NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c++-part1/4565

Internet Message Format | 1996-08-06 | 1.9 KB

Path: sundog.tiac.net!smayo
From: [e-mail] (Scott Mayo)
Newsgroups: comp.lang.c++
Subject: So I have this pointer-to-member function...
Date: Tue, 30 Jan 1996 19:50:55 -0500
Organization: The Internet Access Company
Message-ID: <[e-mail]>
NNTP-Posting-Host: sunspot.tiac.net

I'm trying to do something in C++ which instinct tells me shouldn't be hard to
arrange, but the compiler isn't cooperating.

I have a class, call it C, whose purpose in life is to acquire an input
stream, parse it into tokens, and then have the tokens tell it which member
functions to call. For token "A" call member function A, and so on. An array
of struct {char *token; void (C::*f)();} fits the bill well enough to manage
that.

Now the tricky part. I want to derive class C1 and C2 from C. And I want C1
(and C2, etc) to have their own such table for mapping tokens to member
functions; in the absence of a match, I want to go up and search C's table and
use what I find there.

"Ah," I innocently thought, "I'll just make each array of struct a static
variable inside the class definition it goes with, and then it will be easy to
automatically search the right table at the right time."

The compiler will have none of it. It takes one glance at the static struct
{char*name; void (C::*f)();} list[] = {"A", C::A}; and has a hissy fit at the
attempt to assign initializers. It's fine outside the class definition; it
won't work inside.

I can clearly just put the arrays outside the class definitions, and just
teach the class "search" function to know which array to go after, but it
seems crass. That array is logically a part of the class behaviour, and if I
end up creating a lot of these subclasses, I don't want to rely on
remembering to tell each search function which magic array to use.

Suggestions? Am I missing something simple and useful here? Thanks. Mail
appreciated, posting might be missed.