Introduction#

YAFOWIL targets are rendering form widgets and extracting/validating the data sent by the browser per widget.

YAFOWIL widgets are just configuration.

It provides a factory which can produce widget instances from blueprints.

There is a library of existing blueprints ready to be extended on demand.

YAFOWIL provides blueprints for all HTML standard inputs, lots of helper blueprints for building complex widgets and a bunch of add-ons (usually in the namespace yafowil.widget.*).

Motivation#

The YAFOWIL authors use several Python frameworks and became very tired of inventing widgets again and again.

YAFOWIL was intentionally written to be framework-independent. By just feeding it with configuration it can be used and extended in most existing python web frameworks. Possible candidates include Zope, Pyramid, Django, Flask, CherryPy and similar.

Another common problem with form libraries is a non-unique look and feel of the available widget collection. YAFOWIL tries to provide some useful addon widgets, thereby taking care of a unified user experience.

Dependencies#

YAFOWIL aims to have no dependencies on any web framework. It utilizes the node package which provides a powerful abstraction for lightweight graph data structures.

YAFOWIL does not know about data-storage, but offers a hook to add processing callback handler functions and a mechanism for delegating persistence automatically (to a certain degree).

Integrations#

YAFOWIL currently integrates with the following packages:

For details refer to the chapter Integrations.

Example#

For the impatient, code says more than 1000 words.

A simple example form works like so:

import yafowil.loader
from yafowil.base import factory
from yafowil.controller import Controller

Create a form:

form = factory(
    'form',
    name='myform',
    props={
        'action': 'http://www.domain.tld/someform',
    })

form['someinput'] = factory(
    'label:text',
    props={
        'label': 'Your Text',
    })

def formaction(widget, data):
    data.printtree()

def formnext(request):
    return 'http://www.domain.tld/result'

form['submit'] = factory(
    'submit',
    props={
        'handler': formaction,
        'next': formnext,
        'action': True,
    })

Render the empty form by calling the form object:

rendered = form()

This results in:

<form action="http://www.domain.tld/someform"
      enctype="multipart/form-data"
      id="form-myform"
      method="post">
    <label for="input-myform-someinput">Your Text</label>
    <input id="input-myform-someinput"
           name="myform.someinput"
           type="text"/>
    <input id="input-myform-submit"
           name="action.myform.submit"
           type="submit"
           value="submit" />
</form>

Process the form with a request. Requests are expected to be a dict-like mapping:

request = {
    'myform.someinput': 'Hello World',
    'action.myform.submit': 'submit'
}
controller = Controller(form, request)

The result of processing gets written to controller.data:

.. code-block:: python

controller.data

Creating a widget#

A widget is an instance of a blueprint created by the factory. Factory is a singleton and operates as a registry for blueprints.

By calling the factory, a widget is created.

Here, a naked text input field from the text blueprint:

widget = factory('text')

Blueprints can be chained by colon-separated names, or given as a list:

widget = factory('field:label:text')

This causes the created widget to chain the registered renderers, extractors, and other parts of the named blueprints (field, label and text) in order.

Blueprint chains can be organised using macros to reduce the complexity of factory calls (details below). E.g.:

widget = factory('#errorfield:text')

expands the macro #errorfield to field:label:error and appends :text so the result is field:label:error:text.

Widget trees#

YAFOWIL forms are organized as widget trees. The entire form is the root widget which contains compound nodes (which can contain children) and/or leaf nodes. A widget behaves similarly to an ordered Python dictionary. Compounds may represent the entire HTML form or fieldsets, while leaf objects may represent the various HTML input fields.

Building a widget tree:

form = factory(
    'form',
    name='formname',
    props={
        'action': 'someurl',
    })
form['somefield'] = factory(
    'field:label:text',
    props={
        'label': 'Some Field',
    })
form['somefieldset'] = factory(
    'fieldset',
    props={
        'legend': 'A Fieldset',
    })
form['somefieldset']['innerfield'] = factory(
    'field:label:text',
    props={
        'label': 'Inner Field',
    })
form['submit'] = factory(
    'submit',
    props={
        'handler': formaction,
        'next': formnext,
        'action': True,
    })

Rendering Mode#

The way a widget is rendered is controlled by its mode. Every widget may be given a mode keyword argument to the factory as a string or a callable accepting two parameters (widget and data), and returning a string.

These modes are supported:

edit

Default classic mode, editing of form is possible. Rendering follows the registered edit_renderers.

display

No form elements are rendered, just the data as defined by registered display_renderers.

skip

Renders an empty string.

