Skip to content

Getting Started

This guide covers the essentials to get you validating DataFrames in minutes.

Installation

pip install daffy

Daffy works with pandas, Polars, Modin, or PyArrow — whatever you already have installed.

For row validation with Pydantic:

pip install 'daffy[pydantic]'

Your First Decorator

Import the decorators and add them to your functions:

from daffy import df_in, df_out

@df_in(["name", "price", "quantity"])
def process_orders(orders_df):
    # If orders_df is missing any of these columns,
    # Daffy raises an AssertionError immediately
    return orders_df

That's it! When process_orders is called, Daffy checks that the DataFrame has all required columns.

Input and Output Validation

Validate both what goes in and what comes out:

@df_in(["name", "price"])
@df_out(["name", "price", "total"])
def calculate_totals(df):
    df = df.copy()
    df["total"] = df["price"] * 1.1  # Add 10% tax
    return df

If the returned DataFrame doesn't have the expected columns, you'll know immediately.

Type Checking

Check data types alongside column names:

@df_in({
    "name": "object",
    "price": "float64",
    "quantity": "int64"
})
def process_orders(df):
    return df

If price is accidentally an int64 instead of float64, Daffy catches it.

Value Constraints

Validate column values without row iteration:

@df_in({
    "price": {"dtype": "float64", "checks": {"gt": 0}},
    "quantity": {"dtype": "int64", "checks": {"ge": 0}},
    "status": {"checks": {"isin": ["pending", "shipped", "delivered"]}}
})
def process_orders(df):
    return df

Error Messages

Daffy provides clear, actionable error messages:

AssertionError: Missing columns: ['quantity'] in function 'process_orders'
parameter 'df'. Got columns: ['name', 'price']

AssertionError: Column 'price' in function 'process_orders' parameter 'df'
has wrong dtype. Was int64, expected float64

AssertionError: Column 'price' in function 'process_orders' parameter 'df'
failed check gt: 3 values failed. Examples: [-5, 0, -1]

Next Steps