[Biojava-dev] nio for io?

[mark.schreiber at novartis.com]

Hello -

I have been reading about the NIO api recently (very remiss of me to leave 
it this long). There seems to be potential for this to greatly speed up 
our parsers such as memory mapping etc, however ....

Our current Format objects want to see something like a BufferedReader. We 
would probably need to make an extension to the Format interface to use a 
FileChannel of some sort as input.

If I do something so naive as use the Channels utility to wrap a Channel 
into a BufferedReader I get no advantage (it's even a bit slower).

There has been some talk of Sun reprogramming normal java.io objects to 
use NIO internally, does anyone know if this has happened (if it has there 
is probably not much point in us looking into it much)?

Are there people more experienced with this API that can see some 
potential for it to improve parsing speed (probably the biggest 
bioinformatics bottleneck)? Comments greatly appreciated.

- Mark

Mark Schreiber
Research Investigator (Bioinformatics)

Novartis Institute for Tropical Diseases (NITD)
10 Biopolis Road
#05-01 Chromos
Singapore 138670
www.nitd.novartis.com
www.dengueinfo.org

phone +65 6722 2973
fax  +65 6722 2910