[Freeipa-devel] [PATCH] 990 attribute name in exceptions

[Martin Kosek]

On Mon, 2012-03-19 at 14:43 -0400, Rob Crittenden wrote:
> Petr Viktorin wrote:
> > On 03/19/2012 03:02 PM, Rob Crittenden wrote:
> >> Petr Viktorin wrote:
> >>> On 03/16/2012 10:23 PM, Rob Crittenden wrote:
> >>>> When using *attr we should return the param.name of in the exception
> >>>> and
> >>>> when using a cli option we should return param.cli_name. This didn't
> >>>> work consistently in the framework.
> >>>>
> >>>> This is a bit of a kludge, catching exceptions and re-raising them, but
> >>>> it is a less invasive way of doing it.
> >>>>
> >>>> I added some examples of things to test in the ticket.
> >>>>
> >>>> rob
> >>>>
> >>>
> >>> > (name, error) = err.strerror.split(':')
> >>> > raise errors.ConversionError(name=attr, error=error)
> >>>
> >>> AFAIU this will break when the error message is translated. Why not just
> >>> use err.kw['error']?
> >>>
> >>
> >> Because the attribute name needs to get pulled out of it.
> >>
> >> rob
> >
> > No, you're ignoring the attribute name.
> > Anyway, even the English error messages are 'invalid %(name)r:
> > %(error)s', so "name" part starts with 'invalid ', and the "error" part
> > starts with an extra space.
> 
> The error gets re-raised so there is no dup "invalid". I'll grant you 
> the extra space though.
> 
> What this does is catch something like 'invalid maxfail: something bad', 
> split out the attribute/error and re-raise use the attribute name we want.

I think this is what Petr wanted to point out - you can access the
attributes passed to our PublicErrors directly, i.e. you don't have to
parse it from its string representation:

>>> from ipalib import errors
>>> x = errors.ValidationError(name='foo', error=u'Invalid value!')
>>> print str(x)
invalid 'foo': Invalid value!
>>> x.name
'foo'
>>> x.error
u'Invalid value!'

Martin