with the ganglia_slope.diff, perhaps you could use calls to some gperf-generated perfect hash within cstr_to_slope(). Using a perfect hash would save the multiple calls to stricmp(), although this is probably not huge performance issue ;-)
While in general I hate bit switch statements, in this case, it's only called when _creating_ a new rrd so the performance gain of doing something more exotic is about 0 ;-) (as you mentioend) A minor typo: in the patch (attachment #1 to bug #143), slope_to_ds()
returns "GUAGE" (instead of "GAUGE") for the unknown case.
DOH!!!! Thanks! with GANGLIA_SLOPE_NEGATIVE I don't think COUNTER is correct. I really have no idea what to do with slope negative! Before last week it was impossible to use gmetric to submit a postive or negative slop. I'm not sure what's the right solution is, if any. I'll regenerate the patch with typo fix tomorrow. thanks all. --nickg