Data extraction#

After calling the Controller we have the form processing result on controller.data, which is an instance of yafowil.base.RuntimeData.

Like widgets, runtime-data is organized as tree where each runtime data node refers to a widget node and provides the extracted value and any error(s) that occurred while extracting data from the request.

request = {
    'formname.somefield': 'Hello World',
    'action.formname.submit': 'submit'
}
controller = Controller(form, request)

data = controller.data

value = data.fetch('myform.someinput').extracted

Validation#

In YAFOWIL, validation and extraction happens at the same time. Extraction means to get a meaningful value out of the request. Validation means to check constraints, i.e if a number is positive or an email-address is valid.

If validation fails, ExtractionErrors are collected on runtime-data describing what happened.

Datatype extraction#

There is a set of common blueprints where you can define the datatype of the extracted value. Datatype is either some primitive type like int or float, a class object which can be instantiated with the extracted string value like uuid.UUID, or a callable expecting the extracted string value and converting it to whatever.

form['somefield'] = factory('field:label:text', props={
    'label': 'Some Field',
    'datatype': int
})

Blueprints which provide datatype by default are hidden, proxy, text, lines, select and number.

When providing a datatype to a widget which is not required, we probably want to have a valid emptyvalue, which takes effect if the request contains an empty string for this widget. The empty value must either be of or castable to the defined datatype or UNSET.

form['somefield'] = factory('field:label:text', props={
    'label': 'Some Field',
    'datatype': int,
    'emptyvalue': 0
})

Blueprints which provide emptyvalue by default are hidden, proxy, text, textarea, lines, select, file, password, email, url, search and number.

Invariants#

Invariants are implemented as extractors on compounds. Usually they are defined as a custom blueprint (see below) with one extractor on some parent of the elements to be validated.

Here is a short example (extension of the hello world example) of a custom invariant extractor which checks if one or the other field is filled, but never both or none (XOR):

from yafowil.base import ExtractionError
# ... see helloworld example whats missing here

def myinvariant_extractor(widget, data):
    if data['hello'].extracted == data['world'].extracted:
        error = ExtractionError(
            'provide hello or world, not both or none'
        )
        data['hello'].error.append(error)
        data['world'].error.append(error)
    return data.extracted

def application(environ, start_response):
    # ... see helloworld example for the code that belongs here
    form = factory(
        u'*myinvariant:form',
        name='helloworld',
        props={
            'action': url,
        },
        custom={
            'myinvariant': {
                'extractors': [myinvariant_extractor]
            }
        })
    form['hello'] = factory(
        'field:label:error:text',
        props={
            'label': 'Enter some text here',
        })
    form['world'] = factory(
        'field:label:error:text',
        props={
            'label': 'OR Enter some text here',
        })
    # ... see helloworld example for the code that belongs here

Persistence#

YAFOWIL provides a delegation mechanism for single data model bound forms.

Processing the extracted form data often requires some additional computing and targets several persistent objects. In this case, we simply implement the submit action callback and do what’s necessary:

class Form(object):

    def __init__(self, model):
        self.model = model

    def __call__(self, request):
        controller = Controller(self.form, request)

    def save(self, widget, data):
        # HERE IS THE INTERESTING PART
        self.model.hello = data.fetch('myform.hello').extracted
        self.model.world = data.fetch('myform.world').extracted
        # ...
        transaction.commit()

    form = factory(
        'form',
        name='myform',
        props={
            'action': 'http://www.domain.tld/someform',
        })
    form['hello'] = factory(
        'field:label:error:text',
        props={
            'label': 'Enter hello text here',
        })
    form['world'] = factory(
        'field:label:error:text',
        props={
            'label': 'Enter world text here',
        })
    form['submit'] = factory(
        'submit',
        props={
            'handler': save,
            'action': True,
        })

form = Form(model)
form(request)
# ... should have form data persisted to model now

While fetching the value from data and assigning it to model seems quite reasonable, as long as forms are small, this may get annoying when writing more or complex forms. If forms refer to a single model, data.write can be used to delegate transferring the extracted data to the model.

from yafowil.persistence import attribute_writer

class Form(object):

    # ...

    def save(self, widget, data):
        # HERE IS THE INTERESTING PART
        data.write(self.model)
        transaction.commit()

    form = factory(
        'form',
        name='myform',
        props={
            'action': 'http://www.domain.tld/someform',
            'persist_writer': attribute_writer
        })
    # ...

form = Form(model)
form(request)

The most common way is to add the persist_writer property to the entire form. data.write will walk through the data tree and call attribute_writer with model, target and value arguments for each runtime-data node with the persist property set to True.

