1.00	29	13076.0	44.6	1.02 KB	1.00
Sep_Unescape	Cols	50000	2.369 ms	1.06	29	12315.8	47.4	1.02 KB	1.00
Sylvan___	Cols	50000	3.283 ms	1.47	29	8889.4	65.7	8.46 KB	8.32
ReadLine_	Cols	50000	8.279 ms	3.71	29	3524.9	165.6	88608.23 KB	87,245.02
CsvHelper	Cols	50000	44.021 ms	19.73	29	662.9	880.4	445.61 KB	438.75
Sep______	Asset	50000	22.918 ms	1.00	29	1273.3	458.4	13802.3 KB	1.00
Sep_MT___	Asset	50000	15.072 ms	0.66	29	1936.2	301.4	13997.81 KB	1.01
Sylvan___	Asset	50000	25.305 ms	1.10	29	1153.2	506.1	13962.14 KB	1.01
ReadLine_	Asset	50000	80.250 ms	3.50	29	363.6	1605.0	102133.88 KB	7.40
CsvHelper	Asset	50000	54.268 ms	2.37	29	537.7	1085.4	13970.19 KB	1.01
Sep______	Asset	1000000	441.901 ms	1.00	583	1321.1	441.9	266667.44 KB	1.00
Sep_MT___	Asset	1000000	191.048 ms	0.43	583	3055.7	191.0	267864.88 KB	1.00
Sylvan___	Asset	1000000	529.177 ms	1.20	583	1103.2	529.2	266824.41 KB	1.00
ReadLine_	Asset	1000000	1,312.114 ms	2.97	583	444.9	1312.1	2038834.94 KB	7.65
CsvHelper	Asset	1000000	1,160.448 ms	2.63	583	503.1	1160.4	266840.68 KB	1.00

AMD.Ryzen.9.9950X3D2 - PackageAssets Benchmark Results (Sep 0.17.1.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	1.300 ms	1.00	29	22451.9	26.0	1.02 KB	1.00
Sep_Async	Row	50000	1.450 ms	1.12	29	20131.5	29.0	1.02 KB	1.00
Sep_Unescape	Row	50000	1.313 ms	1.01	29	22220.0	26.3	1.02 KB	1.00
Sylvan___	Row	50000	1.858 ms	1.43	29	15708.8	37.2	8.45 KB	8.32
ReadLine_	Row	50000	7.627 ms	5.87	29	3826.1	152.5	88608.23 KB	87,245.02
CsvHelper	Row	50000	24.656 ms	18.97	29	1183.5	493.1	19.95 KB	19.64
Sep______	Cols	50000	1.851 ms	1.00	29	15766.8	37.0	1.02 KB	1.00
Sep_Unescape	Cols	50000	2.252 ms	1.22	29	12958.3	45.0	1.02 KB	1.00
Sylvan___	Cols	50000	3.199 ms	1.73	29	9122.8	64.0	8.46 KB	8.32
ReadLine_	Cols	50000	8.249 ms	4.46	29	3537.6	165.0	88608.23 KB	87,245.02
CsvHelper	Cols	50000	44.812 ms	24.21	29	651.2	896.2	445.6 KB	438.75
Sep______	Asset	50000	20.932 ms	1.00	29	1394.1	418.6	13801.89 KB	1.00
Sep_MT___	Asset	50000	9.588 ms	0.46	29	3043.5	191.8	13976.54 KB	1.01
Sylvan___	Asset	50000	25.103 ms	1.20	29	1162.5	502.1	13961.71 KB	1.01
ReadLine_	Asset	50000	57.269 ms	2.74	29	509.5	1145.4	102133.38 KB	7.40
CsvHelper	Asset	50000	53.148 ms	2.55	29	549.1	1063.0	13969.95 KB	1.01
Sep______	Asset	1000000	406.995 ms	1.00	583	1434.4	407.0	266664.02 KB	1.00
Sep_MT___	Asset	1000000	183.364 ms	0.45	583	3183.7	183.4	268786.5 KB	1.01
Sylvan___	Asset	1000000	476.032 ms	1.17	583	1226.4	476.0	266822.84 KB	1.00
ReadLine_	Asset	1000000	1,408.037 ms	3.47	583	414.6	1408.0	2038835.55 KB	7.65
CsvHelper	Asset	1000000	1,146.483 ms	2.82	583	509.2	1146.5	266831.14 KB	1.00

Apple.M1.(Virtual) - PackageAssets Benchmark Results (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	2.971 ms	1.00	29	9790.9	59.4	960 B	1.00
Sep_Async	Row	50000	3.025 ms	1.02	29	9614.8	60.5	960 B	1.00
Sep_Unescape	Row	50000	2.981 ms	1.00	29	9755.5	59.6	960 B	1.00
Sylvan___	Row	50000	18.006 ms	6.06	29	1615.3	360.1	7475 B	7.79
ReadLine_	Row	50000	16.749 ms	5.64	29	1736.6	335.0	90734824 B	94,515.44
CsvHelper	Row	50000	43.180 ms	14.54	29	673.6	863.6	20496 B	21.35
Sep______	Cols	50000	4.085 ms	1.00	29	7120.0	81.7	960 B	1.00
Sep_Unescape	Cols	50000	4.628 ms	1.13	29	6284.5	92.6	960 B	1.00
Sylvan___	Cols	50000	21.259 ms	5.21	29	1368.2	425.2	7516 B	7.83
ReadLine_	Cols	50000	16.801 ms	4.11	29	1731.2	336.0	90734824 B	94,515.44
CsvHelper	Cols	50000	65.949 ms	16.15	29	441.0	1319.0	456368 B	475.38
Sep______	Asset	50000	30.072 ms	1.00	29	967.2	601.4	14133394 B	1.00
Sep_MT___	Asset	50000	21.703 ms	0.72	29	1340.2	434.1	14271705 B	1.01
Sylvan___	Asset	50000	49.570 ms	1.65	29	586.8	991.4	14295922 B	1.01
ReadLine_	Asset	50000	79.663 ms	2.66	29	365.1	1593.3	104585174 B	7.40
CsvHelper	Asset	50000	73.557 ms	2.45	29	395.4	1471.1	14305304 B	1.01
Sep______	Asset	1000000	590.295 ms	1.00	581	985.7	590.3	273067136 B	1.00
Sep_MT___	Asset	1000000	429.687 ms	0.73	581	1354.2	429.7	284090744 B	1.04
Sylvan___	Asset	1000000	997.005 ms	1.69	581	583.6	997.0	273230816 B	1.00
ReadLine_	Asset	1000000	1,645.187 ms	2.79	581	353.7	1645.2	2087766816 B	7.65
CsvHelper	Asset	1000000	1,534.770 ms	2.60	581	379.1	1534.8	273236224 B	1.00

