lazytype

Write type hints, implement optional features, define pydantic models without up-front import delays, and more without having to import slow modules; using a lazytype.LazyType, you can wrap a slow-loading class so that its module doesn't load until you instantiate it or run an actual type check.

Pydantic models are also supported through the LazyField interface, though you don't need to have Pydantic installed to use LazyTypes.

Installation

For users:

pip install lazytype

Optionally make sure pydantic is installed with the pydantic option (if you plan to use LazyFields):

pip install lazytype[pydantic]

For developers, clone this repository, change to its directory, and run:

flit install --symlink

Examples

Regular LazyTypes

The interface is the same as the wrapped object:

>>> a = LazyType['numpy.ndarray']((3, 2))
>>> a
<Lazy array([[-1.49166815e-154, -2.68679856e+154],
                [ 1.48219694e-323,  0.00000000e+000],
                [ 0.00000000e+000,  4.17201348e-309]])>
>>> a._instance
array([[-2.00000000e+000,  2.32036240e+077],
        [ 1.48219694e-323,  0.00000000e+000],
        [ 0.00000000e+000,  4.17201348e-309]])
>>> a[:] = 0
>>> a
array([[0., 0.],
        [0., 0.],
        [0., 0.]])
>>> a.dtype
dtype('float64')
>>> isinstance(a._instance, type(a))
True

You can optionally require that a check run for the module availability at definition time:

# this works if you have "numpy" installed
>>> LazyArray = LazyType['numpy.ndarray', 'strict':True]
lazytype.Lazyndarray

# this doesn't (unless you have a package called "numpay"...)
>>> LazyArray = LazyType['numpay.ndarray', 'strict':True]
ImportError: Strict check for module numpay availability failed

pydantic Fields with LazyFields

You can also use LazyTypes with Pydantic models to specify data types and validators. This requires some extra methods provided by the LazyField class; you can specify any built-in Pydantic field-type to use for schema validation, followed by the arguments you would use to create a new LazyType. This allows you to map an existing field-type to a custom, slow-loading field-type that can trivially accept the same input arguments, all without loading the wrapped class's module until your model is instantiated.

Create a lazy-loading field for astropy.time.Time using the built-in datetime string validator:

>>> from datetime import datetime
>>> LazyField[datetime, 'astropy.time.Time', 'strict':True]
lazytype.LazyFieldTime

Actually use the field in a pydantic model:

>>> from pydantic import BaseModel
>>> class LazyTest(BaseModel):
...     foo: str
...     time: LazyField[datetime, 'astropy.time.Time', 'strict':True]

See the JSON schema of the resulting model:

>>> LazyTest.schema()
{'title': 'LazyTest',
 'type': 'object',
 'properties': {'foo': {'title': 'Foo', 'type': 'string'},
  'time': {'title': 'Time', 'type': 'string', 'format': 'date-time'}},
 'required': ['foo', 'time']}

Actually instantiate something, forcing astropy.time.Time to load:

>>> t = LazyTest(foo='bar', time='2019-11-29 13:40:29.197')
>>> t.time
<Lazy <Time object: scale='utc' format='iso' value=2019-11-29 13:40:29.197>>
>>> t.time.gps
1259070047.197

Provide additional schema annotations, e.g. providing an example input value:

>>> class LazyTest(BaseModel):
...     foo: str
...     time: LazyField[str::{'example': '2019-11-29 13:40:29.197'},
...                     'astropy.time.Time', 'strict':True]
>>> LazyTest.schema()
{'title': 'LazyTest',
 'type': 'object',
 'properties': {'foo': {'title': 'Foo', 'type': 'string'},
  'time': {'title': 'Time',
   'type': 'string',
   'example': '2019-11-29 13:40:29.197'}},
 'required': ['foo', 'time']}