cisv 0.1.3

High-performance CSV parser with SIMD optimizations (AVX-512/AVX2)

CISV Python Binding

High-performance CSV parser with SIMD optimizations for Python.

Requirements

• Python 3.7+
• CISV core library (libcisv.so)

Installation

Build Core Library First

cd ../../core
make

Install Python Package

pip install -e .

Or using the Makefile:

make build

Quick Start

from cisv import CisvParser, parse_file, parse_string, count_rows

# Simple file parsing
rows = parse_file('data.csv')
for row in rows:
    print(row)

# Parse with custom options
parser = CisvParser(
    delimiter=';',
    quote="'",
    trim=True
)
rows = parser.parse_file('data.csv')

# Parse from string
csv_data = """name,age,email
John,30,[e-mail]
Jane,25,[e-mail]"""

rows = parse_string(csv_data)

# Fast row counting (without full parsing)
total = count_rows('large.csv')
print(f"Total rows: {total}")

API Reference

CisvParser Class

class CisvParser:
    def __init__(
        self,
        delimiter: str = ',',
        quote: str = '"',
        escape: Optional[str] = None,
        comment: Optional[str] = None,
        trim: bool = False,
        skip_empty_lines: bool = False,
    ):
        """
        Create a new CSV parser.

        Args:
            delimiter: Field separator character (default: ',')
            quote: Quote character for fields (default: '"')
            escape: Escape character (default: None for RFC4180 "" style)
            comment: Comment line prefix (default: None)
            trim: Strip whitespace from fields (default: False)
            skip_empty_lines: Skip empty lines (default: False)
        """

    def parse_file(self, path: str) -> List[List[str]]:
        """Parse a CSV file and return all rows."""

    def parse_string(self, content: str) -> List[List[str]]:
        """Parse a CSV string and return all rows."""

Convenience Functions

def parse_file(
    path: str,
    delimiter: str = ',',
    quote: str = '"',
    **kwargs
) -> List[List[str]]:
    """Parse a CSV file with the given options."""

def parse_string(
    content: str,
    delimiter: str = ',',
    quote: str = '"',
    **kwargs
) -> List[List[str]]:
    """Parse a CSV string with the given options."""

def count_rows(path: str) -> int:
    """Count rows in a CSV file without full parsing."""

Configuration Options

Option	Type	Default	Description
delimiter	str	','	Field delimiter character
quote	str	'"'	Quote character
escape	str	None	Escape character
comment	str	None	Comment line prefix
trim	bool	False	Trim whitespace from fields
skip_empty_lines	bool	False	Skip empty lines

Examples

TSV Parsing

from cisv import CisvParser

parser = CisvParser(delimiter='\t')
rows = parser.parse_file('data.tsv')

Skip Comments and Empty Lines

parser = CisvParser(
    comment='#',
    skip_empty_lines=True,
    trim=True
)
rows = parser.parse_file('config.csv')

Parse CSV String

from cisv import parse_string

data = """
id,name,value
1,foo,100
2,bar,200
"""

rows = parse_string(data, trim=True)
# [['id', 'name', 'value'], ['1', 'foo', '100'], ['2', 'bar', '200']]

Performance

CISV uses SIMD optimizations (AVX-512, AVX2, SSE2) for high-performance parsing. The Python binding uses ctypes to call directly into the native C library with minimal overhead.

Typical performance on modern hardware:

• 500MB+ CSV files parsed in under 1 second
• 10-50x faster than pure Python CSV parsers

License

MIT