[Biojava-l] 1.4 vs 1.5

[Dr. Christoph Gille]

Recently we had a discussion whether Biojava could use the novel
features of Java1.5. Since I just have moved two larger applications
(633 java files) to Java1.5 I would like to share my experiences with
you.

Applying the new features to the source code took me two days: This
was worth doing because I identified two bugs thanks to the Generics
and Annotations of Java 1.5.

1. GENERICS: I added types to all collections E.g.
public List getProteinsV() { ... } was turned into
public List<Protein> getProteinsV() { ... }
I found one bug where I added the wrong Object type!

2. ANNOTATIONS: I preceded all methods that override a method of the
parent class with the annotation @Override.  Indeed I found a hidden
bug where I mistyped the name of a method !  Instead of of overriding
a method I invented a new one which was not intended.  This kind of
bugs remains unnoticed in a Java1.4 environment.

3. Loops: I achieved a more compact source code by using foreach
loops. The code is better readable now. In 1.4 the head of loops
sometimes require 3 java lines which is now condensed to one single
line.


RETROWEAVER

A sound argument against 1.5 was the broken compatibility to
application servers still working with 1.4 and old Macintosh
OSX. I used Retroweaver to convert the class files after compilation
into 1.4 class format. As a result the program works on a 1.4 virtual
machine as well as on a 1.5 machine. Fortunately, I did not find any
problem related to the code conversion by Retroweaver.


PERFORMANCE:

The foreach loops are slightly slower.  The autoboxing feature is
dangerous in terms of performance because expensive object creation
is hidden.
For example the compiler would conveniently replace "10" by new
Integer(10) for method parameters that require "Integer" and not
"int". Therefore, I do not like autoboxing.

I did not try but the alternative to StringBuffer is said to be faster
because thread safty is omitted but still lacks standard String operations
from other languages.


DISADVANTAGES:

1. Jikes can not be used any more. Jikes compiles faster than javac
and has a better error report.

2. The make script takes longer because Retroweaver must be run.

3. Some additional class files shipped with Retroweaver are required
at runtime and makes the binary larger by 60kbytes. Well, that is not
really significant.

Conclusions: I would highly recommend migrating Biojava to 1.5.

I hope this helps to make a decision.

Christoph