User:Dan Bron/Temp/IsEvokeBroken

From J Wiki
Jump to: navigation, search

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)


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: A. 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
 A. 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
 A.  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''~')...


-- Dan Bron <<DateTime(2006-05-23T00:09:16Z)>>


  • 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
}}}[[BR]][[BR]]Perhaps 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.[[BR]][[BR]]-- ["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.


-- Dan Bron <<DateTime(2006-05-23T15:25:10Z)>>

  • 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 <<DateTime(2006-05-23T20:18:16Z)>>

  • 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
}}}  -- [[User:Dan Bron|Dan Bron]] [[DateTime(2006-05-26T16:14:45Z)]]