Cobalt.100 - PackageAssets Benchmark Results (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	5.195 ms	1.00	29	5616.8	103.9	960 B	1.00
Sep_Async	Row	50000	5.715 ms	1.10	29	5106.0	114.3	960 B	1.00
Sep_Unescape	Row	50000	5.432 ms	1.05	29	5372.4	108.6	960 B	1.00
Sylvan___	Row	50000	20.667 ms	3.99	29	1412.0	413.3	7476 B	7.79
ReadLine_	Row	50000	20.739 ms	4.00	29	1407.1	414.8	90734824 B	94,515.44
CsvHelper	Row	50000	54.592 ms	10.53	29	534.5	1091.8	20424 B	21.27
Sep______	Cols	50000	6.963 ms	1.00	29	4190.9	139.3	960 B	1.00
Sep_Unescape	Cols	50000	7.811 ms	1.12	29	3735.7	156.2	960 B	1.00
Sylvan___	Cols	50000	24.814 ms	3.56	29	1176.0	496.3	7480 B	7.79
ReadLine_	Cols	50000	20.856 ms	3.00	29	1399.2	417.1	90734827 B	94,515.44
CsvHelper	Cols	50000	88.627 ms	12.73	29	329.3	1772.5	456368 B	475.38
Sep______	Asset	50000	31.201 ms	1.00	29	935.3	624.0	14133000 B	1.00
Sep_MT___	Asset	50000	14.267 ms	0.46	29	2045.3	285.3	14193488 B	1.00
Sylvan___	Asset	50000	56.442 ms	1.81	29	517.0	1128.8	14295523 B	1.01
ReadLine_	Asset	50000	117.311 ms	3.76	29	248.8	2346.2	104584230 B	7.40
CsvHelper	Asset	50000	97.873 ms	3.14	29	298.2	1957.5	14308394 B	1.01
Sep______	Asset	1000000	775.808 ms	1.00	583	752.5	775.8	273063600 B	1.00
Sep_MT___	Asset	1000000	367.674 ms	0.47	583	1587.8	367.7	281452296 B	1.03
Sylvan___	Asset	1000000	1,237.548 ms	1.60	583	471.7	1237.5	273225288 B	1.00
ReadLine_	Asset	1000000	2,741.695 ms	3.53	583	212.9	2741.7	2087765224 B	7.65
CsvHelper	Asset	1000000	2,130.007 ms	2.75	583	274.1	2130.0	273247280 B	1.00

PackageAssets Benchmark Results (SERVER GC)

The package assets benchmark (Scope Asset) has a very high base load in the form of the accumulated instances of PackageAsset and since Sep is so fast the GC becomes a significant bottleneck for the benchmark, especially for multi-threaded parsing. Switching to SERVER GC can, therefore, provide significant speedup as can be seen below.

With ParallelEnumerate and server GC Sep is >4x faster than Sylvan and up to 18x faster than CsvHelper. Breaking 8 GB/s parsing speed on package assets on 9950X.

AMD.EPYC.7763 - PackageAssets Benchmark Results (SERVER GC) (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	30.67 ms	1.00	29	948.5	613.3	13.48 MB	1.00
Sep_MT___	Asset	50000	17.07 ms	0.56	29	1704.0	341.4	13.53 MB	1.00
Sylvan___	Asset	50000	41.63 ms	1.36	29	698.7	832.5	13.63 MB	1.01
ReadLine_	Asset	50000	58.75 ms	1.92	29	495.0	1175.1	99.74 MB	7.40
CsvHelper	Asset	50000	114.31 ms	3.73	29	254.5	2286.1	13.64 MB	1.01
Sep______	Asset	1000000	640.63 ms	1.00	581	908.3	640.6	260.41 MB	1.00
Sep_MT___	Asset	1000000	348.07 ms	0.54	581	1671.7	348.1	269.03 MB	1.03
Sylvan___	Asset	1000000	811.52 ms	1.27	581	717.0	811.5	260.57 MB	1.00
ReadLine_	Asset	1000000	1,205.96 ms	1.88	581	482.5	1206.0	1991.04 MB	7.65
CsvHelper	Asset	1000000	2,315.88 ms	3.62	581	251.3	2315.9	260.58 MB	1.00

