merlin.schema package

class merlin.schema.Schema(column_schemas=None)

Bases: object

A collection of column schemas for a dataset.

property column_names

select(selector) → Schema

Select matching columns from this Schema object using a ColumnSelector

Parameters:

selector (ColumnSelector) – Selector that describes which columns match

Returns:

New object containing only the ColumnSchemas of selected columns

Return type:

Schema

apply(selector) → Schema

excluding(selector) → Schema

Select non-matching columns from this Schema object using a ColumnSelector

Parameters:

selector (ColumnSelector) – Selector that describes which columns match

Returns:

New object containing only the ColumnSchemas of selected columns

Return type:

Schema

apply_inverse(selector) → Schema

select_by_tag(tags: str | Tags | List[str | Tags], pred_fn=None) → Schema

Select columns from this Schema that match ANY of the supplied tags.

Parameters:

• tags (List[Union[str, Tags]] :) – List of tags that describes which columns match

• pred_fn (any or all) – Predicate function that decides if the column should be selected. Receives iterable of bool values indicating whether each of the provided tags is present on a column schema. Returning True selects this column, False will not return that column.

Returns:

New object containing only the ColumnSchemas of selected columns

Return type:

Schema

excluding_by_tag(tags) → Schema

remove_by_tag(tags) → Schema

select_by_name(names: List[str]) → Schema

Select matching columns from this Schema object using a list of column names

Parameters:

names (List[str] :) – List of column names that describes which columns match

Returns:

New object containing only the ColumnSchemas of selected columns

Return type:

Schema

excluding_by_name(col_names: List[str])

Remove columns from this Schema object by name

Parameters:

col_names (List[str]) – Names of the column to remove

Returns:

New Schema object after the columns are removed

Return type:

Schema

remove_col(col_name: str) → Schema

Remove a column from this Schema object by name

Parameters:

col_name (str) – Name of the column to remove

Returns:

This Schema object after the column is removed

Return type:

Schema

without(col_names: List[str]) → Schema

get(col_name: str, default: ColumnSchema | None = None) → ColumnSchema

Get a ColumnSchema by name

Parameters:

• col_name (str) – Name of the column to get

• default (ColumnSchema :) –

  Default value to return if column is not found.

  (Default value = None)

Returns:

Retrieved column schema (or default value, if not found)

Return type:

ColumnSchema

property first: ColumnSchema

Returns the first ColumnSchema in the Schema. Useful for cases where you select down to a single column via select_by_name or select_by_tag, and just want the value

Returns:

The first column schema present in this Schema object

Return type:

ColumnSchema

Raises:

ValueError – If this Schema object contains no column schemas

to_pandas() → DataFrame

Convert this Schema object to a pandas DataFrame

Returns:

DataFrame containing the column schemas in this Schema object

Return type:

pd.DataFrame

copy() → Schema

Return a copy of the schema

class merlin.schema.ColumnSchema(name: str, tags: ~merlin.schema.tags.TagSet | ~typing.List[str | ~merlin.schema.tags.Tags] | None = <factory>, properties: ~typing.Dict | None = <factory>, dtype: ~merlin.dtypes.base.DType | None = None, is_list: bool | None = None, is_ragged: bool | None = None, dims: dataclasses.InitVar[typing.Union[typing.Tuple, merlin.dtypes.shape.Shape]] = None)

Bases: object

A schema containing metadata of a dataframe column.

name: str

tags: TagSet | List[str | Tags] | None

properties: Dict | None

dtype: DType | None = None

is_list: bool | None = None

is_ragged: bool | None = None

dims: dataclasses.InitVar[Union[Tuple, merlin.dtypes.shape.Shape]] = None

property shape

with_name(name: str) → ColumnSchema

Create a copy of this ColumnSchema object with a different column name

Parameters:

name (str) – New column name

Returns:

Copied object with new column name

Return type:

ColumnSchema

with_tags(tags: str | Tags) → ColumnSchema

Create a copy of this ColumnSchema object with different column tags

Parameters:

tags (Union[str, Tags]) – New column tags

Returns:

Copied object with new column tags

Return type:

ColumnSchema

with_properties(properties: dict) → ColumnSchema

Create a copy of this ColumnSchema object with different column properties

Parameters:

properties (dict) – New column properties

Returns:

Copied object with new column properties

Return type:

ColumnSchema

Raises:

TypeError – If properties are not a dict

with_dtype(dtype, is_list: bool | None = None, is_ragged: bool | None = None) → ColumnSchema

Create a copy of this ColumnSchema object with different column dtype

Parameters:

• dtype (np.dtype) – New column dtype

• is_list (bool :) –

  Whether rows in this column contain lists.

  (Default value = None)

• is_ragged (bool :) –

  Whether lists in this column have varying lengths.

  (Default value = None)

Returns:

Copied object with new column dtype

Return type:

ColumnSchema

with_shape(shape: Tuple | Shape) → ColumnSchema

Create a copy of this object with a new shape

Parameters:

shape (Union[Tuple, Shape]) – Object to set as shape, must be either a tuple or Shape.

Returns:

A copy of this object containing the provided shape value

Return type:

ColumnSchema

Raises:

TypeError – If value is not either a tuple or a Shape

property int_domain: Domain | None

property float_domain: Domain | None

property value_count: Domain | None

class merlin.schema.Tags(value)

Bases: Enum

Standard tags used in the Merlin ecosystem

CATEGORICAL = 'categorical'

CONTINUOUS = 'continuous'

LIST = 'list'

SEQUENCE = 'sequence'

TEXT = 'text'

TOKENIZED = 'tokenized'

TIME = 'time'

EMBEDDING = 'embedding'

ID = 'id'

USER = 'user'

ITEM = 'item'

SESSION = 'session'

CONTEXT = 'context'

TARGET = 'target'

REGRESSION = 'regression'

CLASSIFICATION = 'classification'

BINARY = 'binary'

MULTI_CLASS = 'multi_class'

USER_ID = 'user_id'

ITEM_ID = 'item_id'

SESSION_ID = 'session_id'

TEXT_TOKENIZED = 'text_tokenized'

BINARY_CLASSIFICATION = 'binary_classification'

MULTI_CLASS_CLASSIFICATION = 'multi_class_classification'