Problem: JSON Serialization Issue with Polars Time Data Type in Error Reporting

[ccosming]

Describe the bug

When validating Polars DataFrames containing columns with the Time or pl.Time data type using pandera, a PanicException occurs during error reporting due to an incompatibility with JSON serialization.

Checks

• I have checked that this issue has not already been reported.
• I have confirmed this bug exists on the latest version of pandera.
• (optional) I have confirmed this bug exists on the main branch of pandera.

Code Sample:

Steps to Reproduce:

• Define a pandera schema with a column of type pl.Time.
• Create a Polars DataFrame with data that violates the schema constraint for the Time column.
• Call validate() on the DataFrame using the defined schema.

from datetime import time, datetime

import polars as pl
import pandera.polars as pa
import pandera.typing.polars as pat

class CalendarModel(pa.DataFrameModel):
    date: pat.Series[pl.Date] = pa.Field(nullable=False,coerce=True,unique=True)
    open: pat.Series[pl.Time] = pa.Field(nullable=False, coerce=True)
    close: pat.Series[pl.Time] = pa.Field(nullable=False,coerce=True)

    @pa.dataframe_check
    def open_lower_than_close(cls, data: pa.PolarsData):
        return data.lazyframe.select(pl.col("close").gt(pl.col("open")))

df = pl.DataFrame({
    "date": [datetime(2025,1,1), datetime(2025,2,1)],
    "open": [time(9,0,0), time(16,30,0)],
    "close": [time(9,0,0), time(8,30,0)]
})

CalendarModel.to_schema().validate(df, lazy=True)

Expected behavior

pandera should gracefully handle the validation error and provide a detailed error report, including information about the failed values.

Actual behavior

A PanicException is raised with the message "not yet implemented: Writing Time64(Nanosecond) to JSON". This is because Polars' internal representation of Time as Time64(Nanosecond) does not have a direct JSON equivalent, and pandera's error reporting mechanism relies on JSON serialization.

Screenshots

Workaround:

Currently, a workaround involves catching the PanicException and manually handling the error reporting. However, this approach is not ideal and requires custom code.

Additional Context:

This issue likely stems from the inherent limitation of JSON in representing specific data types like Time. Addressing this within pandera would significantly improve its usability for Polars users.