Re: Confused ⎕UCS

[Dr . Jürgen Sauermann]

    Hi Adam,
      
      thanks for sharing your observation. What you observed is that
      ⎕UCS in GNU APL is an
      extension of the original  ⎕UCS in IBM APL2.
      
      Formally speaking, an extension of a function F1 is a function F2
      which returns
      the same values as F1 for all arguments in the definition domain
      of F1, but
      other results in the definition domain of F2 (i.e. where F1 would
      raise a DOMAIN
      ERROR).
      
      The IBM ⎕UCS had some rather arbitrary limitations, which were
      removed in
      GNU APL's ⎕UCS:
      
      * In IBM ⎕UCS the argument had to be a simple array of integers or
      characters.
      This type of arrays were common in good old APL1, but is no longer
      adequate for
      APL2. GNU APL extended this limitation to allow mixed arrays (of
      characters
      and integers) as well.
      
      * In IBM APL2 the integers had to be positive. However, many such
      characters
      originate from some ⎕FIO function which could return negative
      integers (primarily
      in the range ¯128 ... ¯1) and requiring them to be converted to
      positive integers
      would be useless overhead that can simply be avoided to allow
      negative integers
      as well.
      
      Floating point and complex numbers are not allowed as to avoid
      interference
      with ⎕CT (i.e. how should rounding be performed?).
      
      IBM APL2 ⎕UCS is based on  ISO 10646 and not on Unicode. Therefore
      the "last Unicode" should not make a difference (and is subject to
      change over
      time anyway). In the past the set of Unicode code popints has
      grown over time
      (as new characters were added in different Unicode releases) while
      the encodings
      of these codepoints (e.g, UTF8) has decreased (from 0...0x7FFFFFFF
      in RFC 2279
      down to 0...0x10FFFF in RFC 3629). I therefore came to the
      conclusion that the
      range supported by GNU APL's ⎕UCS should be so large that it
      becomes immune to
      future changes in the various character sets around.
      
      Therefore the full picture of your last example is this:
      
          ⎕ucs 1114112
      ����
        
            ⎕ucs ⎕ucs 1114112
      1114112
      
      Best Regards,
      Jürgen

On 11/22/20 10:00 AM, Adám Brudzewsky wrote:

I don't know if this is one bug or several, but here you go:

      ⎕UCS 100
d
      ⎕UCS 1E2
DOMAIN ERROR+
      ⎕UCS 100
      ^

      ⎕ucs ¯1
�

      ⎕ucs ¯100032+⍳3
ABC

      'ABC'=⎕ucs ¯100032+⍳3
0 0 0

      ⎕ucs 1114112 ⍝ last Unicode char is 1114111
����