Wow. That was amazingly easy to find and change though I do not yet know
what else this breaks. I changed "1e_7" to "1r10000000" in "derivsecant"
and now can do this:
New1=: 1 : 'y-(u y)%0 u sslope_jcalculus_ 1]y'
0j50": (_2+*:) New1^:8 ] 1x
1.41421356237309504880168872420969807856967448493513
0j50": (_2+*:) New1^:10 ] 1x
1.41421356237309504880168872420969807856967187537695
The latter matches the last entry on the "Newton's Method" page.
On Sat, Jan 16, 2021 at 4:18 PM Henry Rich <[email protected]> wrote:
> I guess that something in sslope_jcalculus_ reverts to floating point.
> You could try to find & fix it.
>
> Henry Rich
>
> On 1/16/2021 3:51 PM, Devon McCormick wrote:
> > If this discussion is based on the briefly mentioned "f1 =: #.&1 _1
> _2&>",
> > that is admittedly a hack: it says so on that page. I've already changed
> > the page to use the polynomial version.
> >
> > Of greater concern is that once I update this page with what I've
> proposed,
> > the lower section of the "Newton's Method" page is not so easily remedied
> > as that demonstrates using the method with extended precision arguments
> to
> > calculate the square root of 2 to arbitrary precision. I've always liked
> > that example and would like to be able to retain that ability.
> >
> > On Sat, Jan 16, 2021 at 2:45 PM Hauke Rehr <[email protected]>
> wrote:
> >
> >> In this case, there is no well-defined fixed point,
> >> no way of talking about “the” fixed point.
> >> Any nonboxed value is a fixed point.
> >>
> >> What about “taking the derivative of ": ?”
> >> All strings are fixed points.
> >>
> >> The question raised here is “what is > meant to be?”
> >> Sure you can define it so that it wants to do what
> >> it actually does. But I always thought of it as
> >> unboxing, with the quirk that it doesn’t reject
> >> unboxed values but works like no-op ] instead.
> >> As a convenience.
> >>
> >> As always, YMMV
> >>
> >> Am 16.01.21 um 20:35 schrieb Justin Paston-Cooper:
> >>> The fixed point of a function is a well-defined concept.
> >>>
> >>> On Sat, 16 Jan 2021 at 22:21, Hauke Rehr <[email protected]>
> wrote:
> >>>> There’s no sane way of talking about a “derivative of >”
> >>>>
> >>>> You said it shouldn’t concern itself with functions
> >>>> meant for dealing with boxed arguments, which > is an
> >>>> example of. If you’re not willing to state your numeric
> >>>> function in terms of functions dealing with numeric
> >>>> arguments only, you should be blamed.
> >>>>
> >>>> There is ].
> >>>> This is not by design meant for boxed-only arguments.
> >>>>
> >>>>> 3 works only as a convenience. Semantically, it’s crap.
> >>>> I think it should be undefined behaviour officially.
> >>>> Open to be changed to produce an error without notice.
> >>>>
> >>>> Don’t misunderstand me: I like using &.> and the like.
> >>>> But I think it’s working against intended semantics
> >>>> and always consider using > on unboxed arguments a hack.
> >>>>
> >>>> Am 16.01.21 um 20:12 schrieb Raul Miller:
> >>>>> >3
> >>>>> 3
> >>>>>
> >>>> --
> >>>> ----------------------
> >>>> mail written using NEO
> >>>> neo-layout.org
> >>>>
> >>>> ----------------------------------------------------------------------
> >>>> For information about J forums see
> http://www.jsoftware.com/forums.htm
> >>> ----------------------------------------------------------------------
> >>> For information about J forums see http://www.jsoftware.com/forums.htm
> >>>
> >> --
> >> ----------------------
> >> mail written using NEO
> >> neo-layout.org
> >>
> >> ----------------------------------------------------------------------
> >> For information about J forums see http://www.jsoftware.com/forums.htm
> >>
> >
>
>
> --
> This email has been checked for viruses by AVG.
> https://www.avg.com
>
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
--
Devon McCormick, CFA
Quantitative Consultant
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
