J Users,

Good morning.  I've been looking into Array Based Languages for work I'm doing 
with large arrays of binary data.  Currently, our lab uses a proprietary 4GL 
library for number crunching.  My thought was to get feedback on the 
possibility of switching over to an open source library like J.  The software 
would be Windows based, just like our current package.

Presently, I wrap the entire proprietary system with a Visual Studio based 
interface, due to the clunkiness of the library using Windows.  The library we 
use is geared towards OpenVMS and Linux; we chose it due to are previous VMS 
platform.

The software I've develop reads in mixed binary and text data files containing 
large arrays of data.  The data used to be integer with at most 8 parameters in 
one direction and up to a million events for each parameter (2 x 8 x 1000000 or 
up to 15 MB per file).  Now, however, the data is moving into floating point 
with many more parameters and event counts.  Would J be suitable for arrays of 
this size?  I split off the parameters into single dimension arrays throughout 
the code.  Arrays are combined, multiplied, masked and filtered then displayed 
as histograms, dot plots and contours.  From the literature on J, this all 
seems to be within scope.

One CPU intensive routine we employ is Cluster Analysis using a K-Means 
algorithm to find cluster centroids.  

In addition to the wrapper I wrote around the current library, I've also 
written an IDE for development and testing purposes.  Our library is basically 
a console program with Windows Event handling added.  It appears that J is also 
a console based environment which may be easily incorporated into my IDE.  
Currently, I handle the creation of the process running the proprietary code by 
.NET process functions and output redirection.

Anyway, any thoughts or suggestions on the use of J in this endeavor would be 
appreciated.  I've also looked into the statistical package "R" (what's with 
the letters, you know it is hard to search on google with simply a letter, 
perhaps "xyzzy" would be more unique).

Sincerely,
dgp
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm

Reply via email to