danda

Automatically prepare pandas DataFrames for analysis. Danda automatically:

• removes empty rows
• removes empty columns
• trims whitespace
• detects dates
• detects booleans
• detects categorical columns
• optimizes memory
• generates a cleaning report

A library that aims to be a drop-in replacement for pd.read_csv() and automatically:

• detects column types
• converts low-cardinality strings to category
• parses dates
• optimizes integer/float dtypes
• trims whitespace
• standardizes missing values ("", "NA", "N/A", "null", etc.)
• removes empty rows and columns
• optionally generates a cleaning report

import danda as dg

df = dg.read_csv("sales.csv")
df = dg.read_excel("sales.xlsx")
df = dg.read_parquet("sales.parquet")

Installation

pip install danda

Usage

import pandas as pd
import danda

df = pd.read_csv("employees.csv")

clean_df = df.dg.clean()

What happens automatically

• Remove empty rows
• Remove empty columns
• Strip whitespace [" John ", " John", "John "] -> John
• Normalize missing values ["", " ", "N/A", "NULL", "None", "-"] -> Nan
• Detect dates 2025-01-01->datetime64[ns]
• Detect booleans [Yes , No , TRUE , FALSE ,0, 1] -> bool
• Detect categorical columns unique values < 20 or unique ratio < 15%
• Optimize integer types int64 -> int8
• Detect IDs: do not do these [CustomerID , ZipCode , PostalCode, Phone]
• Memory optimization: Show users the improvement. Memory before : 95 MB, Memory after : 21 MB, Reduction : 77.9%
• Cleaning report:

Features

• Validation: WARNING Column Age: 12 invalid values , converted to NaN
• Profiling: Rows 145,000 Columns 24 Missing 3.2% Duplicates 512 Memory 18 MB Categorical 7 Numeric 10 Datetime 3
• suggestions: Suggestions CustomerID Looks like an ID column. Convert to string? [Y/n]
• Plugin system

dg.read_csv(
    "sales.csv",
    plugins=[
        dg.TrimWhitespace(),
        dg.ParseDates(),
        dg.RemoveDuplicates(),
        dg.NormalizeMissing(),
    ],
)

• Pandas accessor

import pandas as pd
import dataforge

df = pd.read_csv("sales.csv")

df.dg.clean()

df.dg.profile()

df.dg.optimize()

df.dg.report()

Normalize missing values