AMD.Ryzen.7.PRO.7840U.w.Radeon.780M - PackageAssets Benchmark Results (SERVER GC) (Sep 0.12.0.0, Sylvan 1.4.3.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	35.630 ms	1.00	29	819.0	712.6	13.48 MB	1.00
Sep_MT___	Asset	50000	8.892 ms	0.25	29	3281.6	177.8	13.57 MB	1.01
Sylvan___	Asset	50000	38.901 ms	1.09	29	750.1	778.0	13.63 MB	1.01
ReadLine_	Asset	50000	43.945 ms	1.23	29	664.0	878.9	99.74 MB	7.40
CsvHelper	Asset	50000	116.596 ms	3.27	29	250.3	2331.9	13.64 MB	1.01
Sep______	Asset	1000000	654.638 ms	1.00	583	891.8	654.6	260.41 MB	1.00
Sep_MT___	Asset	1000000	244.949 ms	0.37	583	2383.3	244.9	262.32 MB	1.01
Sylvan___	Asset	1000000	825.727 ms	1.26	583	707.0	825.7	260.57 MB	1.00
ReadLine_	Asset	1000000	968.707 ms	1.48	583	602.6	968.7	1991.04 MB	7.65
CsvHelper	Asset	1000000	2,364.922 ms	3.62	583	246.9	2364.9	260.58 MB	1.00

AMD.Ryzen.9.5950X - PackageAssets Benchmark Results (SERVER GC) (Sep 0.9.0.0, Sylvan 1.3.9.0, CsvHelper 33.0.1.24)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	20.951 ms	1.00	29	1392.9	419.0	13.48 MB	1.00
Sep_MT___	Asset	50000	6.614 ms	0.32	29	4411.8	132.3	13.64 MB	1.01
Sylvan___	Asset	50000	27.761 ms	1.33	29	1051.2	555.2	13.63 MB	1.01
ReadLine_	Asset	50000	33.516 ms	1.60	29	870.7	670.3	99.74 MB	7.40
CsvHelper	Asset	50000	77.007 ms	3.68	29	378.9	1540.1	13.64 MB	1.01
Sep______	Asset	1000000	432.887 ms	1.00	583	1348.6	432.9	260.41 MB	1.00
Sep_MT___	Asset	1000000	119.430 ms	0.28	583	4888.1	119.4	261.39 MB	1.00
Sylvan___	Asset	1000000	559.550 ms	1.29	583	1043.3	559.6	260.57 MB	1.00
ReadLine_	Asset	1000000	573.637 ms	1.33	583	1017.7	573.6	1991.05 MB	7.65
CsvHelper	Asset	1000000	1,537.602 ms	3.55	583	379.7	1537.6	260.58 MB	1.00

AMD.Ryzen.9.9950X - PackageAssets Benchmark Results (SERVER GC) (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	14.258 ms	1.00	29	2046.6	285.2	13.48 MB	1.00
Sep_MT___	Asset	50000	3.261 ms	0.23	29	8947.4	65.2	13.65 MB	1.01
Sylvan___	Asset	50000	18.170 ms	1.27	29	1606.0	363.4	13.63 MB	1.01
ReadLine_	Asset	50000	19.166 ms	1.34	29	1522.6	383.3	99.74 MB	7.40
CsvHelper	Asset	50000	49.240 ms	3.45	29	592.6	984.8	13.64 MB	1.01
Sep______	Asset	1000000	287.939 ms	1.00	583	2027.5	287.9	260.41 MB	1.00
Sep_MT___	Asset	1000000	58.468 ms	0.20	583	9984.7	58.5	261.67 MB	1.00
Sylvan___	Asset	1000000	370.560 ms	1.29	583	1575.4	370.6	260.57 MB	1.00
ReadLine_	Asset	1000000	365.042 ms	1.27	583	1599.2	365.0	1991.04 MB	7.65
CsvHelper	Asset	1000000	1,001.476 ms	3.48	583	582.9	1001.5	260.58 MB	1.00

AMD.Ryzen.9.9950X3D2 - PackageAssets Benchmark Results (SERVER GC) (Sep 0.17.1.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	14.645 ms	1.00	29	1992.6	292.9	13.48 MB	1.00
Sep_MT___	Asset	50000	3.216 ms	0.22	29	9074.6	64.3	13.65 MB	1.01
Sylvan___	Asset	50000	18.847 ms	1.29	29	1548.3	376.9	13.63 MB	1.01
ReadLine_	Asset	50000	18.680 ms	1.28	29	1562.2	373.6	99.74 MB	7.40
CsvHelper	Asset	50000	50.322 ms	3.44	29	579.9	1006.4	13.64 MB	1.01
Sep______	Asset	1000000	294.100 ms	1.00	583	1985.0	294.1	260.41 MB	1.00
Sep_MT___	Asset	1000000	52.370 ms	0.18	583	11147.3	52.4	261.64 MB	1.00
Sylvan___	Asset	1000000	379.257 ms	1.29	583	1539.3	379.3	260.57 MB	1.00
ReadLine_	Asset	1000000	463.493 ms	1.58	583	1259.5	463.5	1991.04 MB	7.65
CsvHelper	Asset	1000000	1,026.722 ms	3.49	583	568.6	1026.7	260.58 MB	1.00

Apple.M1.(Virtual) - PackageAssets Benchmark Results (SERVER GC) (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	21.545 ms	1.00	29	1350.0	430.9	13.48 MB	1.00
Sep_MT___	Asset	50000	9.889 ms	0.46	29	2941.4	197.8	13.59 MB	1.01
Sylvan___	Asset	50000	43.137 ms	2.00	29	674.3	862.7	13.63 MB	1.01
ReadLine_	Asset	50000	28.729 ms	1.33	29	1012.4	574.6	99.74 MB	7.40
CsvHelper	Asset	50000	75.601 ms	3.51	29	384.7	1512.0	13.64 MB	1.01
Sep______	Asset	1000000	457.098 ms	1.00	581	1273.0	457.1	260.41 MB	1.00
Sep_MT___	Asset	1000000	224.967 ms	0.49	581	2586.5	225.0	268.14 MB	1.03
Sylvan___	Asset	1000000	894.656 ms	1.96	581	650.4	894.7	260.57 MB	1.00
ReadLine_	Asset	1000000	792.175 ms	1.73	581	734.5	792.2	1991.04 MB	7.65
CsvHelper	Asset	1000000	1,462.667 ms	3.20	581	397.8	1462.7	260.58 MB	1.00

