Skip to main content

Type Annotation

Type annotation is optional but highly recommended in dataset.sh

As a dataset publisher, this small inconvinence can benefit your future users by:

  • making it easier for others to understand and work with your dataset.
  • enabel us to genreate code in multiple languages, which enhances the developer experience significantly.

We create easytype to help dataset creators provide type annotations of their datasets.

easytype Tutorial​

Basic Example​

from easytype import TypeBuilder

TranslationPair = TypeBuilder.create(
    'TranslationPair',
    orginal= str,
    translated=str,
)

Primitive types​

easytype supports the following primitive types:

  • int
  • float
  • str
  • bool
  • list
  • dict
  • any: via typing.Any
all supported primitive types
from easytype import TypeBuilder
from typing import Any

AllPrimitiveTypes = TypeBuilder.create(
    'AllPrimitiveTypes',
    int_value=int,
    float_value=float,
    str_value=str,
    bool_value=bool,
    list_value=list
    dict_value=dict
    any_value=Any # you can escape using Any type
)

Paramertized types​

easytype supports the following primitive types:

  • list (python 3.9+) or typing.List
  • dict (python 3.9+) or typing.Dict
  • tuple (python 3.9+) or typing.Tuple
  • Optional (typing.Optional)
  • Union (typing.Union)
supported paramertized types
from easytype import TypeBuilder
import typing


ParamertizedTypeExample = TypeBuilder.create(
    'ParamertizedTypeExample',
    list_of_int_values=list[int],
    optional_int_value=typing.Optional[int],
    union_type=typing.Union[str, int],
    dict_str_to_int_value=dict[str, int],
    tuple_example=tuple[str, int],
)

# You may have to use typing.List, typing.Dict or typing.Tuple before python 3.9

Referencing other types​

Self Reference (recursive)​

Self Reference
from easytype import TypeBuilder

TreeNode = TypeBuilder.create(
    'TreeNode',
    name= str,
    children=list['TreeNode'],
)


Using other types​

Reference​
from easytype import TypeBuilder, TypeReference

WikidataEntity = TypeBuilder.create(
    'WikidataEntity',
    id=str,
    label=str
)

WrittenWork = TypeBuilder.create(
    'WrittenWork',
    name=str,
    authors=list[WikidataEntity],
    country_of_origin=WikidataEntity,
    entity_id=str,
).reference(WikidataEntity) # You need to reference keyword if used this way.


Inline Types​

You can also create type inline.

from easytype import TypeBuilder, TypeReference

WrittenWork = TypeBuilder.create(
    'WrittenWork',
    name=str,
    country_of_origin=dict(
        id=str,
        label=str
    ),
    entity_id=str,
).reference(WikidataEntity) # You need to reference keyword if used this way.


Use TypeReference With reference function​
export typing with reference to other types
from easytype import TypeBuilder, TypeReference

WikidataEntity = TypeBuilder.create(
    'WikidataEntity',
    id=str,
    label=str
)

WrittenWork = TypeBuilder.create(
    'WrittenWork',
    name=str,


    # create a type reference with TypeReference function.
    country_of_origin=TypeReference('WikidataEntity'),

    # use as a parameter
    authors=list['WikidataEntity'],  

    entity_id=str,
).reference(WikidataEntity) # You need to reference keyword if used this way.

Provide type annotation with dataset.sh API​

Provide type annotation with datafact​

You can provide type annotation by editing type.py in your datafact project, for detailed instruction, please read datafact's documentation.

Link: datafact documentation