Resampling

We can resample by either:

• upsampling (moving data to a higher frequency)
• downsampling (moving data to a lower frequency)
• combinations of these e.g. first upsample and then downsample

Downsampling to a lower frequency

Polars views downsampling as a special case of the group_by operation and you can do this with group_by_dynamic and group_by_rolling - see the temporal group by page for examples.

Upsampling to a higher frequency

Let's go through an example where we generate data at 30 minute intervals:

Python Rust

DataFrame · date_range

df = pl.DataFrame(
    {
        "time": pl.date_range(
            start=datetime(2021, 12, 16),
            end=datetime(2021, 12, 16, 3),
            interval="30m",
            eager=True,
        ),
        "groups": ["a", "a", "a", "b", "b", "a", "a"],
        "values": [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0],
    }
)
print(df)

DataFrame

let df = df!(
"time" => date_range(
    "time",
    NaiveDate::from_ymd_opt(2021, 12, 16).unwrap().and_hms_opt(0, 0, 0).unwrap(),
    NaiveDate::from_ymd_opt(2021, 12, 16).unwrap().and_hms_opt(3, 0, 0).unwrap(),
    Duration::parse("30m"),
    ClosedWindow::Both,
    TimeUnit::Milliseconds, None)?,
    "groups" => &["a", "a", "a", "b", "b", "a", "a"],
    "values" => &[1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0],
)?;
println!("{}", &df);

shape: (7, 3)
┌─────────────────────┬────────┬────────┐
│ time                ┆ groups ┆ values │
│ ---                 ┆ ---    ┆ ---    │
│ datetime[μs]        ┆ str    ┆ f64    │
╞═════════════════════╪════════╪════════╡
│ 2021-12-16 00:00:00 ┆ a      ┆ 1.0    │
│ 2021-12-16 00:30:00 ┆ a      ┆ 2.0    │
│ 2021-12-16 01:00:00 ┆ a      ┆ 3.0    │
│ 2021-12-16 01:30:00 ┆ b      ┆ 4.0    │
│ 2021-12-16 02:00:00 ┆ b      ┆ 5.0    │
│ 2021-12-16 02:30:00 ┆ a      ┆ 6.0    │
│ 2021-12-16 03:00:00 ┆ a      ┆ 7.0    │
└─────────────────────┴────────┴────────┘

Upsampling can be done by defining the new sampling interval. By upsampling we are adding in extra rows where we do not have data. As such upsampling by itself gives a DataFrame with nulls. These nulls can then be filled with a fill strategy or interpolation.

Upsampling strategies

In this example we upsample from the original 30 minutes to 15 minutes and then use a forward strategy to replace the nulls with the previous non-null value:

Python Rust

upsample

out1 = df.upsample(time_column="time", every="15m").fill_null(strategy="forward")
print(out1)

let out1 = df
    .clone()
    .upsample::<[String; 0]>([], "time", Duration::parse("15m"), Duration::parse("0"))?
    .fill_null(FillNullStrategy::Forward(None))?;
println!("{}", &out1);

shape: (13, 3)
┌─────────────────────┬────────┬────────┐
│ time                ┆ groups ┆ values │
│ ---                 ┆ ---    ┆ ---    │
│ datetime[μs]        ┆ str    ┆ f64    │
╞═════════════════════╪════════╪════════╡
│ 2021-12-16 00:00:00 ┆ a      ┆ 1.0    │
│ 2021-12-16 00:15:00 ┆ a      ┆ 1.0    │
│ 2021-12-16 00:30:00 ┆ a      ┆ 2.0    │
│ 2021-12-16 00:45:00 ┆ a      ┆ 2.0    │
│ …                   ┆ …      ┆ …      │
│ 2021-12-16 02:15:00 ┆ b      ┆ 5.0    │
│ 2021-12-16 02:30:00 ┆ a      ┆ 6.0    │
│ 2021-12-16 02:45:00 ┆ a      ┆ 6.0    │
│ 2021-12-16 03:00:00 ┆ a      ┆ 7.0    │
└─────────────────────┴────────┴────────┘

In this example we instead fill the nulls by linear interpolation:

Python Rust

upsample · interpolate · fill_null

out2 = (
    df.upsample(time_column="time", every="15m")
    .interpolate()
    .fill_null(strategy="forward")
)
print(out2)

let out2 = df
    .clone()
    .upsample::<[String; 0]>([], "time", Duration::parse("15m"), Duration::parse("0"))?
    .lazy()
    .with_columns([col("values").interpolate(InterpolationMethod::Linear)])
    .collect()?
    .fill_null(FillNullStrategy::Forward(None))?;
println!("{}", &out2);

shape: (13, 3)
┌─────────────────────┬────────┬────────┐
│ time                ┆ groups ┆ values │
│ ---                 ┆ ---    ┆ ---    │
│ datetime[μs]        ┆ str    ┆ f64    │
╞═════════════════════╪════════╪════════╡
│ 2021-12-16 00:00:00 ┆ a      ┆ 1.0    │
│ 2021-12-16 00:15:00 ┆ a      ┆ 1.5    │
│ 2021-12-16 00:30:00 ┆ a      ┆ 2.0    │
│ 2021-12-16 00:45:00 ┆ a      ┆ 2.5    │
│ …                   ┆ …      ┆ …      │
│ 2021-12-16 02:15:00 ┆ b      ┆ 5.5    │
│ 2021-12-16 02:30:00 ┆ a      ┆ 6.0    │
│ 2021-12-16 02:45:00 ┆ a      ┆ 6.5    │
│ 2021-12-16 03:00:00 ┆ a      ┆ 7.0    │
└─────────────────────┴────────┴────────┘