On Tue, Nov 17, 2015 at 4:08 PM, Darryl Freinkel
<dfreinkel@xxxxxxxxxxxxxxxxx> wrote:
Leading on from it, with the size of the spreadsheets, how do I determine a
good number to use for "ss_begin_object_group".
My number I use now is 100, which seems to be way to small.
My earlier post about not having to worry about garbage collection due
to the nature of Java was clearly naive (because I didn't account for
the mechanics of JNI calls via RPG). Fortunately, that post seems to
have been drowned out by these later ones.
But to me, seeing this profusion of posts is like watching
whack-a-mole. You come up against problems with installation. OK,
things seemed installed, but then there's CLASSPATH issues. OK, those
are resolved, but now what about memory allocation?
It feels endless. I want to stress for folks that are not already
knee-deep in POI-via-RPG (and maybe *especially* for folks that are!)
that there are easier ways. My preferred way is Python, but there are
others. Perhaps the most straightforward, for folks who have already
made the investment of learning Java, is to just use Java. That is,
forget about wrapping POI with RPG, and just call POI directly from
Java. Then you *will* have Java's native garbage collection take care
of everything for you.
Perhaps the most damning thing about POI-via-RPG is that even after
fighting through all the setup issues and finagling with memory
allocation, you *still* don't get very impressive performance. So
you've got almost the worst of all worlds: more complicated setup,
more complicated coding, poor-to-lackluster performance. The *one*
thing you get with POI-via-RPG is that you don't have to learn another
language. (Yes, I understand that can be a big one, but it's still a
little sad.)
John Y.
midrange.com