The persist property indicates widgets to be considered when data.write gets called and is given as part of the widget properties at factory time.

The persist property is True by default on hidden, proxy, text, textarea, lines, password, checkbox, select, email, url and number blueprints.

The model received in persisting callback is the model passed to data.write.

The target received in the persisting callback is an arbitrary Python object and defaults to the runtime-data name for any particular widget. The target can be customized by providing persist_target as part of the widget properties at factory time.

The value received in the persisting callback is the extracted value from runtime-data.

The writer callback can be customized for each widget via the persist_writer property.

data.write can be called with a recursive=False keyword argument. In that case, persistence only happens on the calling level.

When setting the persist property True on compound widgets, make sure its children get persist set to False explicitly if the used child factory blueprint is persistent by default.

If data.write gets called on runtime-data which contains extraction error(s), a RuntimeError is raised.

The following default writer callbacks are provided:

  • yafowil.persistence.attribute_writer

    Write value to target attribute on model.

  • yafowil.persistence.write_mapping_writer

    Write value to target write mapping key on model.

  • yafowil.persistence.node_attribute_writer

    Write value to target node.attrs key on model.

In conjunction with datatype and emptyvalue, this approach gives us a lot of convenience for persisting form data to single models.

Providing blueprints#

General behaviours (rendering, extracting, etc…) can be registered as blueprints in the factory:

factory.register(
    'myblueprint',
    extractors=[myvalidator],
    edit_renderers=[],
    display_renderers=[],
    preprocessors=[],
    builders=[])

and then used as regular blueprints when calling the factory:

widget = factory('field:label:myblueprint:text', props={
    'label': 'Inner Field',
})

Adding custom behaviour#

It’s possible to inject custom behaviour by marking a part of the blueprint chain with the asterisk * character. Behaviours are one or more of a:

extractor

extracts, validates and/or converts form-data from the request.

edit_renderer

build the markup for editing.

display_renderer

build the markup for display only.

builder

Generic hook called once at factory time of the widget. Here e.g. subwidgets can be created.

preprocessor

Generic hook to prepare runtime-data. Runs once per runtime-data instance before extractors or renderers are running.

def myvalidator(widget, data):
   # validate the data, raise ExtractionError if something's wrong
   if data.extracted != 'something:'
       raise ExtractionError("only 'something' is allowed as input.")
   return data.extracted

widget = factory(
    'field:label:*myvalidation:text',
    props={
        'label': 'Inner Field',
    },
    custom={
        'myvalidation': dict(extractor=[myvalidator]),
    })

Delivering resources#

YAFOWIL addon widgets are shipped with their required JavaScript and stylesheet resources. These resources are registered to the factory with additional information like delivery order and resources group.

To help the integrator delivering these resources through their chosen web framework, the helper object yafowil.resources.YafowilResources should be used.

The function configure_resource_directory should be overwritten on deriving class which is responsible to make the given physical resource directory somehow available to the web.

The object can be instantiated with js_skip and css_skip keyword arguments, which contain iterable resource group names to skip when calculating resources. This is useful if basic or dependent resources are already shipped in another way.

The following example shows how to integrate YAFOWIL resources in a pyramid application.

from pyramid.static import static_view
from yafowil.resources import YafowilResources
import mypackage.views

class Resources(YafowilResources):

   def __init__(self, js_skip=[], css_skip=[], config=None):
       self.config = config
       super(Resources, self).__init__(js_skip=js_skip, css_skip=css_skip)

   def configure_resource_directory(self, plugin_name, resourc_edir):
       # instantiate static view
       resources_view = static_view(resourc_edir, use_subpath=True)
       # attach resources view to package
       view_name = '%s_resources' % plugin_name.replace('.', '_')
       setattr(mypackage.views, view_name, resources_view)
       # register view via config
       view_path = 'mypackage.views.%s' % view_name
       resource_base = '++resource++%s' % plugin_name
       self.config.add_view(view_path, name=resource_base)
       return resource_base

def includeme(config):
    # resources object gets instantiated only once
    resources = Resources(config=config)

    # sorted JS resources URLs. Supposed to be rendered to HTML
    resources.js_resources

    # sorted CSS resources URLs. Supposed to be rendered to HTML
    resources.css_resources