[Zope3-dev] Re: Merging schema & interfaces (was: "Sub interface")

[Jim Fulton]

Casey Duncan wrote:
> On Tue, 02 Mar 2004 10:29:56 -0500
> Gary Poster <gary at modernsongs.com> wrote:
> 
> 
>>
>>Jim Fulton wrote:
>>
>>
>>>To deal with constraints on multiple fields, we will add the ability
>>>to model interface/schema-level invariants/constraints.
>>>
>>>def cityStateZip(contact):
>>>    pc = lookupPostalCodeFor(contact.city, contact.state)
>>>    if pc != contact.postal_code:
>>>       raise InvalidPostalCode(contact)
>>>
>>>class IContact(zope.interface.Interface):
>>>
>>>   postal_code = ...
>>>   city = ...
>>>   state = ...
>>>
>>>   invariant(cityStateZip)
>>>
>>>Then, when we modify a contact, we can check individual
>>>field constraints *and* we can check the interface-level invariants.
>>
>>We have that now, yes?  We're using it.  The guts are in Zope 3, from 
>>when Jim and I pair programmed on it quickly a few weeks ago.
>>
>>Zope 3 needs a pattern on how to use the output in forms, and we need
>>to figure out how to spell the invariants and possibly build a library
>>of these.
>>
>>We have a first stab on how to do this.  One of our interfaces has
>>
>>     invariant(isGreater(
>>         "expiration_date", "publication_date",
>>         _("${field1} must be later than ${field2}")))
> 

One of the reasons for this is that we've come to the conclusion that
we should raise exceptions when invariants fail. This allows us to
use exception views when they fail.

> I seems to me like this would generate an excessive amount of trivial
> code. I'd be inclined to use a lambda in simple cases and a function in
> more complex ones as above::
> 
>   invariant(lambda thing: thing.expiration_date >
> thing.publications_date,
>             'Expiration date must be later than the publication date')

So would this raise some sort of exception using the second argument
if the first arguments returns false?  I think this could work well.

I too prefer the lambda in a case like this.

> More complex invariants could be written like Jim's example. Something I
> find a bit strange about that though is that "cityStateZip" becomes like
> an instance method defined as a function assigned to the interface. Its
> a bit funky conceptually, it certainly stretches the notion of what an
> interface is.

Agreed. A lambda is much more readable, IMO.

We'd need to decide on how we'll get exceptions raised in cases like this.

Jim

-- 
Jim Fulton           mailto:jim at zope.com       Python Powered!
CTO                  (540) 361-1714            http://www.python.org
Zope Corporation     http://www.zope.com       http://www.zope.org