Josh Hurst wrote: > On 2/1/07, Vasiliy <vassun at gmail.com> wrote: > > Yes, that db it was result of 3 years long "Regestry 4" project > > claimed to improve performance twice - unfortunately it was > > shown only on test packages. I and Gary was directly involved > > to pulling it off and by just improving packaging and patching > > code (side effect of zone support) really improve performance > > twice for packaging. > > > > There is some code in patchadd shell script or possible > > in other scripts left and it may run some utility which > > check database existance so have a look at patchadd.ksh also. > > Would it help to use ksh93 instead of ksh if the performance depends > on the shell?
Yes and no. Yes it would help. Long ago I reimplemented parts of these scripts using ksh93 mainly using builtin commands, associative arrays and a small "cache" which resulted in a performance boost between 20 and 60 times faster than the original Solaris 8 script code, mainly thanks parsing all the data, using builtin commands and operators (and otherefore avoiding the zillions of |fork()| calls triggered by the original script) putting it in several associative arrays and then using some database-like operators to work on the data (the implementation was still slow and a better algorithm could still squish more performance out of the code). Some operations like applying large patch clusters were suddently done within minutes instead of hours (at the expense of storing all the data in memory). The problem is... this work was more or less obsoleted because someone reimplemented the whole thing in C code (which means the final answer whether we can do this is "no" because there are no scripts to work with). ---- Bye, Roland -- __ . . __ (o.\ \/ /.o) roland.mainz at nrubsig.org \__\/\/__/ MPEG specialist, C&&JAVA&&Sun&&Unix programmer /O /==\ O\ TEL +49 641 7950090 (;O/ \/ \O;)
