-----Original Message-----
From: [EMAIL PROTECTED]
[mailto:[EMAIL PROTECTED] On Behalf Of Roger Hui
Sent: Tuesday, August 15, 2006 9:40 PM
To: Beta forum
Subject: Re: RE: [Jbeta] Linear Representation
Well, don't say I didn't warn you.
----- Original Message -----
From: Henry Rich <[EMAIL PROTECTED]>
Date: Tuesday, August 15, 2006 6:34 pm
Subject: RE: [Jbeta] Linear Representation
> With respect, I disagree. 3!:x changed for release 6, for
> example. As new datatypes are added, 5!:5 seems to me less
> likely to break. (This is not an academic issue: I have
> to keep J5.04 running to use it for DDE)
>
> I agree that to get floats right bit for bit 3!:n would
> be better.
>
> Henry Rich
>
> > -----Original Message-----
> > From: [EMAIL PROTECTED]
> > [EMAIL PROTECTED] On Behalf Of Roger Hui
> > Sent: Tuesday, August 15, 2006 9:26 PM
> > To: Beta forum
> > Subject: Re: [Jbeta] Linear Representation
> >
> > It is best to use 3!:x to exchange between machines
> > or to recover values in a session with perfect
> > fidelity.
> >
> >
> >
> > ----- Original Message -----
> > From: Henry Rich <[EMAIL PROTECTED]>
> > Date: Tuesday, August 15, 2006 6:16 pm
> > Subject: RE: RE: [Jbeta] Linear Representation
> >
> > > I know I encountered this problem years ago, when
> > > 0$0 was changed to '' in the 5!:5 result (IIRC). I
> > > believe the 0$0 that I was using was boxed and
> > > was part of a list of boxes.
> > >
> > > Roger fixed that, and my problems went away. I
> > > am glad to see that Roger expects to change the error
> > > reported here, because the 5!:5 form is a good portable
> > > way to pass values between machines and versions, and
> > > it is important to make sure that the values received
> > > are the same ones that were sent.
> > >
> > > Henry Rich
> > >
> > > > -----Original Message-----
> > > > From: [EMAIL PROTECTED]
> > > > [EMAIL PROTECTED] On Behalf Of Roger Hui
> > > > Sent: Tuesday, August 15, 2006 6:34 PM
> > > > To: Beta forum
> > > > Subject: Re: RE: [Jbeta] Linear Representation
> > > >
> > > > > By the way, a particular application failed to
> > > > > run because a J6 empty vector turned out not to
> > > > > be 'identical' to the corresponding J5 empty
> > > > > vector and it was difficult to see, via their
> > > > > linear representation, that they were actually
> > > > > different in some sense.
> > > >
> > > > There had been no changes in 5!:5 on empty vectors
> > > > or boxed empty vectors between J6 and J5.
> > > >
> > > >
> > > >
> > > > ----- Original Message -----
> > > > From: Jose Mario Quintana
> > > > <[EMAIL PROTECTED]>
> > > > Date: Tuesday, August 15, 2006 2:17 pm
> > > > Subject: RE: [Jbeta] Linear Representation
> > > >
> > > > > > Behalf Of Mark D. Niemiec
> > > > > > Jose Mario Quintana
> > <[EMAIL PROTECTED]> wrote:
> > > > > > > "5!:5 y Linear. The linear representation is a string
> > > > which, when
> > > > > > > interpreted, produces the named object."
> > > > > > >
> > > > > [...]
> > > > > >
> > > > > > One of the idiosyncracies of empty arrays is that they
> > > > > > are
> > > > > > considered equal, even if they have different data types.
> > > > > > This is rarely important, but the underlying data type
> > > > > > can
> > > > > > reveal itself when fill elements are being used (as with
> {.
> > > or {:)
> > > > > > and certainly with 3!:0
> > > > > >
> > > > > [...]
> > > > > > A problem with 5!:5 and 5!:6 is that they produce
> > > > > representations that are
> > > > > > equal to the original, but not necessarily
IDENTICAL in all
> > > > > > aspects. For example, they do not preserve type:
> > > > > >
> > > > > > type f0=:-~1.5 NB. real clone of 0
> > > > > > 8
> > > > > > f0
> > > > > > 0
> > > > > > type "[EMAIL PROTECTED]'f0' NB. This forgets the 'realness' of
the
> zero> > > > > 1
> > > > > > f0+!20x NB. real 0 trumps extended precision
> > > > > > 2.4329e18
> > > > > > ("[EMAIL PROTECTED]'f0')+!20x NB. freeze-dried and reconstituted 0
> > > does not
> > > > > > 2432902008176640000
> > > > > >
> > > > > > (and, in the cited example, lr does not preserve the
> type of
> > > the
> > > > > empty> list either).
> > > > >
> > > > >
> > > > > > Behalf Of Roger Hui
> > > > > > It comes down to whether there is one empty vector
> > > > > > or more than one empty vector. (*)
> > > > > [...]
> > > > > >
> > > > > > I will probably change 5!:5 to preserve the
> > > > >
> > > > > That seems to make more sense than the alternative of
> > > clarifying
> > > > > that "when
> > > > > interpreted, produces the named object" but it might not
> > > always be
> > > > > 'identically' the same object.
> > > > >
> > > > > By the way, a particular application failed to run because
> a
> > > J6
> > > > > empty vector
> > > > > turned out not to be 'identical' to the corresponding J5
> empty
> > > > > vector and it
> > > > > was difficult to see, via their linear representation,
> that
> > > > they were
> > > > > actually different in some sense.
> > > > >
> > > > > > distinction between <i.0 and <'' without answering
> > > > > > (*) one way or the other.
----------------------------------------------------------------------
For information about J forums see
http://www.jsoftware.com/forums.htm
