df.assert_schema(expected_schema)

[KDruzhkin]

df.assert_schema(expected_schema)

[JulianCologne]

yes, agree!
Have already shared by idea about something more general .assert / .expect here: #11064 (comment)

Would love to see:

.assert(...) or .expect

• assert that some conditions are met like: datatypes, max < x, min > 0, schema = ...
• with configuration options to turn on for "testing" (full evalutation) or "production" (ignore assertions)
• df.select(...).assert(col(x)>0, pl.max(y) < 100).with_columns(...).assert(...).group_by...

.inspect(...)

• allow printing / debugung current df values
• df.select(...).with_columns(...).inspect(lambda df: print(df.schema)).select(...).inspect(lambda df: print(df.describe()).group_by...

[KDruzhkin]

Interesting.

But schemas are different from other properties. You can get the schema of a LazyFrame without collecting it.

The more general assert and inspect methods of LazyFrames would only add delayed actions. (Still useful.)

[KDruzhkin]

Wrote #16311 specifically about dropping/forgetting/ignoring values that are used only for assertions.

[stinodego]

You can already do this using pipe:

import polars as pl

def assert_schema(
    lf: pl.LazyFrame, schema: dict[str, pl.PolarsDataType]
) -> pl.LazyFrame:
    if lf.schema != schema:
        msg = (
            "Wrong LazyFrame schema:\n"
            f"• expected: '{schema}',\n"
            f"• observed: '{dict(lf.schema)}'."
        )
        raise AssertionError(msg)
    return lf

lf = pl.LazyFrame({"a": [1, 2, 3], "b": [4.0, 5.0, 6.0]})
lf.select("a").pipe(assert_schema, schema={"a": pl.Float64})

AssertionError: Wrong LazyFrame schema:
• expected: '{'a': Float64}',
• observed: '{'a': Int64}'.

I don't think this is worth adding to the Polars API.

For a more general 'assertion / raising' utility, please see #11064

[KDruzhkin]

You can already do this using pipe:

Nice. I somehow missed pipe.

Maybe, it deserves a more prominent place in the docs? The user guide mentions it only in passing (Pipe littering: "Don't").