Re: [Boost-users] [boost::system] Is the library abstraction general enough?

Subject: Re: [Boost-users] [boost::system] Is the library abstraction general enough?
From: Lian Cheng (rhythm.mail_at_[hidden])
Date: 2009-02-24 01:38:06

Thanks a lot for the recipe! But the asynchronous MySQL API I'm working 
on is based on Asio, which uses Boost.System for error handling. 
Boost.System provides a unified method to handle errors with both 
traditional error code paradigm and the exception paradigm. Thus, there 
may be code like this:

     connector.async_query( "SHOW DATABASES", handle_query );

     void handle_query( boost::system::error_code const& ec ) {
         if( ec ) {
             std::cerr
                 <<  boost::format( "Query error: %1% - %2%" )
                     % ec.value() % ec.message()
                 <<  std::endl;
         }
         else {
             // ...
         }
     }

The ec.message() call above is redirected to the message() method of the 
user defined error_category class (amysql::client_category in my case) 
with an error code (ONLY!).

Currently, I use a boost::thread_specific_ptr<std::string> to store the 
last MySQL error message, and in the amysql::client_category::message() 
method, the error code parameter is ignored, and just return the thread 
specific last error message string.

The pitfall here is that an error_code object should never be stored for 
later usage, since the thread specific last error message would be 
changed whenever a new error occurred.

I think Boost.System should allow user to pass into a function object 
instead of a single error code to the error_category::message() 
function, so that user can have better control on error message 
generation policies.

Cheers
Cheng

Emil Dotchevski:
> On Mon, Feb 23, 2009 at 2:33 AM, Lian Cheng <rhythm.mail_at_[hidden]> wrote:
>> Boris Schaeling:
>>> On Fri, 20 Feb 2009 09:36:53 +0100, Lian Cheng <rhythm.mail_at_[hidden]>
>>> wrote:
>>>
>>> Isn't there a list of error numbers and error strings MySQL uses? I ask as
>>> calling error_category::message() should also return an error string even if
>>> no MySQL connection is used?
>> Unfortunately, no. As I've mentioned, the error strings MySQL uses are
>> printf format strings, such as:
>>
>>    Using unsupported buffer type: %d  (parameter: %d)
>>
>> And you always need an active MySQL connection handle to obtain the full
>> error message by calling mysql_error(). And you can never get the message
>> back once you close the connection:
>>
>>    MYSQL* conn = mysql_init( 0 );
>>    // do something causes an error.
>>    mysql_close( conn );
>>    printf( "error: %s", mysql_error( conn ) ); // Messed up.
> 
> Boost Exception can be used to transport anything in an exception
> object. You can transport error codes as well as error messages:
> 
> #include <boost/exception.hpp>
> 
> typedef boost::error_info<struct tag_mysql_error_code,int> mysql_error_code;
> typedef boost::error_info<struct tag_mysql_error_string,std::string>
> mysql_error_string;
> struct mysql_err: std::exception, boost::exception { };
> 
> ....
> if( int error=mysql_query(conn,....) )
>   throw mysql_err()
>     << mysql_error_code(error)
>     << mysql_error_string(mysql_error(conn));
> 
> Alternatively, if you use shared_ptr to manage the lifetime of the
> MYSQL connection, as in:
> 
> boost::shared_ptr<MYSQL> conn( mysql_init(0), mysql_close );
> 
> you can use an exception object to keep the connection afloat:
> 
> typedef boost::error_info<struct tag_mysql_error_code,int> mysql_error_code;
> typedef boost::error_info<struct
> tag_mysql_error_string,boost::shared_ptr<MYSQL> > mysql_conn;
> struct mysql_err: std::exception, boost::exception { };
> 
> ....
> if( int error=mysql_query(conn,....) )
>   throw mysql_err()
>     << mysql_error_code(error)
>     << mysql_conn(conn);
> 
> Now, when you catch mysql_err, you can talk to mysql to get the error message:
> 
> catch( mysql_err & x )
> {
>   boost::shared_ptr<MYSQL> conn=*boost::get_error_info<mysql_conn>(x);
>   char const * msg=mysql_error(conn.get());
>   ....
> }
> 
> Emil Dotchevski
> Reverge Studios, Inc.
> http://www.revergestudios.com/reblog/index.php?n=ReCode