Introduction

🔥 FlameCsv 🔥 is a fully-featured high performance CSV library for .NET with a simple API, deep customization, and built-in support for UTF8, nativeAOT, and more.

FlameCsv can be thought of as a CSV-equivalent of System.Text.Json. It is designed to simple to use, extensible, and performant, while supporting low-level operations for advanced use-cases. FlameCsv is the most memory-efficient .NET CSV library available.

FlameCsv can process millions of CSV records per second on consumer hardware, and write arbitrarily large datasets with near-zero allocations. See the benchmarks for more details.

The library has thousands of unit tests and has been fuzz-tested.

See Getting Started, view the Examples, or deep dive into the API Reference.

Features

TL;DR: Blazingly fast, trimmable and easy-to-use feature-rich CSV library
Ease of Use
- Fluent API to read/write CSV from/to almost any source/destination
- Built-in support for common CLR types and interfaces like I(Utf8)SpanParsable
- Full feature parity with sync and async APIs
- Read or write UTF-8 bytes directly without a TextReader or TextWriter
- Hot reload support for internal caches
High Performance
- SIMD parsers tuned for each platform (AVX2, AVX512, ARM64)
- Near-zero allocations irrespective of dataset size
- Parallel APIs to read/write records in parallel
- Low-level APIs to handle raw CSV field spans directly
Deep Customization
- Attribute configuration for header names, constructors, field order, etc.
- Support for custom converters and converter factories (like System.Text.Json)
- Read or write multiple CSV documents from/to a single data stream
Source Generators
- Library is fully annotated for NativeAOT and trimming
- Source generated type maps for reflection-free reading and writing
- Source generated enum converters with up to 10x better performance than Enum.TryParse/TryFormat

All of this without sacrificing ease of use.

Example

UTF-16
UTF-8

using FlameCsv;

const string data =
    """
    id,name,age
    1,Bob,42
    2,Alice,37
    3,"Bond, James",39
    """;

List<User> users = [];

// read users from utf16 string
foreach (var user in Csv.From(data).Read<User>())
{
    users.Add(user);
}

// write users to a stream as tab-separated fields
await Csv.To(stream, Encoding.UTF8).WriteAsync(
    users,
    new CsvOptions<char> { Delimiter = '\t' },
    cancellationToken);

using FlameCsv;

byte[] data =
    """
    id,name,age
    1,Bob,42
    2,Alice,37
    3,"Bond, James",39
    """u8.ToArray();

List<User> users = [];

// read users from utf8 bytes
foreach (var user in Csv.From(data).Read<User>())
{
    users.Add(user);
}

// write users to a stream as tab-separated fields
await Csv.To(stream).WriteAsync(
    users,
    new CsvOptions<char> { Delimiter = '\t' },
    cancellationToken);

Dependencies

FlameCsv has two dependencies:

CommunityToolkit.HighPerformance which provides utilities for writing high-performance code, and the string pool used for header values.
A development-time dependency to FastExpressionCompiler is used for extremely performant runtime code-generation in reflection-based code paths.