pydantic-panel 0.1.3

Top-level package for pydantic-panel.

Edit pydantic models with panel.

This is just a small little project i made mostly for my own use and decided to share. Its limited in scope and probably still has bugs, USE AT YOUR OWN RISK.

I will continue to add support for more types as I need them but feel free to open issues with feature requests or better yet PRs with implementations.

• Free software: MIT

• Documentation: https://pydantic-panel.readthedocs.io.

Getting Started

Step 1 - Install

pip install pydantic-panel

Step 2 - Import pydantic_panel and add your models to layouts!

import pydantic
import panel as pn
import pydantic_panel

class SomeModel(pydantic.BaseModel):
    name: str
    value: float

widget = pn.panel(SomeModel)

layout = pn.Column(widget, widget.json)
layout.servable()

Now edit

Basic Usage

If you import pydantic_panel, it will register the widget automatically using the panel.BasePane.applies interface. After importing, calling panel.panel(model) will return a panel.CompositeWidget whos value is the model. When you change one of the sub-widget values, the new value is validated/coerced using the corresponding pydantic field and if it passes validation/coercion the new value is set on the model itself. By default this is a one-way sync, if the model field values are changed via code, it does not sync the widgets. If you want biderectional sync, you can pass bidirectional = True to the widget constructor, this will patch the model to sync changes to the widgets but this may break without warning if pydantic change the internals of their __setattr__ method.

import panel as pn
import pydantic_panel

class SomeModel(pydantic.BaseModel):
    name: str
    value: float

# when passing a model class,
# all widget values will be None including the composite widget value
w = pn.panel(SomeModel)

# if you pass a model instance
# widget values will be the same as the model instance
inst = SomeModel(name='meaning', value=42)
w = pn.panel(inst)

# This will display widgets to e.g. edit the model in a notebook
w

# This will return True
inst is w.value

# This will be None if the widgets have not yet been set to values
# if all the required fields have been set, this will be an instance of SomeModel
# with the validated attribute values from the widgets
w.value

The pn.panel method will return a widget which can be used as part of a larger application or as just a user friendly way to edit your model data in the notebook.

Customizing widgets

You can add or change the widgets used for a given type by hooking into the dispatch mechanism (we use plum-dispatch). This can be used to override the widget used for a supported type or to add supprt for a new type.

from pydantic_panel import get_widget
from pydantic import FieldInfo

# precedence = 1 will ensure this function will be called
# instead of the default which has precedence = 0
@get_widget.dispatch(precedence=1)
def get_widget(value: MY_TYPE, field: FieldInfo, **kwargs):
    # extract relavent info from the pydantic field info here.

    # return your favorite widget
    return MY_FAVORITE_WIDGET(value=value, **kwargs)

Supporting non-serializable types

Panel encodes the data sent to the widget using json serialization. If your type is not json serializable, you can have pydantic-panel convert the data to a json-serializable object before its passed to the widget. To add this conversion, register a conversion function using the json_serializable.dispatch decorator

from pydantic_panel import json_serializable

# precedence = 1 will ensure this function will be called
# instead of the default which has precedence = 0
@json_serializable.dispatch(precedence=1)
def json_serializable(value: TYPE):
    # convert to a serializable object
    value = some_function(value)
    return value

Supported types

• int

• float

• str

• list

• tuple

• dict

• datetime.datetime

• BaseModel

• List[BaseModel]

• pandas.Interval

• numpy.ndarray

FAQ

Q: Why did you decide to use CompositWidget instead of Pane like Param uses?

A: Nested models. This is a recursive problem, so I was looking for a recursive solution. By using a Widget to display models, all fields are treated equally. A field of type BaseModel is edited with a widget that has a .value attribute just like any other field and therefore requires no special treatment. When the parent collects the values of its children it just reads the widget.value attribute and does not need to check whether the value is nested or not. At every level of the recursion the widget only has to care about the fields on its model class and watch only the .value attribute of its children widgets for changes.

Features

• TODO

Credits

This package was created with Cookiecutter and the briggySmalls/cookiecutter-pypackage project template.