TopolearnMost questions about data start with 'which rows' and 'in what order'.
df.filter(expr) keeps the rows where a boolean expression is true. Combine conditions with & (and), | (or) and ~ (not), wrapping each comparison in parentheses because of Python operator precedence. Passing several expressions, df.filter(cond1, cond2), means all of them must hold. Rows where the condition is null are dropped, the same as SQL's WHERE.
Handy predicates:
pl.col("city").is_in(["Oslo", "Lima"])pl.col("age").is_between(18, 65)(inclusive by default)pl.col("name").str.starts_with("A")pl.col("x").is_null()/.is_not_null()
Sorting: df.sort("age"), df.sort("age", descending=True), or several keys df.sort(["city", "age"], descending=[False, True]). Nulls go first by default; pass nulls_last=True to change that. For the n largest values use df.top_k(3, by="score") or df.bottom_k(...), which is faster than a full sort.
Positional slicing still exists for quick looks: df.head(5), df.slice(10, 5), df[2:4]. But because Polars has no index, real logic should select rows by values, not positions.
Removing duplicates is df.unique(subset=["email"], keep="first", maintain_order=True).
Example
import polars as pl
df = pl.DataFrame({
"name": ["Ana", "Ben", "Cy", "Di"],
"age": [34, 17, None, 52],
"city": ["Oslo", "Lima", "Oslo", "Rome"],
})
adults_oslo_or_rome = df.filter(
(pl.col("age") >= 18) & pl.col("city").is_in(["Oslo", "Rome"])
)
print(adults_oslo_or_rome["name"].to_list()) # ['Ana', 'Di'] (Cy's null age is dropped)
print(df.sort("age", descending=True, nulls_last=True)["name"].to_list())
# ['Di', 'Ana', 'Ben', 'Cy']
print(df.top_k(2, by="age")["name"].to_list()) # ['Di', 'Ana']Remember
- Combine conditions with
&,|,~and parentheses - Rows whose condition is null are filtered out
top_kbeats sort-then-head for the largest n
Watch out for
- Writing
cond1 and cond2: Python'sanddoesn't work on expressions; use&.