Comparing Python and Java: Performance & Standard Libs

[neurocomp2003]

Any one have any opinions on the differences between these two language (and their closeness to C)...

chroot posted about python being almost as complete as java interms of standard libs. I attended a presentation which seemed to verify that claim though you have to download some of the added libs like numpy,scipy,wxpython.

Is there a performance difference between the two?

My goal is to try to minimize the number of packages to be downloaded for my application which currently uses both java/python. Not sure why since wxpython seems to be platform independent but they use java's awt/swing because of clientside platform independence.

 

[chroot]

Python's a better language, all around, in my opinion. The syntax, built-in data types, weak-typing, and other features just make it incredibly fluid and flexible.

The standard library is certainly as big as Java's. The only downside is that it's rather messy, since it was written by a ton of different people. There are no universal naming conventions -- some people use initial-capped class names, some use all lowercase. Some have oscure names that make them hard to search for. (If you were looking for an object serialization system, would the word 'pickle' come to mind?)

Python's performance is often much better than Java's. Experienced Python programmers can write code such that the processor spends most of its time in the underlying native, optimized C implementation, rather than in the interpreter code. The best way to use Python is as high-level coordination of native modules like NumPy and wxPython. I hesitate to call Python "glue," since you can do almost anything in pure Python. On the other hand, if speed is essential, try to use libraries with native implementations. In my real-world experience, Java is significantly slower than Python.

You can even compile virtually any Python program to a native executable with free, open-source tools like Pyrex. There are some similar tools for Java, but they're not nearly as capable.

- Warren

 

[haki]

Why does Python have such a poor performance on the language shoot out page?

http://shootout.alioth.debian.org/gp4/benchmark.php?test=all&lang=java&lang2=python

I tought that the power of Java lies in the VM (it does dynamic optimization rather than compile time static optimization).

 

[chroot]

From what I can tell, these people implemented CPU- and memory-intensive algorithms in pure Python. That's a very good way to ensure your Python program is slow. I'd be much more interested in seeing a comparison against Python programs that use native libraries (like NumPy) in a way that a real programmer would use them.

- Warren

 

[-Job-]

Does Python support just-in-time compilation? A second run of a Java application is largely native due to JIT. The values used in the benchmark were probably on runs after JIT has taken place, considering it's a server application.

Otherwise it has probably to do with optimization like haki mentioned (where Java's rigorous yet verbose standards might be a factor).

 

[chroot]

There any many different implementations of Python, some which do JIT and some which do not.

- Warren

 

[-Job-]

The non-JIT Python implementations have to be considerably slower than Java 6.

 

[chroot]

That depends on how the code was written, and how well the programmer takes advantage of natively-coded modules.

- Warren

 

[-Job-]

Same goes for Java though.

 

[igouy]

Why does Python have such a poor performance on the language shoot out page?

http://shootout.alioth.debian.org/gp4/benchmark.php?test=all&lang=java&lang2=python

I tought that the power of Java lies in the VM (it does dynamic optimization rather than compile time static optimization).

Those measurements are OUT-OF-DATE!

Here are up-to-date measurements -

http://shootout.alioth.debian.org/u32/compare.php?lang=python3&lang2=java"

http://shootout.alioth.debian.org/u32q/compare.php?lang=python3&lang2=java"

http://shootout.alioth.debian.org/u32/compare.php?lang=pypy&lang2=java"

 

[igouy]

From what I can tell, these people implemented CPU- and memory-intensive algorithms in pure Python. That's a very good way to ensure your Python program is slow. I'd be much more interested in seeing a comparison against Python programs that use native libraries (like NumPy) in a way that a real programmer would use them.

It's a good way to look at the performance of pure Python ;-)

 

[igouy]

Does Python support just-in-time compilation? A second run of a Java application is largely native due to JIT. The values used in the benchmark were probably on runs after JIT has taken place, considering it's a server application.

No - there's no link between a second run of a Java application and a previous run.

We can force there to be a link by starting Java and then repeatedly running the application once, twice, ... without stopping Java - http://shootout.alioth.debian.org/help.php#measure" .

Java Hotspot JIT compiles hot methods to native code and switches to those compiled methods while the program is being run.

 

[igouy]

There any many different implementations of Python, some which do JIT and some which do not.

- Warren

Which Python JIT implementations are ready for production use?

 

[Colleen1]

Python is a programming language that let's you work more quickly and integrate your systems more effectively. It supports multiple programming paradigms (primarily object oriented, imperative, and functional) and features a fully dynamic type system and automatic memory management.