Re: [boost] Review Request: Boost Exception

From: Emil Dotchevski (emildotchevski_at_[hidden])
Date: 2007-07-09 16:18:58

I guess pointing at the documentation of the proposed Boost Exception counts 
as generating buzz :)

http://www.revergestudios.com/boost-exception/boost-exception.htm

The source code with boost-build jamfiles and tests can be downloaded here:

http://www.revergestudios.com/boost-exception/boost-exception.zip

The source code is also available for online browsing:

http://www.revergestudios.com/boost-exception/

As for selling the library, I can point out the 3 main advantages of using 
Boost Exception:

1) With Boost Exception, exception class design is decoupled from what data 
each exception object needs to transport to the catch site. For example, 
consider boost::filesystem, the basic_filesystem_error class template is 
designed to store 2 paths, called path1 and path2. If Boost Exception was 
used, there would be no need to anticipate all possible data that could be 
available to the throw site (note that some such data could be 
platform-specific), and also it would be possible to give more meaningful 
names to path1 and path2.

2) Contexts above the throw site can add data to any boost::exception that 
passes by. If we consider a boost::filesystem example again, a library that 
uses boost::filesystem but does not handle the exceptions boost::filesystem 
emits could intercept a "file not found" exception and augment it with 
library-specific information about the failure.

3) Without Boost Exception, often it is necessary to pass data to a throw 
site just so it can pack it in exceptions in case of failure. Typical 
example is a file read operation: somehow, you need to make the file name 
available to it, even though the read operation itself doesn't need it. With 
Boost Exception, the context that detects and reports a file read error 
wouldn't put the file name in the exception in throws -- this will be done 
by the context that opened the file, by intercepting _any_ boost::exception 
and augmenting it with the file name (because the file name is relevant to 
any failures, not only I/O failures).

I am contemplating refactoring boost::filesystem to use the proposed 
boost::exception for demonstration purposes. Would anyone be interested in 
seeing such an example?

Emil Dotchevski

Felipe Magno de Almeida wrote:
> On 7/8/07, Tom Brinkman <reportbase_at_[hidden]> wrote:
>> No one has stepped up to volunteer to be the review manager.  I
>> suggest that you generate a little buzz around the library, post a
>> couple of threads and "sell" your library to the boost developers.
>> C++ "exceptions" are a controversial topic with many different points
>> of view.  Tell us why we need boost::exception and why does your
>> library solve that need?
>
> It is not mine library, but I really see potential in this library.
> Every time I need something like this I remember this proposal and
> find that it hasn't been reviewed yet.