Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!shadooby!accuvax.nwu.edu!tank!mimsy!chris
From: chris@mimsy.UUCP (Chris Torek)
Newsgroups: comp.lang.c
Subject: Re: Using data types before they are declared.
Message-ID: <18137@mimsy.UUCP>
Date: 19 Jun 89 06:54:14 GMT
References: <10268@watcgl.waterloo.edu> <548@corona.pb>
Organization: U of Maryland, Dept. of Computer Science, Coll. Pk., MD 20742
Lines: 76

In article <548@corona.pb> db2@pbinfo.UUCP (Vorl. Informationssysteme)
suggests trying the following code:
>union Fred {
>	int		Number;
>	struct _el_str	*Node;
>};
>
>struct _el_str {
>	int	Info;
>	Fred	Name;
>};
>typedef	struct _el_str	*element;
>
>I am practising this in my own programs very often.

I doubt it: as written, it will not compile.  (It will in C++, however.)

The best approach is to ignore `typedef' completely, give all
structures and unions tags, and add typedefs afterward if you
wish.  Thus:

	union Fred {
		int	Number;
		struct	_el_str *Node;
	};

	struct _el_str {
		int	Info;
		union	Fred Name;
	};

(The difference between this and the first 9 lines of code quoted
above is the addition of one `union' keyword.)  Then, if necessary:

	typedef union Fred {
		int	Number;
		struct	_el_str *Node;	/* no synonyms available */
	} Fred;
	/* `Fred' is now a synonym for `union Fred' */

	typedef struct _el_str {
		int	Info;
		union	Fred Name;	/* this can now be abbreviated */
	} *element;
	/* `element' is now a synonym for `struct _el_str *' */

Other approaches are possible, but this one always works: Avoid the
abbreviation until it has been defined, and always use a structure
or union tag so that there *is* an unabbreviated form.  At worst, you
will invent some tags you never use.

Incidentally, I happen to like the practise of naming structure and
union elements in a `type-describing' manner:

	struct glumph {
		char	*g_name;	/* name of this glumph */
		int	g_mass;		/* mass of this glumph, in kg */
		short	g_halflife;	/* biodegredation halflife, years */
		short	g_cost;		/* price in cents per kg */
	};

If I then spot a line like

	p->g_mass

somewhere, I have a clue (the `g_' element prefix) that `p' is an
object of type `pointer to struct glumph'.

(This practise arose from pre-V7 C's approach of putting all structure
elements/offsets in a single global table, making unions unnecessary,
but forcing prefixes to guarantee uniqueness, and weakening type-checking.
The prefixes turned out to be convenient, and were included in new
structures in V7 and BSD.)
-- 
In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 7163)
Domain:	chris@mimsy.umd.edu	Path:	uunet!mimsy!chris