Skip to content

featurebyte.Table.sample

sample(
size: int=10,
seed: int=1234,
from_timestamp: Union[datetime, str, NoneType]=None,
to_timestamp: Union[datetime, str, NoneType]=None,
after_cleaning: bool=False
) -> DataFrame

Description

Returns a DataFrame that contains a random selection of rows of the table based on a specified time range, size, and seed for sampling control. By default, the materialization process occurs before any cleaning operations that were defined at the column level.

Parameters

  • size: int
    default: 10
    Maximum number of rows to sample, with an upper bound of 10,000 rows.

  • seed: int
    default: 1234
    Seed to use for random sampling.

  • from_timestamp: Union[datetime, str, NoneType]
    Start of date range to sample from.

  • to_timestamp: Union[datetime, str, NoneType]
    End of date range to sample from.

  • after_cleaning: bool
    default: False
    Whether to apply cleaning operations.

Returns

  • DataFrame
    Sampled rows from the table.

Examples

Sample 3 rows from the table.

>>> catalog.get_table("GROCERYPRODUCT").sample(3)
                     GroceryProductGuid ProductGroup
0  e890c5cb-689b-4caf-8e49-6b97bb9420c0       Épices
1  5720e4df-2996-4443-a1bc-3d896bf98140         Chat
2  96fc4d80-8cb0-4f1b-af01-e71ad7e7104a        Pains

Sample 3 rows from the table with timestamps.

>>> event_table = catalog.get_table("GROCERYINVOICE")
>>> event_table["Amount"].update_critical_data_info(
...     cleaning_operations=[
...         fb.MissingValueImputation(imputed_value=0),
...     ]
... )

>>> event_table.sample(
...     size=3,
...     seed=111,
...     from_timestamp=datetime(2019, 1, 1),
...     to_timestamp=datetime(2023, 12, 31),
...     after_cleaning=True,
... )

See Also