The last while I've been working on getting the Suneido standard library tests to run on jSuneido.
Mostly this is a matter of implementing built-in functions and methods. Occasionally I find a bug in the existing jSuneido code, but thankfully that hasn't been too frequent.
It's actually been going quite well. The tests run alphabetically and I've got them succeeding all the way up to 'N'. Actually, I'm surprised that many tests succeed since I still have quite a few built-in methods to implement. 80-20 principle, I guess. (i.e. 80% of the tests only require 20% of the built-in functions.)
But one thing that's been nagging me during this process is that the tests report how long they take to run, and they've been running extremely slowly. I wasn't surprized they would run somewhat slower than cSuneido, since I'm not using the fast server JVM and I haven't done any optimization.
But jSuneido was on the order of 100 times slower. That was a little scary. If it was really going to be that much slower then I might as well give up now. But I figured there had to be a reason for it, probably something stupid.
As I expected, it was something stupid. Suneido loads and compiles code from libraries on demand. Once loaded, the compiled version stays in memory. I had this all working, but I missed one small but critical piece - I wasn't saving the compiled version. So it was re-loading and re-compiling on every reference. Yikes! It's not surprising it was slow.
Now the tests actually seem to be running faster on jSuneido than cSuneido, although I'm not sure the resolution of the timing is accurate enough to tell. At this stage, as long as there's not a huge difference, I'm happy.