Altering self.fields in Form.__init__() fails

[Ronny Pfannschmidt]

take a look at the test

"""
>>> TestForm({'text':"foo"}).is_valid()
True

>>> TestForm({'text':"foo"},need_name=True).is_valid()
False

>>> TestForm({'text':"foo"}).is_valid() #this one will fail
True

"""
from django.newforms import *
class TestForm(Form):
    name = CharField(max_length=30)
    text = CharField(widget=Textarea)
    def __init__(self,*k,**kw):
        need_name=kw.pop("need_name",False)
        Form.__init__(self,*k,**kw)

        if need_name:
            self.fields["name"].required=True

Just assigning need_name to self.fields["name"].required would create a race condition for multithreaded servers.

[Adrian Holovaty]

Fixed formatting in description.

[Adrian Holovaty]

(In [4504]) Fixed #3489 -- Changed newforms to use copy.copy() in constructing self.fields, so changes to self.fields in a given form instance do not affect other instances

[Jonathan Buchanan <jonathan.buchanan@…>]

Should copy.deepcopy() be used instead? I ran into problems when modifying widgets, as each field's widget will still be a reference rather than a copy.

from django import newforms as forms

class UserForm(forms.Form):
    username = forms.CharField(max_length=30)

    def __init__(self, disable_name=False, *args, **kwargs):
        super(UserForm, self).__init__(*args, **kwargs)
        if disable_name:
            self.fields['username'].widget.attrs['disabled'] = 'disabled'
        
>>> f1 = UserForm()
>>> print f1['username']
<input id="id_username" type="text" name="username" maxlength="30" />
>>> f2 = UserForm(disable_name=True)
>>> print f2['username']
<input disabled="disabled" id="id_username" type="text" name="username" maxlength="30" />
>>> print f1['username']
<input disabled="disabled" id="id_username" type="text" name="username" maxlength="30" />

[Jonathan Buchanan <jonathan.buchanan@…>]

Copying of fields with deepcopy and regression test

[Jonathan Buchanan]

Reassigning to myself since I've already written a patch and regression tests - will check if it still applies cleanly.

[Jonathan Buchanan]

Previous patch no longer works due to a deepcopy error

[Jonathan Buchanan]

The deepcopy implementation I've just attached a patch for seems like a bit of a hack to me, but the tests pass - can I get a second opinion?

[Malcolm Tredinnick]

I don't think this patch is hackish at all. It's exactly what is required when you write a deepcopy method: copy the bits you need and that's it. This is fine.

[Malcolm Tredinnick]

(In [6156]) Fixed #3489 -- Added proper deepcopying to form fields so that widget instances get copied as well. Patch from Jonathan Buchanan and insin.