[871] in java-interest
"new" revisited
daemon@ATHENA.MIT.EDU (Scott Hudson)
Sat Aug 12 03:25:54 1995
From: hudson@cc.gatech.edu (Scott Hudson)
To: java-interest@java.sun.com
Date: Wed, 9 Aug 1995 21:43:03 -0400 (EDT)
I don't seem to be able to find the original message, but I recall that
recently the precedence of the "new" operator was discussed and that, in
contrast to C++, apparently it is supposed to bind tighter than ".". As I
recall the particular example in question was something like:
"new foo.bar().blee.bas" which is supposed to be legal and parse as:
"(new foo.bar()).blee.bas".
Now if "new" has higher precedence than ".", and (per the spec) "." has
the same precedence as "[]", then this implies that "new" has higher
precedence than "[]". If this is the case, then I think we may have
another ambiguity. In particular, it seems like "new foo.bar[5][5][5]" by
the same reasoning as the example above should parse as:
"(new foo.bar[5])[5][5]" which can't be right since we could never allocate
multidimensional arrays.
Before saying something on the order of "[] is left associative", however,
note that it gets a bit trickier. Consider: "new foo.bar[5][][5]". This one
clearly must mean "(new foo.bar[5][])[5]". (In fact making associativity
arguments here is dubious since "[]" acts as two different "operators" here.
The first two uses of "[]" are a declarative "operation" which is distinct
from the third which is an array access operation.)
This seems to argue for putting "new" precedence back at the level it is in
C++. In that case both "new foo.bar().blee.bas" and "new foo.bar[5][][5]"
would be illegal and the language would force the user to use explicit
parenthesis (which results in more readable code, anyway).
Comments/discussion?
Scott
-
Note to Sun employees: this is an EXTERNAL mailing list!
Info: send 'help' to java-interest-request@java.sun.com
