checkpl

Ergonomic inline validation for Polars DataFrames.

Installation

pip install checkpl

Quick Start

import polars as pl
from assert_polars import verify, is_uniq

df = pl.DataFrame({
    "id": [1, 2, 3],
    "city": ["NYC", "LA", "NYC"],
    "year": [2020, 2020, 2021]
})

# Validate with Polars expressions
df.pipe(verify(pl.col("year") >= 2020))

# Validate uniqueness
df.pipe(verify(is_uniq("id")))

# Composite key uniqueness
df.pipe(verify(is_uniq("city", "year")))

# Chain validations
(
    df
    .pipe(verify(pl.col("year") >= 2020))
    .pipe(verify(is_uniq("id")))
    .filter(pl.col("city") == "NYC")
)

Features

• Pipeable: Integrates with Polars' .pipe() for fluent data pipelines
• Native expressions: Use any Polars boolean expression directly
• Predicates: Built-in checks like is_uniq() for common validations
• Clear errors: CheckError with descriptive messages and failure counts
• LazyFrame support: Works with both DataFrame and LazyFrame

API

verify(check)

Validate a DataFrame using a Polars expression or predicate.

# With Polars expression
df.pipe(verify(pl.col("price") > 0))

# With predicate
df.pipe(verify(is_uniq("id")))

Raises CheckError if validation fails, returns DataFrame unchanged if valid.

is_uniq(*cols)

Check that column(s) have no duplicate values.

df.pipe(verify(is_uniq("id")))           # Single column
df.pipe(verify(is_uniq("id", "year")))   # Composite key

Error Handling

from assert_polars import CheckError

try:
    df.pipe(verify(is_uniq("id")))
except CheckError as e:
    print(f"Check '{e.check_name}' failed: {e}")

License

MIT