Re: [boost] [scope_guard] Naming roundup

From: Andrey Semashev (andysem_at_[hidden])
Date: 2005-09-24 04:24:53

Peter Dimov wrote:
> Andrey Semashev wrote:
>
>>> FILE * f = fopen( ... );
>>>
>>> if( f == 0 ) throw ...;
>>>
>>> scoped_guard file_closer = make_guard( fclose, f );
>>
>> In this (or similar) case I might rather make a tiny wrapper around
>> FILE* instead of guard.
>
> Interesting... when Andrei Alexandrescu and Petru Marginean introduced
> ScopeGuard, this was one of their primary motivating examples, and my
> initial reaction was that, indeed, biting the bullet and writing a
> separate File class might be a better option.
>
> Eliminating resource management as a motivation leaves me wondering
> what are the primary uses for scope[d]_guard that you have in mind?

First, I didn't say the resource management area is not for scope guard. I 
just said that in this case it is questionable weither to use scope guard or 
making a tiny wrapper. And agreed that in this guard use case the "arm" 
method would be useless. Nothing more.

Second, in addition to Pavel's note, there also is a lot of use cases of 
scope guard class which cannot be put to resouce management or transactions 
category.

One of them, for example, is implementing post-processing queues which may 
help when you have complicated interactions between different objects that 
causes some of these objects get to an unexpected state. You may want to 
have such queue to perform the postponed actions after the call-stack is 
unrolled.

The another use case is ensuring some action will be made regardless of the 
result of some part of program (the resource management comes as 
sub-category of this use case). By the way, the "arm" method is mostly used 
in the latter field.

PS: Why scoped_guard? The object watches for the scope end to alarm, IOW he 
guards the scope. So scope_guard looks more applicable.