Cobalt.100 - PackageAssets Benchmark Results (SERVER GC) (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	32.30 ms	1.00	29	903.3	646.1	13.48 MB	1.00
Sep_MT___	Asset	50000	11.63 ms	0.36	29	2509.7	232.5	13.54 MB	1.00
Sylvan___	Asset	50000	54.58 ms	1.69	29	534.6	1091.6	13.63 MB	1.01
ReadLine_	Asset	50000	48.26 ms	1.49	29	604.6	965.3	99.74 MB	7.40
CsvHelper	Asset	50000	98.61 ms	3.05	29	295.9	1972.2	13.64 MB	1.01
Sep______	Asset	1000000	645.55 ms	1.00	583	904.3	645.6	260.41 MB	1.00
Sep_MT___	Asset	1000000	199.49 ms	0.31	583	2926.4	199.5	266.18 MB	1.02
Sylvan___	Asset	1000000	1,090.59 ms	1.69	583	535.3	1090.6	260.57 MB	1.00
ReadLine_	Asset	1000000	1,001.06 ms	1.55	583	583.2	1001.1	1991.04 MB	7.65
CsvHelper	Asset	1000000	1,986.34 ms	3.08	583	293.9	1986.3	260.58 MB	1.00

PackageAssets with Quotes Benchmark Results

NCsvPerf does not examine performance in the face of quotes in the csv. This is relevant since some libraries like Sylvan will revert to a slower (not SIMD vectorized) parsing code path if it encounters quotes. Sep was designed to always use SIMD vectorization no matter what.

Since there are two extra chars to handle per column, it does have a significant impact on performance, no matter what though. This is expected when looking at the numbers. For each row of 25 columns, there are 24 separators (here ,) and one set of line endings (here \r\n). That's 26 characters. Adding quotes around each of the 25 columns will add 50 characters or almost triple the total to 76.

AMD.EPYC.7763 - PackageAssets with Quotes Benchmark Results (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	11.42 ms	1.00	33	2913.9	228.4	1.02 KB	1.00
Sep_Async	Row	50000	11.72 ms	1.03	33	2840.1	234.4	1.02 KB	1.00
Sep_Unescape	Row	50000	10.82 ms	0.95	33	3076.9	216.3	1.02 KB	1.00
Sylvan___	Row	50000	24.56 ms	2.15	33	1355.4	491.1	8.47 KB	8.34
ReadLine_	Row	50000	26.64 ms	2.33	33	1249.5	532.7	108778.73 KB	107,105.21
CsvHelper	Row	50000	77.46 ms	6.78	33	429.7	1549.2	19.95 KB	19.64
Sep______	Cols	50000	12.66 ms	1.00	33	2629.7	253.1	1.02 KB	1.00
Sep_Unescape	Cols	50000	13.32 ms	1.05	33	2498.0	266.5	1.02 KB	1.00
Sylvan___	Cols	50000	28.17 ms	2.23	33	1181.4	563.4	8.47 KB	8.34
ReadLine_	Cols	50000	27.55 ms	2.18	33	1208.2	550.9	108778.73 KB	107,105.21
CsvHelper	Cols	50000	106.05 ms	8.38	33	313.8	2121.1	445.68 KB	438.82
Sep______	Asset	50000	45.66 ms	1.00	33	728.9	913.2	13802.21 KB	1.00
Sep_MT___	Asset	50000	31.85 ms	0.70	33	1044.8	637.1	13865.49 KB	1.00
Sylvan___	Asset	50000	68.62 ms	1.50	33	485.0	1372.4	13961.94 KB	1.01
ReadLine_	Asset	50000	153.43 ms	3.36	33	216.9	3068.5	122305.14 KB	8.86
CsvHelper	Asset	50000	127.77 ms	2.80	33	260.5	2555.4	13971.34 KB	1.01
Sep______	Asset	1000000	968.61 ms	1.00	665	687.4	968.6	266679.21 KB	1.00
Sep_MT___	Asset	1000000	566.21 ms	0.58	665	1175.9	566.2	271900.91 KB	1.02
Sylvan___	Asset	1000000	1,412.90 ms	1.46	665	471.2	1412.9	266824.41 KB	1.00
ReadLine_	Asset	1000000	3,224.56 ms	3.33	665	206.5	3224.6	2442318.43 KB	9.16
CsvHelper	Asset	1000000	2,561.55 ms	2.64	665	259.9	2561.6	266835.7 KB	1.00

