In other words, what is the behavior when a constraint is violated? Do we
simply say this is non conforming? That it is implementation-dependent,
or implementation-defined? That it returns an error? That it returns a
specific error?
Ideally we'd want to be specific about what the constraint is and describe
these uniformly. I imagine we have domain constraints, data type
constraints, range constraints, ordering constaints, syntactical
constraints, etc. Ideally we'd have a clear taxonomy of these types of
constraints, apply them clearly, and map violation of each of these
constraints to a unique error class: domain error, data type error, range
error, etc.
However, we're a bit hampered by the restricted and illogical error codes
used by legacy spreadsheets. For example in Excel #NUM! conflates
entirely different errors types including numeric overflow like in
EXP(1000) or violation or ordering constraints like RANDBETWEEN(5,0).
Other spreadsheets do equally strange things.
Of course, it is possible to be more precise in OpenFormula, but if we did
we'd need to allow implementations to treat these error classes
interchangeably. This is unfortunate, but unless the vendors step up and
agree to rationalize their error codes, we have this legacy mess.
-Rob
"Andreas J. Guelzow" <> wrote on 01/06/2010
12:19:50 PM:
>
> Hi Rob,
>
> On Wed, 2010-01-06 at 12:04 -0500, wrote:
> > Is the objection over the reference to PI() rather than directly using
the
> > Greek letter Pi?
> >
> > Or is the choice of convention the problem?
> >
> > According to http://mathworld.wolfram.com/InverseCotangent.html there
are
> > two conventions: (0,Pi) or (-Pi/2, Pi/2). The former is continuous,
while
> > that later has a discontinuity at 0.
>
> No, my problem is much more pedestrian:
>
> If a constraint is violated the function should return an error. So how
> can a constraint talk about the function result?
>
> Suppose the return value is a number not in the range from 0 to PI(),
> then by the return value must be an error, since the constraint is
> violated. But if it is an error then it doesn't violate teh constriant
> anymore....
>
> I think that should not be a constraint.
>
> (We might want to add to the semantics that we are returning the
> principal value or the value derived from the branch through (1,PI()/4,
> but that wasn't my problem.)
>
> Andreas
>
> > "Andreas J. Guelzow" <> wrote on
01/06/2010
> > 11:15:44 AM:
> >
> > >
> > > ACOT has the constraint "The result must be between 0 and PI()."
This
> > > doe snot look like a valid constraint to me.
> > >
> > > ANdreas
> > > --
> > > Andreas J. Guelzow, PhD, FTICA
> > > Mathematical & Computing Sciences
> > > Concordia University College of Alberta
> > >
> >
> >
> > ---------------------------------------------------------------------
> > To unsubscribe from this mail list, you must leave the OASIS TC that
> > generates this mail. Follow this link to all your TCs in OASIS at:
> > https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php
> >
>
> --
> Andreas J. Guelzow, PhD, FTICA
> Mathematical & Computing Sciences
> Concordia University College of Alberta
>
>
> ---------------------------------------------------------------------
> To unsubscribe from this mail list, you must leave the OASIS TC that
> generates this mail. Follow this link to all your TCs in OASIS at:
> https://www.oasis-open.org/apps/org/workgroup/portal/my_workgroups.php
>
