I'm not to "happy" with with the standard prefix format of the log message, and so I wonder what would be the best approach to customize it.

Typically I'm after something along the lines of

Timestamp, threadid, severity name - message itself

In addition to the above trace info (file and/or function name) may be relevant, possibly depending on severity. The severity names themselves will be defined in the application.

Lars Ivar

> I'm not to "happy" with with the standard prefix format of the log
message, and so I wonder what would be the best approach to customize it.
>
> Typically I'm after something along the lines of
>
> Timestamp, threadid, severity name - message itself

First, the "official" way of doing it is only the way adopted by the stock back-ends.

It's important to note that Pantheios is designed as a logging API, and the user is intended (and encouraged) to implement their own back-end, which could/would prepare the output statement formats in whatever way they deemed fit. The stock back-ends (and the de-facto standard formatting) is just there as a convenience to users, not in any way any officially prescribed or recommended way.

In some of the commercial customisations of front-/back-end systems that I've done the clients have had a variety of different requirements - some including all the fields you list - all of which are reasonably easily achieved.

> In addition to the above trace info (file and/or function name) may be
relevant, possibly depending on severity.

This can be achieved by using the Trace API for the ones you want file+line for, e.g. perhaps DEBUG, INFO and NOTICE.

To get __FUNCTION__ in, you can override (via the pre-processor). Check out the examples/cpp/example.cpp.tracing.with_function/example.cpp.tracing.with_func tion.cpp example.

> The severity names themselves will be defined in the application.

This is quite a bit more invasive. One client's had me do this for them, but it involved a degree of effort, as it was not something anticipated in the original Pantheios design (which assumes known-at-compile-time severity names in order to have absolute maximum name processing efficiency).

HTH

Matt

Matthew Wilson wrote:

>> I'm not to "happy" with with the standard prefix format of the log
> message, and so I wonder what would be the best approach to customize it.
>>
>> Typically I'm after something along the lines of
>>
>> Timestamp, threadid, severity name - message itself
> 
> First, the "official" way of doing it is only the way adopted by the stock back-ends.
> 
> It's important to note that Pantheios is designed as a logging API, and the user is intended (and encouraged) to implement their own back-end, which could/would prepare the output statement formats in whatever way they deemed fit. The stock back-ends (and the de-facto standard formatting) is just there as a convenience to users, not in any way any officially prescribed or recommended way.
> 
> In some of the commercial customisations of front-/back-end systems that I've done the clients have had a variety of different requirements - some including all the fields you list - all of which are reasonably easily achieved.

Yes, I'm already on to this.

> 
>> In addition to the above trace info (file and/or function name) may be
> relevant, possibly depending on severity.
> 
> This can be achieved by using the Trace API for the ones you want file+line for, e.g. perhaps DEBUG, INFO and NOTICE.
> 
> To get __FUNCTION__ in, you can override (via the pre-processor). Check
> out the
>
examples/cpp/example.cpp.tracing.with_function/example.cpp.tracing.with_func
> tion.cpp example.

Yup.

> 
>> The severity names themselves will be defined in the application.
> 
> This is quite a bit more invasive. One client's had me do this for them, but it involved a degree of effort, as it was not something anticipated in the original Pantheios design (which assumes known-at-compile-time severity names in order to have absolute maximum name processing efficiency).

Not sure I understand you fully here - the names will be known at compile time. They will just not be exactly the same as the predefined ones (DEBUG, INFO, etc) - but rather augmentations of those like INFO_1, INFO_2, etc

-- 
Lars Ivar Igesund
blog at http://larsivi.net
DSource, #d.tango & #D: larsivi
Dancing the Tango

> >> The severity names themselves will be defined in the application.
> >
> > This is quite a bit more invasive. One client's had me do this for them, but it involved a degree of effort, as it was not something anticipated in the original Pantheios design (which assumes known-at-compile-time severity names in order to have absolute maximum name processing efficiency).
>
> Not sure I understand you fully here - the names will be known at compile time. They will just not be exactly the same as the predefined ones (DEBUG, INFO, etc) - but rather augmentations of those like INFO_1, INFO_2, etc

Still not sure what you mean.

Can you enumerate the full list of desired severity levels and, if possible, indicate their equivalents in the stock (Syslog-based)
eight levels.

Cheers

Matt