0j1 is complex. Why not 1j0?
We try an experiment:
ea=. each
NB. Generate a series of complex integers in text:
({'1j'),ea":ea i:3
│1j_3│1j_2│1j_1│1j0│1j1│1j2│1j3│
NB. Convert the text to complex numbers:
".ea({'1j'),ea":ea i:3
│1j_3│1j_2│1j_1│1│1j1│1j2│1j3│
NB. Find the datatypes:
datatype ea".ea({'1j'),ea":ea i:3
│complex│complex│complex│integer│complex│complex│complex│
So the datatype can change in the vector
NB. On the other hand:
NB. Generate a series of mixed integer & floating point numbers:
5,0,10%i:3
5 0 _3.33333 _5 _10 _ 10 5 3.33333
NB. Find the datatypes:
datatype ea 5,0,10%i:3
│floating│floating│floating│floating│floating│floating│floating│floating│floating│
NB. Hmmm. seems like there should be some integers in the datatypes
NB. Take the floors (all integers)
>. 5,0,10%i:3
5 0 _3 _5 _10 _ 10 5 4
NB. Find the datatypes of the floored vector (should be all integers):
datatype ea >.5,0,10%i:3
│floating│floating│floating│floating│floating│floating│floating│floating│floating│
According to the first example, the datatypes should change if the types
change to integer in the vector.
If we take the floor, all the results should be integer.
Seems like J should be more regular and predictable in its treatment of
datatypes.
Skip Cave
On Fri, Nov 22, 2019 at 2:12 PM Raul Miller <[email protected]> wrote:
> I think it's partially a matter of implementation convenience.
>
> Note, also, that J currently does not support gaussian integers except
> for those which can be represented using ieee-754 64 bit binary
> floating point, and:
>
> datatype 100000000000000000000000x 1
> extended
> datatype 100000000000000000000000x 1j0
> |ill-formed number
>
> Thanks,
>
> --
> Raul
>
> On Fri, Nov 22, 2019 at 12:44 PM Skip Cave <[email protected]>
> wrote:
> >
> > So the question is still: why not simply always store 1j0 as a complex
> type
> > internally?
> > What is preventing the J interpreter J from detecting 1j0, and storing it
> > internally as a complex value?:
> >
> > datatype 1j0
> > complex
> >
> > ?????
> >
> > Skip Cave
> > Cave Consulting LLC
> >
> >
> > On Fri, Nov 22, 2019 at 11:36 AM Don Guinn <[email protected]> wrote:
> >
> > > datatype{.0 1j1
> > > complex
> > >
> > > On Fri, Nov 22, 2019, 10:29 AM Skip Cave <[email protected]>
> wrote:
> > >
> > > > datatype >.1j0 1j1
> > > >
> > > > complex
> > > >
> > > > datatype 1j0 1j0
> > > >
> > > > integer
> > > >
> > > >
> > > > Why not just always store 1j0 internally as a complex type? 1j1 gets
> > > stored
> > > > as compex.
> > > > Skip Cave
> > > > Cave Consulting LLC
> > > >
> > > >
> > > > On Fri, Nov 22, 2019 at 10:58 AM Raul Miller <[email protected]>
> > > > wrote:
> > > >
> > > > > (1 j.0) would be one way of writing a constant that produces a
> complex
> > > > > precision with imaginary part 0.
> > > > >
> > > > > Thanks,
> > > > >
> > > > > --
> > > > > Raul
> > > > >
> > > > > On Fri, Nov 22, 2019 at 5:20 AM Henry Rich <[email protected]>
> > > wrote:
> > > > > >
> > > > > > 1/ Because the decoration comes before the value, it was easier
> to
> > > put
> > > > > > the 0 at the beginning of the display. That's a weak argument,
> but
> > > > > > that's why I did it that way.
> > > > > >
> > > > > > 2/ I was just thinking about this last night. You're right, for
> > > > > > consistency complex values should be decorated. But the point
> of the
> > > > > > decoration is to have a linear form that faithfully reproduces
> the
> > > > > > value. 1j0 does not produce a complex value. Before the
> decoration,
> > > > we
> > > > > > need to agree on a way to write a constant that produces a
> complex
> > > > > > precision with imaginary part 0.
