Bump?

Sorry, I didn't have time to look at it yet, will try to do it a.s.a.p. (but ideally it should really be @vslavik who should look at it, I still don't know all this code that well myself...).

Where does !! syntax come from BTW?

Bump?

Really, after only 23 hours?

I apologize for the removal of indentation.

You can always discard or unstage parts of the diff with git… But removing trailing whitespace that shouldn't be there in the first place and is in the code you touch is perfectly fine and appreciated.

At a glance, this code still won't work with references lists (or other values) that have conditionally set values.

Basically, it's equivalent to #34 as far as I can tell, but expressed a bit differently. My point was that references within interpolated strings need to be handled differently from "outside" references and their concatenations. This patch still treats them the same (case in point: there's no difference between ReferenceAsStringExpr and ReferenceExpr; to_string() seems to be (mostly?) identical to as_py() as well.

Ack...I thought ConcatExpr was only used when strings were interpolated. I'm checking to see where it actually occurs now.

@vslavik And...I'm lost. Could you point me to where variables get interpolated? I've grepped for ReferenceExpr, 'def reference(', and get_value, all of which have returned nothing useful.

Could you point me to where variables get interpolated?

They aren't now, because the current grammar treats this case as concatenation of literals with ordinary variable reference. That's why I wrote in in #34 that it needs to be treated specially, with a dedicated Expr specialization that doesn't do just a lookup of a variable, but also converts it to a string (note: "string" in the sense of an Expr tree of StringType (i.e. can be concatenated with other literals), not in the sense of a Python string).

Ok...I think I've got something working. It rejects normal concatenation between different types but specializes string interpolation. The only problem now is that I think it set something off that's causing stuff like $(toolset) == gnu to goof up.