fastCSV

The fastest CSV parser for Python. Written in C with AVX2 SIMD acceleration, memory-mapped I/O, and zero-copy columnar NumPy output.

Benchmarks

Hardware: AMD Ryzen 5 7235HS

Installation

pip install fastcsv-python

Usage

import fastcsv

# Returns dict of column_name -> numpy.ndarray
result = fastcsv.read_csv("data.csv")
result = fastcsv.read_csv("data.csv", delimiter=",", has_header=True, error_mode="strict")

# Streaming — constant memory regardless of file size
for chunk in fastcsv.reader("data.csv", chunk_size=10_000):
    process(chunk)

Options

How it's fast

• mmap I/O — the OS pages in only what is needed. No read() syscalls.
• AVX2 SIMD — scans 32 bytes per cycle to find delimiters. Falls back to SSE4.2 (16 bytes) or scalar automatically.
• Zero-copy output — int and float columns are handed to NumPy directly from the C buffer. No data is copied.
• Single-pass type inference — column types resolved in one pass, never re-scanned.
• GIL released — the entire parse phase runs without the GIL. Safe for multi-threaded use.

Error recovery

• strict — raises ValueError on any RFC 4180 violation.
• skip — silently drops malformed rows.
• replace — replaces malformed fields with empty string.

Building from source

pip install numpy
pip install -e .
make test-all

License

MIT 904c2c43-028b-4195-83f8-cd7f0dcb5d2b