AMD.Ryzen.7.PRO.7840U.w.Radeon.780M - PackageAssets with Quotes Benchmark Results (Sep 0.12.0.0, Sylvan 1.4.3.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	11.228 ms	1.00	33	2972.8	224.6	1.01 KB	1.00
Sep_Async	Row	50000	9.916 ms	0.88	33	3365.9	198.3	1.01 KB	1.00
Sep_Unescape	Row	50000	9.953 ms	0.89	33	3353.4	199.1	1.01 KB	1.00
Sylvan___	Row	50000	23.829 ms	2.12	33	1400.7	476.6	7.66 KB	7.60
ReadLine_	Row	50000	21.909 ms	1.95	33	1523.4	438.2	108778.73 KB	107,935.48
CsvHelper	Row	50000	71.571 ms	6.38	33	466.4	1431.4	19.95 KB	19.79
Sep______	Cols	50000	12.798 ms	1.00	33	2608.0	256.0	1.01 KB	1.00
Sep_Unescape	Cols	50000	13.689 ms	1.07	33	2438.2	273.8	1.01 KB	1.00
Sylvan___	Cols	50000	27.287 ms	2.13	33	1223.2	545.7	7.67 KB	7.61
ReadLine_	Cols	50000	23.201 ms	1.81	33	1438.6	464.0	108778.73 KB	107,935.48
CsvHelper	Cols	50000	104.298 ms	8.15	33	320.0	2086.0	445.6 KB	442.15
Sep______	Asset	50000	55.957 ms	1.00	33	596.5	1119.1	13802.68 KB	1.00
Sep_MT___	Asset	50000	44.668 ms	0.80	33	747.2	893.4	13934.89 KB	1.01
Sylvan___	Asset	50000	68.353 ms	1.22	33	488.3	1367.1	13961.88 KB	1.01
ReadLine_	Asset	50000	260.030 ms	4.65	33	128.4	5200.6	122304.52 KB	8.86
CsvHelper	Asset	50000	127.992 ms	2.29	33	260.8	2559.8	13971.32 KB	1.01
Sep______	Asset	1000000	1,145.307 ms	1.00	667	583.0	1145.3	266673.49 KB	1.00
Sep_MT___	Asset	1000000	696.277 ms	0.61	667	959.0	696.3	267979.84 KB	1.00
Sylvan___	Asset	1000000	1,499.974 ms	1.31	667	445.1	1500.0	266827.47 KB	1.00
ReadLine_	Asset	1000000	4,060.701 ms	3.55	667	164.4	4060.7	2442318 KB	9.16
CsvHelper	Asset	1000000	2,601.519 ms	2.27	667	256.7	2601.5	266839.95 KB	1.00

AMD.Ryzen.9.5950X - PackageAssets with Quotes Benchmark Results (Sep 0.9.0.0, Sylvan 1.3.9.0, CsvHelper 33.0.1.24)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	7.046 ms	1.00	33	4737.2	140.9	1.04 KB	1.00
Sep_Async	Row	50000	8.137 ms	1.15	33	4101.8	162.7	1.04 KB	1.00
Sep_Unescape	Row	50000	7.473 ms	1.06	33	4466.7	149.5	1.04 KB	1.00
Sylvan___	Row	50000	17.571 ms	2.38	33	1899.5	351.4	7.69 KB	7.41
ReadLine_	Row	50000	14.336 ms	1.94	33	2328.2	286.7	108778.75 KB	104,689.33
CsvHelper	Row	50000	52.672 ms	7.12	33	633.7	1053.4	20.05 KB	19.29
Sep______	Cols	50000	8.126 ms	1.00	33	4107.5	162.5	1.04 KB	1.00
Sep_Unescape	Cols	50000	9.748 ms	1.20	33	3424.0	195.0	1.05 KB	1.01
Sylvan___	Cols	50000	20.503 ms	2.52	33	1628.0	410.1	7.7 KB	7.39
ReadLine_	Cols	50000	16.513 ms	2.03	33	2021.3	330.3	108778.76 KB	104,394.99
CsvHelper	Cols	50000	74.224 ms	9.13	33	449.7	1484.5	445.85 KB	427.88
Sep______	Asset	50000	39.523 ms	1.00	33	844.5	790.5	13802.63 KB	1.00
Sep_MT___	Asset	50000	23.386 ms	0.59	33	1427.2	467.7	13981.76 KB	1.01
Sylvan___	Asset	50000	50.803 ms	1.29	33	657.0	1016.1	13962.08 KB	1.01
ReadLine_	Asset	50000	114.306 ms	2.89	33	292.0	2286.1	122304.45 KB	8.86
CsvHelper	Asset	50000	88.786 ms	2.25	33	375.9	1775.7	13970.43 KB	1.01
Sep______	Asset	1000000	752.681 ms	1.00	667	887.1	752.7	266669 KB	1.00
Sep_MT___	Asset	1000000	377.733 ms	0.50	667	1767.7	377.7	267992.5 KB	1.00
Sylvan___	Asset	1000000	1,091.345 ms	1.45	667	611.8	1091.3	266825.09 KB	1.00
ReadLine_	Asset	1000000	2,615.390 ms	3.47	667	255.3	2615.4	2442319.06 KB	9.16
CsvHelper	Asset	1000000	1,756.409 ms	2.33	667	380.2	1756.4	266839.53 KB	1.00

AMD.Ryzen.9.9950X - PackageAssets with Quotes Benchmark Results (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	4.186 ms	1.00	33	7973.2	83.7	1.02 KB	1.00
Sep_Async	Row	50000	4.303 ms	1.03	33	7756.4	86.1	1.02 KB	1.00
Sep_Unescape	Row	50000	4.170 ms	1.00	33	8004.6	83.4	1.02 KB	1.00
Sylvan___	Row	50000	10.639 ms	2.54	33	3137.2	212.8	8.46 KB	8.33
ReadLine_	Row	50000	9.464 ms	2.26	33	3526.7	189.3	108778.73 KB	107,105.21
CsvHelper	Row	50000	27.532 ms	6.58	33	1212.3	550.6	19.95 KB	19.64
Sep______	Cols	50000	5.125 ms	1.00	33	6512.0	102.5	1.02 KB	1.00
Sep_Unescape	Cols	50000	5.701 ms	1.11	33	5854.7	114.0	1.02 KB	1.00
Sylvan___	Cols	50000	11.930 ms	2.33	33	2797.9	238.6	8.46 KB	8.33
ReadLine_	Cols	50000	10.103 ms	1.97	33	3303.7	202.1	108778.73 KB	107,105.21
CsvHelper	Cols	50000	40.502 ms	7.90	33	824.1	810.0	445.61 KB	438.75
Sep______	Asset	50000	28.119 ms	1.00	33	1187.0	562.4	13802.38 KB	1.00
Sep_MT___	Asset	50000	16.974 ms	0.60	33	1966.4	339.5	13995.39 KB	1.01
Sylvan___	Asset	50000	33.211 ms	1.18	33	1005.0	664.2	13961.94 KB	1.01
ReadLine_	Asset	50000	94.445 ms	3.36	33	353.4	1888.9	122304.82 KB	8.86
CsvHelper	Asset	50000	51.841 ms	1.84	33	643.8	1036.8	13970.67 KB	1.01
Sep______	Asset	1000000	520.110 ms	1.00	667	1283.8	520.1	266667.34 KB	1.00
Sep_MT___	Asset	1000000	228.902 ms	0.44	667	2917.0	228.9	269331.98 KB	1.01
Sylvan___	Asset	1000000	715.715 ms	1.38	667	932.9	715.7	266826.29 KB	1.00
ReadLine_	Asset	1000000	2,134.339 ms	4.10	667	312.8	2134.3	2442318.89 KB	9.16
CsvHelper	Asset	1000000	1,114.437 ms	2.14	667	599.1	1114.4	266832.37 KB	1.00

