In our Polars vs. Pandas article, we showed that Polars is faster. The reason it’s faster is its Expression API. In this article, we’ll take a closer look at the Polars Expression API and why it matters for performance.
Instead of telling Pandas how to do something step-by-step, Polars asks you to define what you want. It then optimizes the query and runs it using all your CPU cores.
The two magic words are:
pl.col("name"): Selects a column.pl.lit("value"): Creates a “literal” value.
1. select(): Choosing and Renaming Columns
select is for columns.
import polars as pl
df = pl.DataFrame({
"temp_c": [10, 20, 30],
"location": ["A", "B", "A"]
})
# Select a column and rename it with an alias
df.select(
pl.col("temp_c").alias("temp_in_celsius")
)2. filter(): Choosing Rows
filter is for rows.
# Find all rows where temp_c is greater than 15
df.filter(
pl.col("temp_c") > 15
)3. with_columns(): Creating New Columns
This is the most powerful part. You can create new columns based on expressions.
# Let's create a new 'temp_f' column
df.with_columns([
(pl.col("temp_c") * 9/5 + 32).alias("temp_f")
])Chaining It All Together
This is the “Polars way.” Find all temps over 15 from location A, convert them to Fahrenheit, and select just that column.
results = (
df.filter(
(pl.col("temp_c") > 15) & (pl.col("location") == "A")
)
.with_columns([
(pl.col("temp_c") * 9/5 + 32).alias("temp_f")
])
.select("temp_f")
)
print(results)
