Runtime Filters
Runtime Filters
Stumpless offers a number of functions that allow logging calls to be filtered out at compile time, causing zero performance impact to running workloads. However, sometimes you need the added flexibility to set and adjust filters in an existing binary, and compile-time filters are not feasible. Fortunately, there is also a runtime filtering mechanism that you can use to do just that.
By default, all targets have a runtime filter set that inspects the severity
mask that’s been set on a target. These masks are set to allow any entry
severity through by default, so if you haven’t messed with them then it’s like
they aren’t even there. This default filter mimics that available in the
standard syslog.h
setlogmask
capability, and is compatible with these
masks if you’re using them already in standard syslog code.
// the syslog.h code would look like this:
setlogmask( LOG_UPTO( LOG_ERR ) );
// the minimum change to use stumpless would look like this:
stumplog_set_mask( LOG_UPTO( LOG_ERR ) );
When the syslog.h
header is available, most syslog calls have a match in
stumpless with the same signature and semantics. For setlogmask
, that function
is stumplog_set_mask
. It sets the log mask for the current target, and returns
the mask that was previously being used. That’s what allows us to make the
substition in the above code so simple.
If you want to write code that is portable to systems without the syslog.h
header (Windows, for example), then you can replace the other syslog-specific
macros with the stumpless lookalikes:
// portable to any system that stumpless supports, which is a lot!
stumplog_set_mask( STUMPLESS_SEVERITY_MASK_UPTO( STUMPLESS_SEVERITY_ERR ) );
Yes this is a little more verbose, but it will work everyhere stumpless does. Of course, you can combine masks with bitwise ors and everything else you would expect from a mask as well.
// this will allow EMERG, ALERT, CRIT, ERR, and DEBUG severity messages
stumplog_set_mask( STUMPLESS_SEVERITY_MASK( STUMPLESS_SEVERITY_DEBUG )
| STUMPLESS_SEVERITY_MASK_UPTO( STUMPLESS_SEVERITY_ERR ) );
The default target filter will use the mask of the target to make these
decisions. If you want to work with a specific target instead of just using the
current one, you can use the stumpless_get_target_mask
and
stumpless_set_target_mask
functions to do this.
// let's see what we have now
mask = stumpless_get_target_mask( target );
// and make sure that DEBUG messages are NOT included
new_mask = mask & ~STUMPLESS_SEVERITY_MASK( STUMPLESS_SEVERITY_DEBUG ) );
stumpless_set_target_mask( target, new_mask );
But perhaps we have some magical filtering that we’d like to do. Maybe we want
to make sure that no entries are logged that contain an element named ignore
.
This seems reasonable.
Stumpless provides a simple way to accomplish this! First, we need to write a filter function that takes the target and entry as parameters, and returns a decision on whether to allow the entry. In our case, we just check for the existence of an element with our chosen name.
bool
ignore_element_filter( const struct stumpless_target *target,
const struct stumpless_entry *entry ) {
return !stumpless_get_element_by_name( entry, "ignore" );
}
And then we apply the new filter to the target like so:
stumpless_set_filter( target, ignore_element_filter );
That’s it! Now any messages sent through this target will use our filter function instead of the default mask filter. Just remember that if you want to combine filters, you’ll need to call all of them in your own filter, using whatever logic you’d like. For example, if we wanted our filter to do both mask filtering and element name filtering, it would look like this:
bool
ignore_element_filter( const struct stumpless_target *target,
const struct stumpless_entry *entry ) {
return !stumpless_get_element_by_name( entry, "ignore" )
&& stumpless_mask_filter( target, entry );
}
For a list of available filters provided by stumpless, check out the
stumpless/filter.h
header installed with the library or in the
documentation.