String

[Polars Documentation]

In the string module, OpenDP currently only supports parsing to temporal data types.

import polars as pl
import opendp.prelude as dp
dp.enable_features("contrib")

context = dp.Context.compositor(
    # Many columns contain mixtures of strings and numbers and cannot be parsed as floats,
    # so we'll set `ignore_errors` to true to avoid conversion errors.
    data=pl.scan_csv(dp.examples.get_france_lfs_path(), ignore_errors=True),
    privacy_unit=dp.unit_of(contributions=36),
    privacy_loss=dp.loss_of(epsilon=1.0, delta=1e-7),
    split_evenly_over=2,
)

Strptime, To Date, To Datetime, To Time

Dates can be parsed from strings via .str.strptime, and its variants .str.to_date, .str.to_datetime, and .str.to_time.

query = (
    context.query()
    .with_columns(pl.col.YEAR.cast(str).str.to_date(format=r"%Y"))
    .group_by("YEAR")
    .agg(dp.len())
)
query.release().collect().sort("YEAR")

shape: (9, 2)

YEAR	len
date	u32
2005-01-01	342193
2006-01-01	339683
2007-01-01	350429
2008-01-01	348574
2009-01-01	416966
2010-01-01	500385
2011-01-01	517166
2012-01-01	515460
2013-01-01	480615

While Polars supports automatic inference of the datetime format from reading the data, doing so can lead to situations where the data-dependent inferred format changes or cannot be inferred upon the addition or removal of a single individual, resulting in an unstable computation. For this reason, the format argument is required.

OpenDP also does not allow parsing strings into nanosecond datetimes, because the underlying implementation throws data-dependent errors (not private) for certain inputs.

query = (
    context.query()
    .with_columns(pl.col.YEAR.cast(str).str.to_datetime(format=r"%Y", time_unit="ns"))
    .group_by("YEAR")
    .agg(dp.len())
)
try:
    query.release()
    assert False, "unreachable!"
except dp.OpenDPException as err:
    assert "Nanoseconds are not currently supported" in str(err)

Parsed data can then be manipulated with temporal expressions.