> I think, for plotting purposes, that ]&.": should usually suffice.
That is nifty, Raul.
It took me all of 5 minutes to work out how that worked. And now I see
it lets me process any numeric array with a string manipulation
function that handles each numeric atom like a string, but returns the
array in numeric form. I have loads of uses for that.
Let me record this here, for the sake of other novices ...
]z=: 0.45+i.6
0.45 1.45 2.45 3.45 4.45 5.45
foo=: 3 : 'y-.''.''' NB. simply removes the decimal point
foo &.": z
45 145 245 345 445 545
> This defect seems to be a "fundamental feature" of plot. I cannot
> think how to remove it without redesigning plot.
I appreciate that plot needs to accept complex arrays, to plot
functions over the complex plane. And I had other things on my mind
than to get diverted into debugging plot. But maybe just it could
crash in glcmds a little more elegantly, without hiding the evidence?
(Maybe Bill's fix does that?)
I'm totally in accord with Bill when he says "equality of numbers
[should] not depend on internal storage representation". It's always
been the big selling point of the APL family of languages, at least
since 1968 when I first met APL on the 1100. When the received wisdom
in the DP world at large was that a number variable should be typed
for the range of data intended to fit in it. And it's significant that
it broke down for me not inside J code, but where J has to interface
with foreign code. So I wasn't really surprised when plot choked over
a pathological piece of data: a "complex" array that wasn't.
What surprised me (...perhaps it shouldn't?) is that a noun in J
carries around its provenance with it, and infects (perhaps "poisons"
is too emotive a term) any noun it's used to build. And I discover it
doesn't just happen with 'complex', but with 'boolean' too...
datatype b=: 1 0 1 1 0 1 0
boolean
]z=: {. 0 1 2.22
0
datatype z
floating
datatype b,z
floating
Which surely means that "datatype", or 3!:0, is at best only a guess
at the logical content of a numeric array? I shall be more alert to
that possibility in future, when using 3!:0.
On Wed, Jul 20, 2011 at 3:55 PM, Raul Miller <[email protected]> wrote:
> On Wed, Jul 20, 2011 at 10:36 AM, Ian Clark <[email protected]> wrote:
>> 1. Can anyone think of a neater way of cleaning up z than my verb:
>> real? I expect there's a primitive capable of filtering the "poison"
>> but I haven't found it yet.
>
> I think, for plotting purposes, that ]&.": should usually suffice.
>
>> 2. Can someone fix the plot package so that it either tolerates or
>> correctly reports "poisoned" arrays, and doesn't just fall over in a
>> puzzling way?
>
> This defect seems to be a "fundamental feature" of plot. I cannot
> think how to remove it without redesigning plot.
>
> --
> Raul
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
>
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
