The case for linear boxed representation:

linear representation is easier to manipulate than atomic.
atomic provides a way to interpret the difference between strings and functions.
linear boxed also provides this formalism, but with the ease of manipulation of 
linear.
linear boxed structures representations are in a 2 column table.
The type column is easily discarded to take the linear forms.
A single "tie" modifier can attach any form of speech to the representation 
accumulator. (though that could be developed for atomic too)
Its easier to process anonymously if you assume either a bias to functions or 
nouns (parse strings as assumed modifiers or assumed data).
Anonymous processing is a key to functionality.
The lack of complete tools for processing atomic representations is what 
motivated linear boxed rep, and the ease of making those tools.



----- Original Message -----
From: "[email protected]" <[email protected]>
To: [email protected]
Cc: 
Sent: Monday, March 9, 2015 11:17 PM
Subject: Re: [Jprogramming] linear boxed representation : A new representation 
format for mixed type words

Pascal Jasmin wrote:
PJ>
PJ> if you want to pass a gerund (or boxed noun) to a function, it must
PJ> either be all verbs (with single adverbs as strings permitted), or all
PJ> nouns.  So this is possible.
PJ>   '`a b c' =. '/';+`- 
PJ> but it fails for an adverb phrase.
PJ>   '`a b c' =. '/\';+`- 
PJ> |spelling error

You are missing out on the more general notion of Atomic Representations.

Gerunds are arrays of verb ARs.  ARs are uniquely defined for *any* syntactic
class, though, and you are free to mix them:

   a =. +
   b =. -
   c =. /\
   d =. 42

   (ar =. 5!:1) ;:'a b c d'
+-+-+---------+------+
|+|-|+-+-----+|+-+--+|
| | ||5|+-+-+|||0|42||
| | || ||/|\|||+-+--+|
| | || |+-+-+||      |
| | |+-+-----+|      |
+-+-+---------+------+

   '`aa bb cc dd' =. ar ;:'a b c d'

   aa
+
   bb
-
   cc
/ \
   dd
42

   nameclass ;:'aa bb cc dd'
3 3 1 0

   (aa cc ,: bb cc) 1 2 3
1  3 6
1 _1 2

Further reading:  http://jsoftware.com/help/dictionary/dx005.htm
for 5!:1 (and, if you like, its reverse 5!:0);  then re-read
http://jsoftware.com/help/dictionary/d001.htm  to re-check
how the back-tick in '`a b c' is defined.

Further further reading:
Bernecky, Robert, and R.K.W. Hui, Gerunds and Representations, APL91, ACM.

                            Martin

PS:  Apropos "reading":  I cannot read your (or anybody else's) message
parts with unicode box characters on the machine which receives the
newsified forum messages.  If you want me to be able to read those
message parts, please adhere to the Forum rules as stated in
<http://www.jsoftware.com/jwiki/System/Forums>:

    **Formats**

    The mailing lists send out emails in plain text format. For
    best results, please send messages to the forum in this

    format.

    When copying J session output to an email, first set
    box-drawing characters to Ascii. The IDEs have a menu item
    for this, e.g. in J8 Qt, select View|Toggle Ascii Box
    Drawing.

Having said that, and *not* having been able to read your code, just
guessing from your "Subject:  I'm afraid you are re-inventing a wheel.


----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to