csv-dataclass-gen 1.0.0

Generate Python dataclasses, loaders, from CSV/TSV files

CSV Dataclass Generator

Generate Python dataclasses and loader functions from CSV/TSV files.

Features

• Automatic Type Inference: Detects int, float, and str types based on a sample of rows.
• Dialect Detection: Automatically identifies CSV delimiters (including TSV).
• Name Sanitization: Converts CSV column headers into valid Python identifiers.
• Iterator Loading: Generates a loading generator that yields dataclass instances.

Installation

pip install csv-dataclass-gen

Usage

Command Line Interface (CLI)

The package provides a csv-dataclass-gen command.

# Generate code to stdout
csv-dataclass-gen data.csv

# Generate code and save to a directory
csv-dataclass-gen data.csv --output ./generated_models/

# Specify a custom class name and sample size for type inference
csv-dataclass-gen data.csv --name my_custom_data --sample-size 500

CLI Help Message:

Usage: csv-dataclass-gen [OPTIONS] INPUT_FILE

  Generate dataclass and loader code from CSV files.

Options:
  -o, --output TEXT          Output directory for generated files. "-" outputs
                             the result to stdout.
  -s, --sample-size INTEGER  Number of rows to sample for type inference
  -n, --name TEXT            Alternative name for the generated name. Snake
                             case / spaced words is recommended.
  --help                     Show this message and exit.

Arguments:

• INPUT_FILE: Path to the CSV/TSV file (required).

Example Generated Code

Given a CSV like users.csv:

id,user_name,score
1,alice,95.5
2,bob,88.0

The generator will produce:

from dataclasses import dataclass
from pathlib import Path
from typing import Iterator
import csv

@dataclass
class Users:
    id: int  # Original: "id"
    user_name: str  # Original: "user_name"
    score: float  # Original: "score"

def load_users(csv_path: Path, max_rows: int | None = None, delimiter: str = ',') -> Iterator[Users]:
    # ... loading logic ...
    pass

Development

Code Template

The generator comes with a Dataclass and Loader template.

Dependencies

This project uses uv for dependency management.

uv sync --all-groups

Running Tests

We use pytest for testing.

uv run pytest

Tests include:

• tests/test_name_sanitizer.py: Logic for sanitizing names into different formats.
• tests/test_type_inferrer.py: Logic for detecting data types.
• tests/test_csv_analyzer.py: Logic for CSV structure analysis.
• tests/test_code_gen_e2e.py: End-to-end tests that generate code and verify it by reconstructing the original CSV.

License

MIT License – see the LICENSE file for details.