I had a possibly analogous problem writing something in R which does not
have comparison tolerance.  So, instead of a simple comparison like this

maxSect<- seclim <= sectorSums;

I had to do something like this

maxSect<- 1e-08<seclim-sectorSums;     # R lacks tolerance

However, I've been unable to uncover something like your problem in J after
running several million tests based on different guesses about how the two
vectors might be related.

On Thu, Jul 2, 2015 at 11:34 AM, John Baker <[email protected]> wrote:

> This has been a rare intermittent error. I am beginning to suspect
> something of a a non J'ey nature is going on.  If I can ever collect a
> datastream that forces the issue I will have it bronzed.
>
> On Wed, Jul 1, 2015 at 9:04 PM, Devon McCormick <[email protected]>
> wrote:
>
> > I ran a test for a few million lookups using i. on floating point numbers
> > and found no discrepancy but I created "sl" by a random selection from
> > "fl", so this may not properly replicate your issue.
> >
> > This was in the new J v. 8.04.
> >
> > On Wed, Jul 1, 2015 at 3:59 PM, <[email protected]> wrote:
> >
> > > Can you give an example, exactly?
> > >
> > > I think this would fail only if some sl is not found in fl.
> > >
> > > Henry Rich
> > >
> > > ---- John Baker <[email protected]> wrote:
> > > > I have observed that this sometimes fails
> > > >
> > > > fl=.  78.9999 .....  some floating list
> > > > sl=. 78.9999 .... some subset of (fl)
> > > >
> > > > (666.666 #~ #sl) (fl i.  sl) } fl  NB. searching floats
> > > >
> > > > where this always works
> > > >
> > > > (666.666 #~ #sl) ((<.fl) i. <. sl) } <. fl  NB. searching integers
> > > >
> > > > When I came across this I thought this could be fixed by comparison
> > > > tolerance.
> > > >
> > > > (666.666 #~ #sl) (fl (i. !.0) sl) } fl
> > > >
> > > > but no luck.
> > > >
> > > > Either this is a bug in J or I am not properly setting comparison
> > > tolerance
> > > > to 0.
> > > >
> > > >
> > > > Floating point - one of life's more intractable necessary evils.
> > > >
> > > > --
> > > > John D. Baker
> > > > [email protected]
> > > >
> ----------------------------------------------------------------------
> > > > For information about J forums see
> http://www.jsoftware.com/forums.htm
> > >
> > > ----------------------------------------------------------------------
> > > For information about J forums see http://www.jsoftware.com/forums.htm
> > >
> >
> >
> >
> > --
> > Devon McCormick, CFA
> > ----------------------------------------------------------------------
> > For information about J forums see http://www.jsoftware.com/forums.htm
> >
>
>
>
> --
> John D. Baker
> [email protected]
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>



-- 
Devon McCormick, CFA
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to