Pandas Cheatsheet (Avoid indexes)

pandas_cheatsheet.py

import pandas as pd

### Data import ###

# https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.to_csv.html 
df = pd.read_csv("file.csv")  


### Data types ###

df.info()

### DataFrame operations ###

# Subset, rename, retype columns
df[["a", "b"]]
df.rename(columns={"y":"year"})
df.astype({"a": int})
df.sort_values("col", ascending=False)

# Unique and duplicated values
df["col"].unique  # numpy array of unique values
df["col"].is_unique
df.drop_duplicates(subset=["col"])
df.duplicated(subset="col")  # Boolean Series

# Summary statistics
df["col"].value_counts(dropna=False)
df["col"].isna().sum()

# Table join 
# Warning: NULLs are joined together, unlike SQL
# https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.DataFrame.merge.html

# Use on="key" if joining columns are named the same
df.merge(df2, how="left", left_on="lk", right_on="rk")

# SQL-style GROUP BY named aggregation functions 
# Common functions: size, count (non-NAs), nunique (count distinct), first, max, min, mean
df.groupby("group", as_index=False).agg(
   a_sum=('a', "sum"),
   d_range=('d', lambda x: x.max() - x.min())
)

### Data export
df.to_csv("file.csv", index=False)