[antlr-interest] wildcard in tree grammar

[Gavin Lambert]

At 11:47 27/11/2008, Sam Harwell wrote:
 >Note that when w.ChildIndex == 0 and its next sibling is EOA, it 

 >means the . is the child of a closure such as .* , since the 
other
 >possibility ^(.) is not allowed in a tree parser.

Why would "^(.)" not be allowed?  That should simply match any 
node with no children.  I don't think there's any other syntax 
that could express that particular condition (barring a semantic 
predicate of course), as the equivalent 3.0 construct (simply ".") 
now represents any node with any number of children.

I'm not sure it's an especially *useful* construct, but it ought 
to be supported for consistency at least.

In other words (in vaguely ANTLR-like syntax):

   wildtree : ANYNODE | ^(ANYNODE wildtree*) ;

   .          => wild
   ^(.)       => ANYNODE
   ^(. FOO)   => ^(ANYNODE FOO)
   ^(FOO .)   => ^(FOO wildtree)
   ^(FOO . .) => ^(FOO wildtree wildtree)
   ^(FOO .+)  => ^(FOO wildtree+)
   ^(FOO .*)  => ^(FOO wildtree*)

To put it another way, given a single node FOO, all of the 
following should be able to match it:
   FOO
   ^(FOO)
   ^(FOO BAR?)
   ^(FOO xyz*)
   .
   ^(.)
   ^(. BAR?)
   ^(. xyz*)