Well, my little lambda-like experimental language is getting rather large. I'm really having fun with it and am close to actually doing something almost useful with it. :) Unfortunately I've run into a problem I can't seem to isolate to a small testcase. I'm doing something like this: proto::display_expr( list [ obj["name1"].attr["a"], obj["name2"].attr["a"], obj["name3"].attr["b"] ] ); Presume I have all of the terminals, members, etc. declared/defined correctly (at least "correctly" as I understand it). display_expr dumps out an expression without any comma operators. It just shows a subscript off "list" with one "index," obj["name3"].attr["b"]. It's the behavior I would expect if the operands of the commas were not protoized expressions, but they _should_ be protoized. The funny thing is, if I write a small testcase just isolating the terminals, members, etc. needed to construct the expression, it works as expected. But in the context of the (much) larger language, it fails. However, the grammar specification shouldn't have any bearing on what display_expr dumps out. So any ideas as to why commas might be gobbled up prematurely? -Dave