NetNews Usenet Archive 1992 #31/NN_1992_31.iso/spool/comp/lang/c/19154

Text File | 1993-01-03 | 9.2 KB | 199 lines

Newsgroups: comp.lang.c
Path: sparky!uunet!ferkel.ucsb.edu!taco!gatech!destroyer!cs.ubc.ca!newsserver.sfu.ca!darcys
From: [e-mail] (Darcy Smith)
Subject: Re: Newbie Wants Advice on C-Programming
Message-ID: <[e-mail]>
Sender: [e-mail] (D'Arcy Smith)
Organization: Simon Fraser University, Burnaby, B.C., Canada (sortof)
References: <[e-mail]> <[e-mail]> <[e-mail]>
Date: Sun, 3 Jan 1993 19:00:02 GMT
Lines: 187


From newsserver.sfu.ca!cs.ubc.ca!destroyer!gatech!swrinde!zaphod.mps.ohio-state.edu!cs.utexas.edu!uunet!pipex!bnr.co.uk!uknet!mcsun!sun4nl!and!jos Sun Jan  3 09:27:40 PST 1993
Article: 34226 of comp.lang.c
Xref: newsserver.sfu.ca comp.lang.c++:27833 comp.lang.c:34226
Path: newsserver.sfu.ca!cs.ubc.ca!destroyer!gatech!swrinde!zaphod.mps.ohio-state.edu!cs.utexas.edu!uunet!pipex!bnr.co.uk!uknet!mcsun!sun4nl!and!jos
From: [e-mail] (Jos Horsmeier)
Newsgroups: comp.lang.c++,comp.lang.c
Subject: Re: Newbie Wants Advice on C-Programming
Message-ID: <[e-mail]>
Date: 3 Jan 93 10:14:04 GMT
References: <[e-mail]> <[e-mail]> <[e-mail]>
Followup-To: comp.lang.c++
Organization: AND Software BV Rotterdam
Lines: 57

>In article <[e-mail]> [e-mail] (Bjarne Stroustrup) writes:
>|
>|[e-mail] (Jim Lynch @ Netcom Online Communications Services (408-241-9760 login: guest)) wrote
>|
>| > >You should always start out with the official books:
>| > >
>| > >For C, it is _The_C_Programming_Langauge_ 2nd ed. by Kernighan & Ritchie
>| > >For C++, it is _The_C++_Programming_Language 2nd ed. by Stroustrup
>| >
>| > No no no no no no no no no no no no. These books are _unreadible_ to a
>| > beginner.
>| > However, they can be quite valuable as a reference (whenever you get
>| > around to
>| > learning how to read them...)
>
>|It is worth remembering that `beginner' means different things in different
>|contexts and what `unreadable' means depends critically on what kind of
>|beginner we are talking about.
>|
>|If by `beginner' you mean `someone who approaches programming for the first
>|time,' you are almost certainly right (though I have heard of exceptions to
>|that rule). If, on the other hand, `beginner' means `experienced programmer
>|who just happens never to have encountered C or C++ before,' you are with
>|a high propability wrong (though again I suspect a few exceptions could be
>|found).
>|
>|The backgrounds and expectations of want-to-be C or C++ programmers seem
>|infinitely varied and textbooks, approaches to teaching, and advise must
>|try to take that diversity into account. No one approach or one book can
>|cover all angles.
>
>I think there's a bit more to leaning C++ than there is in learning C.
>C++ uses quite a different paradigm: the object oriented programming
>approach. Here at my company, they made me volunteer (with a gun pointed
>to my head ;-) to teach the programmers team the concept of object
>oriented design and programming. My first attempt was using your book
>along the line of explaining inheritance, polymorphism, data encapsulation
>etc. etc. We ended up discussing the misty details of the C++ language.
>I completely failed here. My second attempt was, not using any programming
>language at all, but just a clear explanation of the object oriented
>concepts and mechanisms. When all that was clear, we switched to C++.
>
>This approach was much more fruitful: the C++ language became just
>a mechanism to express these object oriented mechanisms. I kept on
>convincing these people that it doesn't matter _how_ you express these
>concepts, as long as the language offers you the building blocks to
>express them. IMHO there are tree types of `beginners':
>
>- First time programmers;
>- Experienced programmers but still unaware of the OOD concepts;
>- Experienced programmers aware of OOD concepts, but they don't know
>  anything about the C++ language.
>
>I think that one has to be a beginner of the third kind to be able
>to learn the C++ language successfully.
>
>kind regards and happy new year,
>
>Jos aka [e-mail]
>
I'd like to provide some exerpts from an "Interview with Bjarne Stroustrup"
found invol. 1 no.3 1991 of "The C++ JOURNAL"  - Bjarne please forgive me!


