Re: [Jython-users] Extending the runtime flexibility of my Java application using Jython

[Bill K. <bi...@ct...>]

From: "Carlos Quiroz" <car...@we...>
>
> On Thursday 04 October 2001 00:23, Mat...@i2... wrote:
> 
> > Heh, yeah, I wrote a GUI jython script launcher initially in jython, it
> > was about 1000 lines, had to rewrite it in Java (for speed) and it was >
> > 2300 lines, so I feel your pain =)
> 
> Hi.
> Do you have any estimation how much slower is using jython than java? I'm 
> very interested to know that kind of figure

I have extensive timing and profiling numbers for our jython html
parser and layout engine, and a rewrite of the layout engine in
java.  Well, 'extensive' in the sense that I accumulated a lot of
them in trying to tweak the jython code in the hopes of not having
to rewrite it. :)

The parser, interestingly perhaps, hasn't needed to be rewritten
in Java.  The parser, however, is very simple, relying on the 
lexer to do most of the work - and the lexer is just a big
regexp - and so in that sense we can say the lexer is 'written
in java' (in terms of the performance of the regexp engine, I
mean.  Certainly no-one in their right mind would write a lexer
>in java< when they could do it in a couple lines of regexp. ;-)

Okay.  So there's a point I'm trying to get to about the ratio
of jython-to-java, or rather, the percentage or degree to 
which a bunch of jython code relies heavily on calls to Java
classes to do much of its work - sort of a 'glue' or scripting
for the Java components, to put them together and get them to
do things; vs. jython code implementing an algorithm strictly
in pure jython.  I think our layout engine falls squarely in
the latter category, while our parser falls to a degree toward
the former; . . . well anyway, here are some numbers :)

The test data for these was a simple html document of 1364 
words (9893 characters) with a couple dozen tags strewn 
throughout, and an inline image contained in a simple table.

The jython implementation:

parse x 100: 3.5049999952316284 secs (0.03504999995231629/per)

layout x 10: 3.8459999561309814 secs (0.38459999561309816/per)

So layout was accomplished in about 2/5ths of a second.  But,
this is timed on a 1.3GHz athlon system, and our product's
minimum requirements specify a Pentium 166Mhz (or similar
speed Macintosh.)  And unfortunately, the _three seconds_
the layout might be expected to take on such a machine (and
indeed did) were a bit more that it seemed reasonable to 
ask users to tolerate. =)

Anyway here's the java reimplementation (port really) of
the layout engine (again on the GHz athlon)

layout x 1000: 8.332000017166138 secs (0.008332000017166137/per)

Zoom!  Not quite two orders of magnitude but, just about
one-and-a-half . . . 

But, again, what we're doing in the layout engine - 'pure'
jython, no Java methods called at all, really, 'cept, well
okay stringWidth() from FontMetrics, not much else - is
about as biased an example toward the poor-performance 
end of the spectrum as one can get.  We're quite happy
with the performance of the rest of our jython modules,
including the parser - and the ones that do things like
glue swing components together - I mean, speed just hasn't
been a problem for those kinds of modules.

So I didn't want to sound in any of this like I'm down on
jython or anything.  I love it actually, it's the best
alternative to Java going right now !   O:-)

That runs on the JVM I mean.  <g>

Okay.  Hope this helps,

Regards,

Bill