Re: Wide_Text_IO: bad data in text file

[adam@irvine.com (Adam Beneschan)]

"Grein, Christoph" <christoph.grein@eurocopter.com> wrote in message news:<mailman.1035522182.23022.comp.lang.ada@ada.eu.org>...
> > The RM doesn't specifically address this issue, as far as I can tell. 
> > Data_Error seems to be intended for cases where characters can be read
> > from the file but they're not in the proper syntax for an integer
> > (when using an instance of Integer_IO), or a float, or whatever.  The
> > closest thing I could find was A.13(10), which says "Use_Error is
> > propagated if an operation is attempted that is not possible for
> > reasons that depend on characteristics of the external file".  When I
> > tried a test using GNAT, it raised Constraint_Error, which seems
> > counterintuitive.
> 
> Exactly when did Gnat raise Constraint_Error? At the Get? Or later when you
> use the value read?

The exception was raised by Get.

> My interpretation of the RM is that Data_Error should be raised, however it
> need not be raised if this would be too much of a burden (cannot be easily
> done). If Data_Error is not raised, you have invalid data.
> 
> The 'Valid attribute can be used to insure that the data read really is valid 
> (i.e. has a legal interpretation). This attribute is there for just this 
> purpose.

Sorry, this doesn't make sense to me.  Unless I'm missing something, I
don't think it's possible for Wide_Character'Value ever to return
False in GNAT or in  most other implementations.  This is because a
Wide_Character would be represented in memory by a 16-bit object, and
all 2**16 possible 16-bit values are valid representations for a
Wide_Character.  The issue here isn't whether an Ada object (which
would be in memory or registers) is valid but whether a bit sequence
in the *file* is valid.  (In UTF-8, e.g., it takes 8 to 24 bits to
represent a Wide_Character.)  I don't see how 'Valid can be used to
test that.

				-- Adam