Equivalent of pandas.Series.dt.ceil in polars

I’m trying to round timestamp to the next minutes in polars.

For example:

• 2023-01-01 10:05:00 should stay 2023-01-01 10:05:00
• 2023-01-01 10:05:01 should be 2023-01-01 10:06:00

This works in pandas with ceil:

import polars as pl
import datetime

df = pl.DataFrame(
    {'timestamp' :[
        datetime.datetime(2023, 1, 1, 10, 5, 0),
        datetime.datetime(2023, 1, 1, 10, 5, 30),
        datetime.datetime(2023, 1, 1, 10, 6, 0),
        datetime.datetime(2023, 1, 1, 10, 6, 1),
    ]
    }
)

df['timestamp'].to_pandas().dt.ceil('1min')

The only way I found in polars is the following:

df.with_columns(
    pl.when(pl.col('timestamp').dt.truncate('1m') == pl.col('timestamp'))
    .then(pl.col('timestamp'))
    .otherwise(pl.col('timestamp').dt.truncate('1m') + datetime.timedelta(minutes=1))
)

>Solution :

you can use polars-xdt for this

installation:

pip install polars-xdt

usage:

import polars_xdt as xdt

df.with_columns(timestamp_ceil = xdt.ceil('timestamp', '1m'))

shape: (4, 2)
┌─────────────────────┬─────────────────────┐
│ timestamp           ┆ timestamp_ceil      │
│ ---                 ┆ ---                 │
│ datetime[μs]        ┆ datetime[μs]        │
╞═════════════════════╪═════════════════════╡
│ 2023-01-01 10:05:00 ┆ 2023-01-01 10:05:00 │
│ 2023-01-01 10:05:30 ┆ 2023-01-01 10:06:00 │
│ 2023-01-01 10:06:00 ┆ 2023-01-01 10:06:00 │
│ 2023-01-01 10:06:01 ┆ 2023-01-01 10:07:00 │
└─────────────────────┴─────────────────────┘