Polars cheat sheet

import polars as pl

df = pl.read_csv("sales.csv", try_parse_dates=True)
df = pl.read_excel("sales.xlsx", sheet_name="Data")
lf = pl.scan_parquet("sales/*.parquet")  # lazy: nothing runs yet
lf.collect_schema()                      # names and types, no run

df.select("date", "region", "amount")
df.filter((pl.col("region") == "North") & (pl.col("amount") > 100))
df.filter(pl.col("region").is_in(["North", "South"]))
df.unique(subset="order_id", keep="first")
df.head(10)

df.with_columns(
    net=pl.col("amount") - pl.col("discount"),
    month=pl.col("date").dt.truncate("1mo"),
    band=pl.when(pl.col("amount") >= 1000).then(pl.lit("Large"))
           .otherwise(pl.lit("Small")),
)
df.rename({"amt": "amount"})

(df.group_by("region")
   .agg(total=pl.col("amount").sum(),
        orders=pl.len(),
        customers=pl.col("customer_id").n_unique())
   .sort("total", descending=True))

orders.join(customers, on="customer_id", how="left")
orders.join(returns, on="order_id", how="anti")   # not returned
orders.join(prices, left_on="sku", right_on="product_sku")
pl.concat([jan, feb])                             # stack rows

df.with_columns(
    share=pl.col("amount") / pl.col("amount").sum().over("region"),
    rank=pl.col("amount").rank("dense", descending=True)
           .over("region"),
)
# Latest order per customer
df.filter(pl.col("date") == pl.col("date").max().over("customer_id"))

df.pivot(on="month", index="region", values="amount",
         aggregate_function="sum")
wide.unpivot(index="region", variable_name="month",
             value_name="amount")

pl.col("name").str.strip_chars().str.to_titlecase()
pl.col("email").str.split("@").list.get(1)
pl.col("text_date").str.to_date("%d/%m/%Y")
pl.col("date").dt.year()
pl.col("amount").fill_null(0)

result = (
    pl.scan_parquet("sales/*.parquet")
      .filter(pl.col("year") == 2026)
      .group_by("region")
      .agg(pl.col("revenue").sum())
      .sort("region")       # order is not guaranteed otherwise
      .collect()            # streaming engine by default in 2.0
)

pl.sql("""
    SELECT region, SUM(amount) AS total
    FROM df
    GROUP BY region
""").collect()

df.write_parquet("out/sales.parquet")
df.write_csv("out/sales.csv")
df.write_excel("out/sales.xlsx")   # needs xlsxwriter
df.to_pandas()                     # needs pandas and pyarrow