Re: Exception specifications unfortunate, and what about their future?
on Thu Nov 20 2008, DeMarcus <demarcus-AT-hotmail.com> wrote:
Hi,
Recently I bought the book C++ Coding Standards by Sutter and
Alexandrescu. The book is good but there's one thing that they comment
in a way that doesn't feel right. I talk about item 75 in the book.
Item 75 says: "Avoid exception specifications.".
The authors have a lot of good arguments for not using exception
specifications, but there's one thing they're missing; the
specifications have a purpose, or at least should have.
Of course every feature in a language *should* have a purpose, but
sometimes, well, they just don't.
The authors neglect this fact with following argument: "People often
suggest switching from dynamically checked exception specifications to
statically checked ones, as provided in Java and other languages. In
short, that just trades one set of problems for another; users of
languages with statically checked exception specifications seem to
equally often suggest switching to dynamically checked ones."
I don't blame the authors, don't take me wrong, but to me it seems like
the exception specifications need to be fixed when two of the most
renowned C++ programmers recommend us to avoid them.
Yes; they should be removed.
A little less-glibly, they should be replaced with a statically-checked
"nothrow" specification. Unlike "is this going to be exception A or
exception B," failure to account for nothrow actually affects
correctness in most cases.
I think statically checked exception specifications (the same way as in
Java) should be implemented because of three things.
1. Prefer *compile time* check before run-time.
Some things just don't need to be checked at compile-time. Can you
imagine having to declare every derived class that can come out of a
factory function?
struct Base { virtual ~Base(); };
std::auto_ptr<Base> factory() emits(Derived1, Derived2, ... DerivedN);
It amounts to the same thing for correctness, maintainability, coupling,
etc.: a nightmare.
I usually dislike postings containing lists of links, but in this case
I've been over the argument so many times...
*
http://groups.google.com/group/comp.lang.c++.moderated/msg/c765149dbb040098
* http://tech.groups.yahoo.com/group/boost/message/22807
*
http://groups.google.com/group/comp.lang.c++.moderated/msg/9474fbe0c98d5a4a?
*
http://groups.google.com/group/comp.lang.c++.moderated/msg/9c9d52e31ecbedbd
* http://groups.google.com/group/comp.lang.c++/msg/4752eb5ceeda257a
etc.
--
Dave Abrahams
BoostPro Computing
http://www.boostpro.com
[ See http://www.gotw.ca/resources/clcm.htm for info about ]
[ comp.lang.c++.moderated. First time posters: Do this! ]