Column Constraints API

MeanBetween

`ydata.constraints.columns.constraint.MeanBetween`

Bases: Interval

Check that the mean of each column falls inside (lower_bound, upper_bound].

Parameters:

Name	Type	Description	Default
`lower_bound`	`int \| float`	Lower bound (exclusive).	required
`upper_bound`	`int \| float`	Upper bound (inclusive).	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

Example

MeanBetween(lower_bound=30, upper_bound=40, columns="age")

StandardDeviationBetween

`ydata.constraints.columns.constraint.StandardDeviationBetween`

Bases: Interval

Check that the standard deviation of each column falls inside (lower_bound, upper_bound].

Parameters:

Name	Type	Description	Default
`lower_bound`	`int \| float`	Lower bound (exclusive).	required
`upper_bound`	`int \| float`	Upper bound (inclusive).	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

QuantileBetween

`ydata.constraints.columns.constraint.QuantileBetween`

Bases: Interval

Check that a quantile of each column falls inside (lower_bound, upper_bound].

Parameters:

Name	Type	Description	Default
`quantile`	`float`	Quantile to compute, between 0 and 1 (e.g. `0.95`).	required
`lower_bound`	`int \| float`	Lower bound (exclusive).	required
`upper_bound`	`int \| float`	Upper bound (inclusive).	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

Example

QuantileBetween(quantile=0.95, lower_bound=0, upper_bound=5000, columns="amount")

MinBetween

`ydata.constraints.columns.constraint.MinBetween`

Bases: Interval

Check that the minimum value of a column falls inside (lower_bound, upper_bound].

Parameters:

Name	Type	Description	Default
`lower_bound`	`int \| float`	Lower bound on the column minimum (exclusive).	required
`upper_bound`	`int \| float`	Upper bound on the column minimum (inclusive).	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

MaxBetween

`ydata.constraints.columns.constraint.MaxBetween`

Bases: Interval

Check that the maximum value of a column falls inside (lower_bound, upper_bound].

Parameters:

Name	Type	Description	Default
`lower_bound`	`int \| float`	Lower bound on the column maximum (exclusive).	required
`upper_bound`	`int \| float`	Upper bound on the column maximum (inclusive).	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

NullRateLowerThan

`ydata.constraints.columns.constraint.NullRateLowerThan`

Bases: LowerThan

Check that the proportion of null values in a column is at most a threshold.

Unlike :class:NullValuesCountLowerThan, which requires knowing the absolute count, this constraint uses a rate in [0, 1] and therefore works regardless of dataset size.

Parameters:

Name	Type	Description	Default
`value`	`float`	Maximum allowed null rate (inclusive), e.g. `0.05` for 5 %.	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

NullValuesCountLowerThan

`ydata.constraints.columns.constraint.NullValuesCountLowerThan`

Bases: LowerThan

Check that the number of null (NaN) values in each column is at most value.

Nulls are detected with numpy.isnan, so the columns must be numeric. For a size-independent check, use :class:NullRateLowerThan.

Parameters:

Name	Type	Description	Default
`value`	`int \| float`	Maximum allowed number of nulls (inclusive).	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

UniqueValuesBetween

`ydata.constraints.columns.constraint.UniqueValuesBetween`

Bases: Interval

Check that the number of distinct values in each column falls inside (lower_bound, upper_bound].

Parameters:

Name	Type	Description	Default
`lower_bound`	`int \| float`	Lower bound on the distinct count (exclusive).	required
`upper_bound`	`int \| float`	Upper bound on the distinct count (inclusive).	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

Constant

`ydata.constraints.columns.constraint.Constant`

Bases: Interval

Check that each column holds a single distinct value.

Parameters:

Name	Type	Description	Default
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

SumLowerThan

`ydata.constraints.columns.constraint.SumLowerThan`

Bases: LowerThan

Check that the sum of each column is at most value.

