Don Guinn wrote:
>  Read your backtick and it looks good. Easy reading. But I still like the
>  original in the dictionary better.

Agree vehemently here.  But the problem is by this point I'm pickled in the
Dictionary.  The question is: what is the best way to help someone who's not
pickled in the Dictionary understand (`)  ?

>  While looking at the original dictionary I noticed 
>  a couple of things that the dictionary said that 
>  were different from what you had. 

And this is the problem.  If you bring all the context to the Vocab page for
`  that the DoJ expects, I don't think you'd find a discrepancy.  What I
tried to do in that link was to weave all that context into a single
narrative.  Well, a lot of it, anyway.  I tend to agree with Roger that it
is an impossible task to "assume nothing" about our audience, as our Vocab
entries must balance completeness with "referenceability" (i.e. brevity).  

>  Also shows that "u`n" is valid. Well, the example "*`9" does
>  give an error as you said.

Here's an example.  You are absolutely right: the DoJ is unequivocal that
u`n is valid.  It is also unequivocal about what u`n means: (ar of u), n.
The issue is you must know the definition (or limitations) of (,) to predict
the result of u`n .  In this case, (ar of u) is always boxed, whereas 9 is
always numeric.  Boxes and numbers are non-conformable; you may not append
them.  If you try to, then (,) will balk & produce domain error, as you
noticed.  Worse still, from J's point of view, an array with a box and a
number would be heterogeneous, and J supports only homogeneous arrays; so it
is not really  `  or even  ,  which is balking, and you will not find this
potential complaint documented in either official Vocab entry.

It is exactly this kind of "conceptual gap" we must try to bridge in the new
Vocabulary.  Note how even you, a J veteran, we thrown by the existing
Vocabulary's (non-) treatment of this issue, were able to grasp the
situation more clearly with a little embedded exposition.

>  The dictionary says that the name "add" would be 
>  converted to "+" (5!:1).

Similar remarks apply here.  You must understand J's parsing & stacking
rules to know why this wouldn't happen.  I agree that without all that
context, in the sandbox of the Vocab entry for  ` (,) the assertion that
atomic representation -: 5!:1 is "clear". 

That is the problem.

-Dan

PS:  re 
>  In the definition part you have a sentence ending with "as with ,.". 
>  The problem is that the comma is immediately followed by a period

In my browser, the distinction is clear; this is because I use a custom
stylesheet (CSS) to distinguish "code" from "prose".  The solution is to
make the default stylesheet more J-aware (or better, mark all J code
explicitly, e.g. with a custom parser.  Then make a very focused stylesheet
that has syntax highlighting). 

Of course, this solution won't work for (ASCII) email.  In this medium, I
use a mixture of extra spacing (e.g. blah blah blah  a.i.'j code'  blah
blah) and Tracy's convention of parenthesizing J code.  

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

Reply via email to