AMD.Ryzen.9.9950X3D2 - PackageAssets with Quotes Benchmark Results (Sep 0.17.1.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	4.252 ms	1.00	33	7850.6	85.0	1.02 KB	1.00
Sep_Async	Row	50000	4.180 ms	0.98	33	7984.6	83.6	1.02 KB	1.00
Sep_Unescape	Row	50000	3.989 ms	0.94	33	8367.3	79.8	1.02 KB	1.00
Sylvan___	Row	50000	10.795 ms	2.54	33	3091.8	215.9	8.46 KB	8.33
ReadLine_	Row	50000	9.123 ms	2.15	33	3658.5	182.5	108778.73 KB	107,105.21
CsvHelper	Row	50000	28.260 ms	6.65	33	1181.1	565.2	19.95 KB	19.64
Sep______	Cols	50000	5.014 ms	1.00	33	6656.3	100.3	1.02 KB	1.00
Sep_Unescape	Cols	50000	5.367 ms	1.07	33	6219.5	107.3	1.02 KB	1.00
Sylvan___	Cols	50000	12.688 ms	2.53	33	2630.6	253.8	8.46 KB	8.33
ReadLine_	Cols	50000	9.400 ms	1.87	33	3550.9	188.0	108778.73 KB	107,105.21
CsvHelper	Cols	50000	41.837 ms	8.34	33	797.8	836.7	445.6 KB	438.75
Sep______	Asset	50000	25.479 ms	1.00	33	1310.0	509.6	13801.9 KB	1.00
Sep_MT___	Asset	50000	12.111 ms	0.48	33	2755.9	242.2	13975.38 KB	1.01
Sylvan___	Asset	50000	31.375 ms	1.23	33	1063.8	627.5	13961.65 KB	1.01
ReadLine_	Asset	50000	68.645 ms	2.70	33	486.2	1372.9	122304.07 KB	8.86
CsvHelper	Asset	50000	50.970 ms	2.00	33	654.8	1019.4	13969.95 KB	1.01
Sep______	Asset	1000000	470.303 ms	1.00	667	1419.7	470.3	266663.86 KB	1.00
Sep_MT___	Asset	1000000	226.690 ms	0.48	667	2945.5	226.7	268001.88 KB	1.01
Sylvan___	Asset	1000000	667.054 ms	1.42	667	1001.0	667.1	266822.84 KB	1.00
ReadLine_	Asset	1000000	1,738.841 ms	3.70	667	384.0	1738.8	2442318.52 KB	9.16
CsvHelper	Asset	1000000	1,084.758 ms	2.31	667	615.5	1084.8	266831.14 KB	1.00

Apple.M1.(Virtual) - PackageAssets with Quotes Benchmark Results (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	7.063 ms	1.00	33	4712.5	141.3	960 B	1.00
Sep_Async	Row	50000	7.049 ms	1.00	33	4721.6	141.0	960 B	1.00
Sep_Unescape	Row	50000	7.173 ms	1.02	33	4640.0	143.5	960 B	1.00
Sylvan___	Row	50000	20.460 ms	2.90	33	1626.7	409.2	7516 B	7.83
ReadLine_	Row	50000	19.699 ms	2.79	33	1689.5	394.0	111389416 B	116,030.64
CsvHelper	Row	50000	46.922 ms	6.64	33	709.3	938.4	20496 B	21.35
Sep______	Cols	50000	8.268 ms	1.00	33	4025.4	165.4	960 B	1.00
Sep_Unescape	Cols	50000	9.084 ms	1.10	33	3663.8	181.7	961 B	1.00
Sylvan___	Cols	50000	23.584 ms	2.85	33	1411.2	471.7	7516 B	7.83
ReadLine_	Cols	50000	19.731 ms	2.39	33	1686.8	394.6	111389416 B	116,030.64
CsvHelper	Cols	50000	72.376 ms	8.75	33	459.8	1447.5	456296 B	475.31
Sep______	Asset	50000	35.076 ms	1.00	33	948.8	701.5	14133372 B	1.00
Sep_MT___	Asset	50000	21.356 ms	0.61	33	1558.4	427.1	14273964 B	1.01
Sylvan___	Asset	50000	51.378 ms	1.47	33	647.8	1027.6	14295882 B	1.01
ReadLine_	Asset	50000	117.825 ms	3.37	33	282.5	2356.5	125240038 B	8.86
CsvHelper	Asset	50000	79.105 ms	2.26	33	420.7	1582.1	14305304 B	1.01
Sep______	Asset	1000000	712.793 ms	1.00	665	934.1	712.8	273067136 B	1.00
Sep_MT___	Asset	1000000	480.678 ms	0.67	665	1385.1	480.7	283217688 B	1.04
Sylvan___	Asset	1000000	1,120.908 ms	1.57	665	594.0	1120.9	273226744 B	1.00
ReadLine_	Asset	1000000	2,250.766 ms	3.16	665	295.8	2250.8	2500933480 B	9.16
CsvHelper	Asset	1000000	1,737.967 ms	2.44	665	383.1	1738.0	273241080 B	1.00