Parameters:

Name	Type	Description	Default
`value`	`int \| float`	Maximum allowed sum (inclusive).	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

Interval

`ydata.constraints.columns.constraint.Interval`

Bases: ColumnConstraint

Check that a statistic of each column falls inside an interval.

check computes one scalar from the column (e.g. numpy.mean); the column passes when that value lies in the interval. Most column constraints are specializations of this class.

Parameters:

Name	Type	Description	Default
`check`	`Callable`	Callable `Series -> scalar` that computes the statistic.	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`
`available_columns`	`list[str] \| None`	Accepted for interface compatibility with row constraints; not used by column constraints.	`None`
`entity`	`str \| None`	Accepted for interface compatibility with row constraints; not used by column constraints.	`None`
`lower_bound`	`int \| float`	Lower bound of the interval. Defaults to `-inf`.	`-inf`
`upper_bound`	`int \| float`	Upper bound of the interval. Defaults to `inf`.	`inf`
`closed`	`str`	Which bounds are inclusive: `"right"` (default, lower exclusive and upper inclusive), `"left"`, `"both"` or `"neither"`.	`'right'`

Example

Interval(check=numpy.median, columns="age", lower_bound=30, upper_bound=45)

`validate(dataset)`

Validate the constraint against a dataset.

Evaluates the check on the full column (no sampling).

Parameters:

Name	Type	Description	Default
`dataset`	`Dataset`	Dataset to check.	required

Returns:

Name	Type	Description
`pdDataFrame`	`DataFrame`	One row with one boolean per constrained column, `True` when the column satisfies the constraint.

Equal

`ydata.constraints.columns.constraint.Equal`

Bases: Interval

Check that a statistic of each column equals a value, within a tolerance.

Parameters:

Name	Type	Description	Default
`check`	`Callable`	Callable `Series -> scalar` that computes the statistic.	required
`value`	`int \| float`	Expected value of the statistic.	required
`tolerance`	`int \| float`	Allowed absolute deviation from `value` (inclusive). Defaults to 0.	`0`
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

Example

Equal(check=numpy.mean, value=0.5, tolerance=0.05, columns="rate")

GreaterThan

`ydata.constraints.columns.constraint.GreaterThan`

Bases: Interval

Check that a statistic of each column is strictly greater than a value.

Parameters:

Name	Type	Description	Default
`check`	`Callable`	Callable `Series -> scalar` that computes the statistic.	required
`value`	`int \| float`	The statistic must be greater than this value.	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

Example

GreaterThan(check=numpy.min, value=0, columns="amount")

LowerThan

`ydata.constraints.columns.constraint.LowerThan`

Bases: Interval

Check that a statistic of each column is lower than or equal to a value.

Parameters:

Name	Type	Description	Default
`check`	`Callable`	Callable `Series -> scalar` that computes the statistic.	required
`value`	`int \| float`	Maximum allowed value of the statistic (inclusive).	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`

Example

LowerThan(check=numpy.max, value=120, columns="age")

ColumnCustomConstraint

`ydata.constraints.columns.constraint.ColumnCustomConstraint`

Bases: ColumnConstraint

Check a user-defined condition on each column as a whole.

The callable receives the full column as a pandas.Series and returns a boolean: True when the column satisfies the condition.

Parameters:

Name	Type	Description	Default
`check`	`Callable`	Callable `Series -> bool` applied to each column.	required
`columns`	`str \| list[str] \| None`	Column(s) to check.	`None`
`name`	`str \| None`	Human-readable label shown in engine summaries.	`None`
`available_columns`	`list[str] \| None`	Accepted for interface compatibility with row constraints; not used by column constraints.	`None`
`entity`	`str \| None`	Accepted for interface compatibility with row constraints; not used by column constraints.	`None`

Example

ColumnCustomConstraint(check=lambda s: s.is_monotonic_increasing, columns="date")

`validate(dataset)`

Validate the constraint against a dataset.

Evaluates the check on the full column (no sampling).

Parameters:

Name	Type	Description	Default
`dataset`	`Dataset`	Dataset to check.	required

Returns:

Name	Type	Description
`pdDataFrame`	`DataFrame`	One row with one boolean per constrained column, `True` when the column satisfies the constraint.