[Zope3-Users] Re: Annoying thing about schema

[Philipp von Weitershausen]

On 5 Jun 2008, at 11:53 , Peter Bengtsson wrote:
> 2008/6/5 Philipp von Weitershausen <philipp at weitershausen.de>:
>> Peter Bengtsson wrote:
>>>
>>> Suppose you've written this:
>>>
>>> class ISomething(zope.interface.Interface):
>>>    foo = zope.schema.TextLine(title="Foo")
>>>
>>> then you get the error below.
>>> First of all, why can't it just help me and just convert the str  
>>> to a
>>> unicode. All it'd need to do is just::
>>>
>>> title = unicode(title)
>>>
>>> Django does this.
>>> If someone does something bad like `TextLine(title="Ölmage")` that
>>> developer should rightly get a UnicodeDecodeError.
>>> Secondly, if there is an actual reason for this extreme analism, why
>>> can't the error be wrapped into something more user friendly?
>>
>> Very good points. I think UnicodeDecodeErrors have a bad  
>> reputation, though
>> (mostly because of Python's stupid implicit conversion). I think  
>> Grok's
>> policy is very sane in this regard: wherever human-readable text  
>> has to be
>> entered, either a unicode object or a string just containing ASCII
>> characters are acceptable. A very explicit error message stating  
>> that exact
>> policy should be raised. Look at Grok's error messages, we've tried  
>> hard
>> making them understandable.
>>
> This was actually Grok development. But the interface and schema part
> is all zope3.
> I should have posted to zope3-dev instead.

zope-dev (my mistake to suggest zope3-dev, it's been dead for a while  
now)

>> I think a proposal on zope3-dev (plus a patch) would be welcomed.
>
> I'm not sure how to write the proposal. Can't I just make a feature
> request on launchpad or something?

An email to zope-dev would be the best. Writing a proposal is easy, it  
can be done in a couple of paragraphs:

1. State the current situtation
2. Identify the problem that you have with the current situation
3. Suggest a solution
4. Identify potential risks


P.S.: As you may have noticed, zope.schema fields use fields  
themselves to define the types of their attributes/parameters. For  
instance, the 'title' and 'description' arguments are  
zope.schema.TextLines. So, in order to make those parameters accept  
both unicode and pure ASCII, there are two choices:

a) change zope.schema.Text and TextLine to accept both unicode and str  
objects, as long as the 'str' object contained just ASCII (-> validate  
and _validate methods)

b) introduce another schema field that has the aforementioned  
properties, thus leaving Text and TextLine unchanged.

I personally prefer a), but there might be good reasons to do b).  
Definitely something worth discussing in your proposal :).