[Mulgara-dev] Performance problem with StackTrace generation

Paul Gearon gearon at ieee.org
Fri Aug 22 04:58:23 UTC 2008


Hi Mourten,

On Thu, Aug 21, 2008 at 5:10 AM, Morten Grouleff <mgr at trifork.com> wrote:
> While profiling a call to "Fedora risearch", I noticed a lot of time spent
> in StackTrace.getStringTrace(), called by HybridTuples and CacheLine. In my
> particular run, the time spent was 54% of all time spent processing the
> query. That is a high price to pay for debugging messages, that were never
> even printed.
>
> So, in line with what is done in other classes in Mulgara, I suggest to
> encapsulate these stacktrace-generation and printouts in "if
> (logger.isDebugEnabled())". The code still logs errors on double close, but
> does not make stacktraces while not in debug-logging mode.
>
> Proposed patch attached.

Thanks for this. I've checked this into trunk. I'm doing a minor
release soon, so you'll see it then.

Regards,
Paul Gearon



More information about the Mulgara-dev mailing list