DanBron/Temp/IsEvokeBroken

Discussion of the ~ bug report continued:

Discussion

• I'm not sure that this is a bug. I think this behavior fits what the dictionary says about resolution for verbs: verbs are resolved as names, and the name is used at evaluation time to determine the context for execution. Here's a definition of evokeU that does what I think you want:

  evokeU=:1 :'(5!:1<''u.'')5!:0'

  -- ["Raul Miller"] DateTime(2006-05-22T19:39:30Z)

• I considered that interpretation; see the last example on this page.

  
  Clearly, + 1 : '''u''~'' is producing 'u'~: a single name to be executed (contrast + 1 : ' $ ''u''~'). The problem is that the name is executed in global (immex) context, not local context. Where is that behavior is described by the Dictionary? Here are the quotes I believe relevant:

  1. First, verify that u is is a valid name:

        qNM        =.  2080 200  qdoj 'dict1'
        qNM #~ -. (+.~:/\) '()' e.~ djqNM=. CRLF-.~ djqNM
     Names begin with a letter and may continue with letters, underlines, and digits.
     
     
        NB.  Definition to test for "name"-ness
        ld         =.  ; 'ltr dgt' =. ;: 'abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ 0123456789'
        isName     =.  (ltr ,&:< ld) *./@:(;@:(e.&.>)~ {. ,&:< }.)

     and

        qVLD       =.   1019 148 qdoj 'dx004'
        LF (] #~ ] -.@:E.~ 2 # [) CR -.~  djqVLD
     4!:0 y  
     Name Class.  Class of (boxed) name:
     _2   invalid
     
        
        NB.  Definition to test for "valid name"-ness
        validName  =.  (_2 -.@-: 4!:0) @:  (;:  ^:  (0 -: L.)) 

     therefore

        NB.  u is a valid name
        + 1 : ' (validName *. isName) ''u'' ' 
     1

  2. Then support the expectation of its value:

        qARG       =.  (CRLF,'()') -.~ 3654 389 qdoj ':'  NB.  Moreover, it refers to the adverb's argument
        _15&(]\)&.;: qARG {~ <<< ; _2 <@:(+ i.)/\ 253 14  156 58  117 15  88 16  81 3  56 12  24 19
     The names x and y denote arguments. a conjunction may refer to its arguments using
     u and v ; likewise , an adverb may refer to its using u                           

  3. Finally, invoke the evoke rule:

        ]qEVK      =.  1071 73 qdoj '~'
     If m is a name, then 'm'~ is equivalent 
     to m

     .
  Given these quotes, I assert the Dictionary says

  •    + adverb define 
    'u'~
    )

  and

  •    + adverb define
     u
    )

  are equivalent, but the interpreter says they're not. Can you argue, from the text of the Dictionary, that this is not a bug?

  
  Obviously, to do so is merely an exercise. The behavior is not a an obstacle in real life, as your code aptly demonstrates (thanks for that; I have a similar utility in my scripts).

  
  On top of that, the deprecation of the special inflected names u.v.m.n.x.y. rendered it even more ambiguous. After all, what's the difference between 1 :'''u''~' and 1 : '''global_function''~'?

  
  On the other hand, consider 1 : ('local_function=.+/';'''local_function''~')...

  
  -- DanBron 2006-05-23 00:09:16

• As I understand the dictionary, the following lines executed in typical j session should always give the indicated result. Furthermore, the lines can be re-arranged and, as long as the final line is left as the final line, the result should be the same.
  {{{ a=:b

  • b=:c d_e_=:3 :'noun' noun_e_=:i.3 noun=:'abc' c=:d_e_ a 1

0 1 2}}}BRBRPerhaps you are arguing that local names should get special treatment in the dictionary, but I don't think you can validly argue that the dictionary currently grants them the behavior you've implied, above.BRBR-- ["Raul Miller"] DateTime(2006-05-23T05:23:07Z)

• I understood your point. What you are arguing is that the behavior is correct in the sense of:

     + adverb define
   q =. +/
   q
  )
  |value error: q

  But what I am arguing is that the behavior is buggy in the sense of:

     + adverb define
   u =. +/
   u
  )
  +/

  Giving local names special treatment would eradicate the inconsistency demonstrated above (WRT q vs u). That might be a useful feature, but indeed a new one. That is, it would require a change to the Dictionary. I am not arguing otherwise.

  
  My argument is that J should be consistent about giving special names special treatment. Those special names are the are the pre-assigned argument referents, which previously were inflected.

  
  In short: I just want the sentence "If m is a name, then 'm'~ is equivalent to m" to be true.

  
  -- DanBron 2006-05-23 15:25:10

• This, I think, is a different issue -- not related to evoke.

  
  What you are talking about here is that (for reasons that I don't think are documented in the dictionary) some local variables get treated differently from others.

  
  That said, I see what you are saying -- I just don't think it's the right thing for locative names. Consider x__y as an example.

  
  -- Raul Miller 2006-05-23 20:18:16

• I don't understand that last paragraph. Where did locatives come in? Now that you've brought them up, though, wouldn't you prefer a different result for the last two sentences here?

     x_z_ =: 42
  
     3 : ' x__y ,&< ''x__y''~ ' {.;:'z'
  ┌──┬──┐
  │42│42│
  └──┴──┘
  
     'abc' 4 : ' x__y ,&< ''x__y''~ ' {.;:'z'
  ┌──┬──┐
  │42│42│
  └──┴──┘
  
     'abc' 2 : ' x__y ,&< ''x__y''~ ' ({.;:'z')
  ┌──┬──┐
  │42│42│
  └──┴──┘
  
     x_z_ =: +/
  
     'abc' 2 : ' x__y ,&< ''x__y''~ ' ({.;:'z')
  x__y ,&< x__y
  
     'abc' 2 : ' x__n ,&< ''x__n'~ ' ({.;:'z')
  x__n ,&< x__n

  -- DanBron DateTime(2006-05-26T16:14:45Z)