Cobalt.100 - PackageAssets with Quotes Benchmark Results (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Row	50000	10.92 ms	1.00	33	3057.5	218.3	960 B	1.00
Sep_Async	Row	50000	11.08 ms	1.02	33	3012.0	221.6	962 B	1.00
Sep_Unescape	Row	50000	10.98 ms	1.01	33	3040.5	219.6	962 B	1.00
Sylvan___	Row	50000	23.03 ms	2.11	33	1449.2	460.6	7478 B	7.79
ReadLine_	Row	50000	24.21 ms	2.22	33	1378.4	484.3	111389416 B	116,030.64
CsvHelper	Row	50000	62.88 ms	5.76	33	530.8	1257.7	20424 B	21.27
Sep______	Cols	50000	12.47 ms	1.00	33	2677.2	249.3	962 B	1.00
Sep_Unescape	Cols	50000	13.73 ms	1.10	33	2430.4	274.7	962 B	1.00
Sylvan___	Cols	50000	27.41 ms	2.20	33	1217.5	548.3	7481 B	7.78
ReadLine_	Cols	50000	25.01 ms	2.01	33	1334.7	500.2	111389416 B	115,789.41
CsvHelper	Cols	50000	95.34 ms	7.65	33	350.1	1906.8	456368 B	474.40
Sep______	Asset	50000	36.81 ms	1.00	33	906.7	736.3	14133000 B	1.00
Sep_MT___	Asset	50000	14.89 ms	0.40	33	2241.8	297.8	14190284 B	1.00
Sylvan___	Asset	50000	58.64 ms	1.59	33	569.2	1172.9	14295523 B	1.01
ReadLine_	Asset	50000	126.40 ms	3.43	33	264.1	2528.0	125238788 B	8.86
CsvHelper	Asset	50000	106.21 ms	2.89	33	314.3	2124.2	14305310 B	1.01
Sep______	Asset	1000000	912.09 ms	1.00	667	732.1	912.1	273063648 B	1.00
Sep_MT___	Asset	1000000	389.94 ms	0.43	667	1712.3	389.9	282220464 B	1.03
Sylvan___	Asset	1000000	1,276.04 ms	1.40	667	523.3	1276.0	273225288 B	1.00
ReadLine_	Asset	1000000	3,123.47 ms	3.42	667	213.8	3123.5	2500931904 B	9.16
CsvHelper	Asset	1000000	2,292.40 ms	2.51	667	291.3	2292.4	273234848 B	1.00

PackageAssets with Quotes Benchmark Results (SERVER GC)

Here again are benchmark results with server garbage collection, which provides significant speedup over workstation garbage collection.

AMD.EPYC.7763 - PackageAssets with Quotes Benchmark Results (SERVER GC) (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	36.95 ms	1.00	33	900.8	738.9	13.48 MB	1.00
Sep_MT___	Asset	50000	22.04 ms	0.60	33	1510.2	440.8	13.53 MB	1.00
Sylvan___	Asset	50000	60.86 ms	1.65	33	546.8	1217.2	13.63 MB	1.01
ReadLine_	Asset	50000	64.87 ms	1.76	33	513.1	1297.4	119.44 MB	8.86
CsvHelper	Asset	50000	118.69 ms	3.21	33	280.4	2373.7	13.64 MB	1.01
Sep______	Asset	1000000	763.29 ms	1.00	665	872.3	763.3	260.41 MB	1.00
Sep_MT___	Asset	1000000	429.16 ms	0.56	665	1551.4	429.2	261.76 MB	1.01
Sylvan___	Asset	1000000	1,223.33 ms	1.60	665	544.3	1223.3	260.57 MB	1.00
ReadLine_	Asset	1000000	1,393.90 ms	1.83	665	477.7	1393.9	2385.07 MB	9.16
CsvHelper	Asset	1000000	2,365.35 ms	3.10	665	281.5	2365.4	260.58 MB	1.00

AMD.Ryzen.7.PRO.7840U.w.Radeon.780M - PackageAssets with Quotes Benchmark Results (SERVER GC) (Sep 0.12.0.0, Sylvan 1.4.3.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	39.09 ms	1.00	33	854.0	781.7	13.48 MB	1.00
Sep_MT___	Asset	50000	25.90 ms	0.66	33	1288.9	517.9	13.58 MB	1.01
Sylvan___	Asset	50000	59.92 ms	1.53	33	557.0	1198.4	13.63 MB	1.01
ReadLine_	Asset	50000	60.27 ms	1.54	33	553.8	1205.4	119.44 MB	8.86
CsvHelper	Asset	50000	111.54 ms	2.85	33	299.2	2230.8	13.64 MB	1.01
Sep______	Asset	1000000	810.33 ms	1.00	667	824.0	810.3	260.41 MB	1.00
Sep_MT___	Asset	1000000	433.80 ms	0.54	667	1539.2	433.8	261.43 MB	1.00
Sylvan___	Asset	1000000	1,216.78 ms	1.50	667	548.8	1216.8	260.57 MB	1.00
ReadLine_	Asset	1000000	1,079.99 ms	1.33	667	618.3	1080.0	2385.07 MB	9.16
CsvHelper	Asset	1000000	2,275.20 ms	2.81	667	293.5	2275.2	260.58 MB	1.00