> > > > > >
> > > > > > 3/ I defer to Bill on this.
> > > > > >
> > > > > > Henry Rich
> > > > > >
> > > > > > On 11/22/2019 1:04 AM, Kirk Iverson wrote:
> > > > > > > A recent thread in the programming forum (00 strange?)
> inspired me
> > > to
> > > > > look at this new behaviour in 901.
> > > > > > >
> > > > > > > NB. Numeric
> > > > > > > datatype&.> 'b i f x c'=. 0 1;(0 1+0);(0 1%1);0
> 1x;0,1j1-0j1
> > > > > > > +-------+-------+--------+--------+-------+
> > > > > > > |boolean|integer|floating|extended|complex|
> > > > > > > +-------+-------+--------+--------+-------+
> > > > > > > b&+
> > > > > > > 0 1&+
> > > > > > > i&+
> > > > > > > 00 1&+ <== (1)
> > > > > > > f&+
> > > > > > > 0 1.&+
> > > > > > > x&+
> > > > > > > 0 1x&+
> > > > > > > c&+
> > > > > > > 0 1&+ <== (2)
> > > > > > >
> > > > > > >
> > > > > > > NB. Character
> > > > > > > d=. (0{a.);'abc'
> > > > > > > 'lctrl lascii'=. d
> > > > > > > 'uctrl uascii'=. u:&.> d
> > > > > > > 'Uctrl Uascii'=. 10 u:&.> d
> > > > > > > datatype&.>lctrl;lascii;uctrl;uascii;Uctrl;Uascii
> > > > > > > +-------+-------+-------+-------+--------+--------+
> > > > > > > |literal|literal|unicode|unicode|unicode4|unicode4|
> > > > > > > +-------+-------+-------+-------+--------+--------+
> > > > > > > lctrl&,
> > > > > > > (00{a.)&,
> > > > > > > lascii&,
> > > > > > > 'abc'&,
> > > > > > > uctrl&,
> > > > > > > (u: 00)&,
> > > > > > > uascii&,
> > > > > > > (u: 97 98 99)&, <== (3)
> > > > > > > Uctrl&,
> > > > > > > (10&u: 00)&,
> > > > > > > Uascii&,
> > > > > > > (10&u: 97 98 99)&, <== (3)
> > > > > > >
> > > > > > >
> > > > > > > 1/ Is there a reason that it is the first element of an integer
> > > array
> > > > > which has extra decoration, rather than the last element (as in the
> > > other
> > > > > cases)?
> > > > > > >
> > > > > > > 2/ Shouldn't this display as 0 1j0&+ ?
> > > > > > > Also, when interpreting constants in a line of J:
> > > > > > > 0 1 is boolean
> > > > > > > 00 1 is integer
> > > > > > > 0 1. is floating
> > > > > > > 0 1x is extended (as is 0 1r1)
> > > > > > > 0 1j0 is integer
> > > > > > > Shouldn't this last one be complex?
> > > > > > >
> > > > > > > 3/ Can the representations of uascii and Uascii make use of
> > > literals
> > > > > (like lascii does)?
> > > > > > >
> > > > > > > /K
> > > > > > >
> > > >
> ----------------------------------------------------------------------
> > > > > > > For information about J forums see
> > > > http://www.jsoftware.com/forums.htm
> > > > > >
> > > > > >
> > > ----------------------------------------------------------------------
> > > > > > For information about J forums see
> > > http://www.jsoftware.com/forums.htm
> > > > >
> ----------------------------------------------------------------------
> > > > > For information about J forums see
> http://www.jsoftware.com/forums.htm
> > > > >
> > > >
> ----------------------------------------------------------------------
> > > > For information about J forums see
> http://www.jsoftware.com/forums.htm
> > > >
> > > ----------------------------------------------------------------------
> > > For information about J forums see http://www.jsoftware.com/forums.htm
> > >
> > ----------------------------------------------------------------------
> > For information about J forums see http://www.jsoftware.com/forums.htm
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
