I agree with your analysis. I think the current behavior of $!.f makes
sense & is correct. I can understand why 1$'' is an error & that
1$!.''5 should be different.
I disagree only in whether the Dictionary describes the behavior. Let
us say the English could be argued either way, which means the text is
failing in its mission to define the language.
Henry Rich
On 4/23/2013 4:36 PM, Raul Miller wrote:
Oh! I had not paid attention to $!.n
But, basically, it's similar to take ({.) - it uses y for any cells
specified there, and uses fill after that. (I had not paid attention
and thought that fill was only for the empty y case)
That said, given this behavior, I'm still not seeing why there should
be anything special about empty y for $!.n - we use n to specify the
fill, should it really matter that y was empty and the entire thing is
fill?
Taking a step back, the unadorned x$y case says that y is the data and
x is the shape, and it's an error if we have a non-empty shape but we
have not specified any data. The x$!.n y case is different because we
have specified the data - that spec is split across n and y, and n has
a special case handling if it's empty instead of being an atom. In
other words, it seems to me that the motivation for the 1$'' error is
absent from the 1$!.'' '' case.
You clearly have a different perspective on this from me, and it's
entirely possible that I've overlooked something else important for
this situation, but "I still don't get it".
Thanks,
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