AMD.Ryzen.9.5950X - PackageAssets with Quotes Benchmark Results (SERVER GC) (Sep 0.9.0.0, Sylvan 1.3.9.0, CsvHelper 33.0.1.24)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	26.42 ms	1.00	33	1263.1	528.5	13.48 MB	1.00
Sep_MT___	Asset	50000	11.53 ms	0.44	33	2894.1	230.7	13.64 MB	1.01
Sylvan___	Asset	50000	43.05 ms	1.63	33	775.3	861.1	13.63 MB	1.01
ReadLine_	Asset	50000	37.30 ms	1.41	33	894.8	746.0	119.44 MB	8.86
CsvHelper	Asset	50000	78.91 ms	2.99	33	423.0	1578.1	13.64 MB	1.01
Sep______	Asset	1000000	538.48 ms	1.00	667	1240.0	538.5	260.43 MB	1.00
Sep_MT___	Asset	1000000	213.29 ms	0.40	667	3130.5	213.3	261.37 MB	1.00
Sylvan___	Asset	1000000	879.04 ms	1.63	667	759.6	879.0	260.57 MB	1.00
ReadLine_	Asset	1000000	642.57 ms	1.19	667	1039.1	642.6	2385.07 MB	9.16
CsvHelper	Asset	1000000	1,598.79 ms	2.97	667	417.6	1598.8	260.58 MB	1.00

AMD.Ryzen.9.9950X - PackageAssets with Quotes Benchmark Results (SERVER GC) (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	17.929 ms	1.00	33	1861.7	358.6	13.48 MB	1.00
Sep_MT___	Asset	50000	5.872 ms	0.33	33	5684.4	117.4	13.64 MB	1.01
Sylvan___	Asset	50000	26.385 ms	1.47	33	1265.0	527.7	13.63 MB	1.01
ReadLine_	Asset	50000	21.798 ms	1.22	33	1531.2	436.0	119.44 MB	8.86
CsvHelper	Asset	50000	47.041 ms	2.62	33	709.5	940.8	13.64 MB	1.01
Sep______	Asset	1000000	358.149 ms	1.00	667	1864.3	358.1	260.41 MB	1.00
Sep_MT___	Asset	1000000	105.507 ms	0.29	667	6328.6	105.5	261.97 MB	1.01
Sylvan___	Asset	1000000	542.348 ms	1.51	667	1231.1	542.3	260.57 MB	1.00
ReadLine_	Asset	1000000	399.659 ms	1.12	667	1670.7	399.7	2385.07 MB	9.16
CsvHelper	Asset	1000000	942.707 ms	2.63	667	708.3	942.7	260.58 MB	1.00

AMD.Ryzen.9.9950X3D2 - PackageAssets with Quotes Benchmark Results (SERVER GC) (Sep 0.17.1.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	18.010 ms	1.00	33	1853.3	360.2	13.48 MB	1.00
Sep_MT___	Asset	50000	6.742 ms	0.37	33	4950.8	134.8	13.64 MB	1.01
Sylvan___	Asset	50000	28.010 ms	1.56	33	1191.6	560.2	13.63 MB	1.01
ReadLine_	Asset	50000	24.633 ms	1.37	33	1355.0	492.7	119.44 MB	8.86
CsvHelper	Asset	50000	48.301 ms	2.68	33	691.0	966.0	13.64 MB	1.01
Sep______	Asset	1000000	365.281 ms	1.00	667	1827.9	365.3	260.41 MB	1.00
Sep_MT___	Asset	1000000	107.768 ms	0.30	667	6195.8	107.8	261.79 MB	1.01
Sylvan___	Asset	1000000	571.951 ms	1.57	667	1167.4	572.0	260.57 MB	1.00
ReadLine_	Asset	1000000	512.824 ms	1.40	667	1302.0	512.8	2385.07 MB	9.16
CsvHelper	Asset	1000000	970.257 ms	2.66	667	688.2	970.3	260.58 MB	1.00

Apple.M1.(Virtual) - PackageAssets with Quotes Benchmark Results (SERVER GC) (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	25.99 ms	1.00	33	1280.6	519.8	13.48 MB	1.00
Sep_MT___	Asset	50000	11.90 ms	0.46	33	2797.0	238.0	13.6 MB	1.01
Sylvan___	Asset	50000	45.83 ms	1.77	33	726.2	916.6	13.63 MB	1.01
ReadLine_	Asset	50000	35.97 ms	1.39	33	925.3	719.4	119.44 MB	8.86
CsvHelper	Asset	50000	76.69 ms	2.95	33	434.0	1533.8	13.64 MB	1.01
Sep______	Asset	1000000	542.50 ms	1.00	665	1227.3	542.5	260.41 MB	1.00
Sep_MT___	Asset	1000000	239.81 ms	0.44	665	2776.4	239.8	266.31 MB	1.02
Sylvan___	Asset	1000000	961.86 ms	1.77	665	692.2	961.9	260.57 MB	1.00
ReadLine_	Asset	1000000	929.89 ms	1.71	665	716.0	929.9	2385.07 MB	9.16
CsvHelper	Asset	1000000	1,584.75 ms	2.92	665	420.1	1584.8	260.58 MB	1.00

Cobalt.100 - PackageAssets with Quotes Benchmark Results (SERVER GC) (Sep 0.13.0.0, Sylvan 1.4.4.0, CsvHelper 33.1.0.26)

Method	Scope	Rows	Mean	Ratio	MB	MB/s	ns/row	Allocated	Alloc Ratio
Sep______	Asset	50000	39.23 ms	1.00	33	850.7	784.7	13.48 MB	1.00
Sep_MT___	Asset	50000	14.62 ms	0.37	33	2283.8	292.3	13.53 MB	1.00
Sylvan___	Asset	50000	56.71 ms	1.45	33	588.5	1134.3	13.63 MB	1.01
ReadLine_	Asset	50000	55.29 ms	1.41	33	603.7	1105.8	119.44 MB	8.86
CsvHelper

---

README truncated. Read the full version on GitHub.