NetNews Offline 2/NetNews Offline Volume 2.iso/news/comp/lang/c++-part1/3019

Internet Message Format | 1996-08-06 | 3.6 KB

Path: keats.ugrad.cs.ubc.ca!not-for-mail
From: [e-mail] (Kazimir Kylheku)
Newsgroups: comp.lang.ada,comp.lang.c++,comp.lang.c,comp.lang.modula3,comp.lang.modula2,comp.lang.eiffel
Subject: Re: Hungarian notation
Date: 21 Jan 1996 10:02:24 -0800
Organization: Computer Science, University of B.C., Vancouver, B.C., Canada
Message-ID: <[e-mail]>
References: <[e-mail]> <4cvu68$[e-mail]> <4d21og$[e-mail]> <4d2ok0$[e-mail]>
NNTP-Posting-Host: keats.ugrad.cs.ubc.ca

In article <4d2ok0$[e-mail]>, Jos A. Horsmeier <[e-mail]> wrote:
>Prefixing 'psz' to every zero terminated string is plain silly if we're
>talking about, say, employee names or beer brand names; something like
>'empnm' and 'beernm' should have been used instead. Only then, statements
>like:
>
>    if (!strcmp(empnmBoss, beernmMyFavBeer))
>
>trigger the reader that something fishy is going on here ... HN, as it
>is used now, doesn't add any clarity to the code at all, i.e. the statement:
>
>    if (!strcmp(pszBoss, pszMyFavBeer))
>
>just tells me that I want to compare my favorite beer and a boss's name,
>which is silly too, but it's not the 'psz' prefix that rings a bell here;
>it's just the 'Boss' and 'MyFavBeer' name parts that catch the eye.

I agree. The "psz" is just a redundant prefix that pollutes the symbol table.
If anyone wrote code with random consonants thrown in front of every perfectly
good name, I'd fire them on the spot, and tell them to go rearrange their
closet.

>Just adding domain name prefixes to the names of objects doesn't help
>anyone; notion of domains should be present in the language itself if
>one really wants domains, instead of just comparable types as implemented
>in the C language. Any (decent) C compiler feels perfectly fine if one
>feeds it something like:
>
>    #include <stdio.h>
>    #include <string.h>
>
>    typedef char* empnm_t;
>    typedef char* beernm_t;

Even that is going overboard. Someone will later see "empnm_t", and wonder what
the heck it is. If you mean "char *", just say "char *". The typedef operator
has its uses, but redefining simple types, like char * or int is not one of
them. The chief advantage is that you can use it with structs so that you can
eliminate having to write the struct keyword later on.

I just wrote a 12000 line project that doesn't have one typedef in it (save for
automatically generated XDR definitions). The code is easy to read; I can go
back to it and understand precisely what is going on without hunting for nested
definitions.

I gave the code to another programmer who is doing a Windows interface, and he
said that everything was nicely laid out for him, enabling him to incoroporate
the bits in a short time.

Typedef-like constructs are useful in poorly thought-out languages like Modula
or Pascal, whose compilers can typically perform typechecks based only on name
equivalence. That is to say, if I declare two types that are both integers, the
compiler will say that the two types are not equivalent, because it goes only
as far as to check the name. This is a cop out which simplifies doing one of
the more difficult phases of compiler construction, which is type checking. It
has nothing to do with implementing "domains", and everything to do with lazy
compiler construction. Real compilers will detect structure equivalence between
types so that you don't have to obfuscate your code with confusing type names.
To implement domains, just lobotomize the compiler by removing the structure
equivalence checking so that size_t and time_t become incompatible types.
--