recently I worked on a 120MB XML subset of the Wikipedia dump with the dbpedia extraction framework in Scala. I found that REPL uses much more system resources than the compiled program: 2GB vs. 600mb RAM and 3 cores vs. 1 core.
I know that REPL is actually compiling internally. However, I just wrote a few lines with one reference to the framework. So why is the difference so big?
Update
I found the problem. I create an iterable on top of the Wikipedia dump. REPL then iterates over it to out put it. If I create a view that does not happen.