NetNews Usenet Archive 1992 #16/NN_1992_16.iso/spool/comp/lang/c/11712

Internet Message Format | 1992-07-29 | 1.5 KB

Path: sparky!uunet!mcsun!sunic!news.funet.fi!hydra!klaava!wirzeniu
From: [e-mail] (Lars Wirzenius)
Newsgroups: comp.lang.c
Subject: Re: Variable length records
Message-ID: <[e-mail]>
Date: 29 Jul 92 16:06:18 GMT
References: <[e-mail]> <[e-mail]> <[e-mail]>
Organization: University of Helsinki
Lines: 22

In an earlier article I asked for what the advantages of using the
non-standard "struct { ...; float data[];}" or the tricky but more
standard "struct {...; float data[1];}" for structures where the data
array may be of varying lengths.  I offered the "struct {...; float *}"
version instead, as it is strictly conforming and will work everywhere.

The reasons for the trickery versions were convenience (no need for
managing a separately allocated array), less wasted space (no additional
pointer), and no need for a pointer dereference.  Your mileage may vary,
but I don't think these are very good reasons, at least under normal
circumstances.  If speed or size is critical, one's values may be
different.

One point which I didn't think of until now (and which doesn't seem to
have been pointed out by others) is that the '[]' and '[1]' versions
don't allow using the structures as values, like for assignment,
parameter passing and returning as function values.  This may or may not
be important, but it is something you have to check for, as it could
lead to difficult bugs, as such usage doesn't look wrong in itself.

--
[e-mail]