Q) What do you recommend to people who want to learn C++ but have the idea
   of the learning curve hanging over their heads?  I don't think that this
   is the case anymore but the idea of this huge learning curve still exists.

A) ...Get a text book or two, at least two because you can never trust just
   one author.  The problem is you can't know what are the view points and
   what are the facts.


Q) At which point should they start learning object-oriented concepts.  Is
   that something to be learned initially or something to be learned later
   on?

A) It depends.  It strongly depends.


Q) ...when I teach...C++, I start with object-oriented concepts and then map
   them into C++ but there are others who go the other way.

A) Lets talk about this because it is tricky.  ...I invariably start with
   history amd background and aims of C++.  ...I don't see how you can
   understand things, unless you understand thier background.  ...These are
   the fundamental 3 ways to use C++:

   1.  as a better C
   2.  supporting data abstraction
   3.  supporting object-oriented programming

   ...


Q) Have you incorporated these ideas into you new book?

A) No, "The Annotated C++ Reference Manual" is a straight reference manual.
   You should not try to learn C++ from that.  If you are more than a casual
   user, you should have it, though.  There is a statement here in the book
   states the aim: "this book provides a complete language reference for the
   expert C++ user."  It is not a tutorial.  It has more about C++ than you
   ever wanted to know, because that is what a good reference manual has to
   have. ...[his first book] That book is not a manifesto for writing
   programs the way that I think you should write them. ...There are many
   ways of writing programs and "the best way" depends on the application,
   personality, and the background of the person writing.  there is no one
   right way.  There are so many people that write manifesto's saying,
   "Do thing my way!"  If anything, my manifesto is, "Think very hard about
   rhat you are doing, then do it your way!"  I am trying to provide tools
   for supporting such a view.
   ...If you have a lot of C code already and you need to do a change in
   some corner of it, then it is probably very wise to start up slowly.  If
   you have to interface with a lot of C code, then the object oriented
   features are not going to fit in all that well.  So what you do is to tool
   up first.  You get used to your new tools.  You use C++ as a better C,
   tighten  the type system.  You use type-safe linkage to make sure you
   don't have linkage problems.  Start throwing in a little bit of data
   abstraction. We work that way in real production environments.  We have
   found that you can't take just a bunch of programmers out of their
   productive lives for, say a year and a half to teach tham to "do
   everything right."  Forget it.  We have to deliver a product within a half
   a year, a year or whatever.  Does that mean you can't go C++?  No, you can.
   Can you go use C++ in a radically new style?  Not really.  So you can
   throw in the "better C" and a bit of data abstraction.


Q) At which point does the object-oriented programming and design come into
   play?

A) You can not do effective OOP if you already have a framework that is
   designed to prevent it. ...[the gist of this is try a new project OO
   when you are ready!].


Q) There are some people who feel if you are going to use C++, you should
   use it in a completely object-oriented manner.  How would you respond to
   that?

A) What does complete object-oriented mean?  Some of the worst C++ programs
   that I have seen, were written Lisp and Smalltalk styles.  C++ is not
   Smalltalk and Smalltalk is not C++.  Same with Lisp.  If you coded an
   imitation Smalltalk in C++, why would you expect it to run faster than
   Smalltalk?  You are not as smart at implementing Smalltalk's basic notions
   as Peter Deutch.  Simulating something is always less conveinient and
   less efficient than using the real thing.  [Bjarne, didn't you write a
   simulator a part of your Ph.D.? - just kidding] If I had wanted to to
   build an imitatin Smalltalk, I would have built a good imitation.  I was
   building a good C++.  C++ is not a good Smalltalk amd Smalltalk is a
   lousy C++.


OK - sorry that this is so big - and there is much much more - however I will
stop here! - you welcome!

So my point, so nicely expressed by Bjarne, is do not use C++ in a way that it
shouldn't be used.  If your project is suited to OO ideas and you are
comfortable in useing tham - the USE them where APPROPRIATE.

You do not ever have to use a Class, inheritence, polymorphism or any other
OO idea if you don't know how [but mabey you should learn!] or if it is not
applicable to your program.

Fanaly here are two other C++ books:

   The C++ Programing Language - Bjarne Stroustrup
   C++ Primer - Stanley B. Lippman

I'm gonna go play in the snow now [usualy it rains here]

D'Arcy