Sorry, not
“_ is an identity to both (= <:) and (= >:)“
but
“ _ and __ are identities to bot <: and >:”
or
“_ and __ are the only values satisfying both (= <:) and (= >:)”
… except for NaN (_.), but that’s not exactly the topic here I think.
Am 02.08.20 um 12:52 schrieb Hauke Rehr:
I was not surprised by the results.
What concept of _ do you have in ^:_ if not one of an “integer?”
Furthermore, (<: <: <.) *. (>: >: >.) is true for any numeric value.
I think it’s obvious that _ is an identity to both (= <:) and (= >:)
– and so both <. and >. must return _ as well (likewise for __).
I am not at all surprised and would not want it to be defined diffently.
finiteinteger =: ((= <.) *. (~: <:))
keepfinint =: (#~ finiteinteger)
Am 02.08.20 um 12:29 schrieb Raul Miller:
The floor of infinity being infinity is not the real problem, opinion.
Or at least not the only problem. And, integer infinity is not a
particularly new concept: https://simple.wikipedia.org/wiki/Aleph_null
Infinity is defined as larger than any number, and larger is not
equal. Or, these sorts of "numeric quality" things can defy logic
because they are not specific values.
But infinity is not the only example of a problem with that
expression. Consider this:
(= <.) 0.5+2^128
1
Basically, floating point notation is not capable of representing
large fractional values.
Meanwhile, you could claim that J's implementation of infinity is an
integer infinity (since all values greater than a limit in floating
point notation are integer values).
Anyways, with that out of the way -- what is it that you're trying to
do? And, why is infinity a problem there? And, is this an issue for
you?
datatype <.2^10
integer
datatype <.2^100
floating
Thanks,
--
----------------------
mail written using NEO
neo-layout.org
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
