Omega
A group of us
from a variety of different insitutions have recently started a project
to develop a new statistical environment.
The specific goals are to provide
- an extensible interactive (console and GUI) environment
similar to R, S, Xlisp, Matlab, etc.
- extensible data structures and algorithms
for data analysis,
- access to modern facilities such as JDBC, GUIs, multi-media,
network support, etc.
- an embeddable interpreter that can be used in
different environments such as web pages, interactive
analysis, distributed clusters, etc.
- support for compiling functions for efficient
performance and stand-alone operation (outside the
interpreter) aiding technology transfer and distribution of
algorithms.
More general considerations include
- the ability to handle large datasets,
- a spectrum of programming levels
(using optional type-checking, compilation, etc.)
- reuse existing software (via CORBA, or some such mechanism),
- distributed and parallel algorithms,
The project represents a unique opportunity for the developers of
three commonly used research environments to collaborate with each
other and the statistical community in general.
The project is open
source, meaning that individuals will be able to use, redistribute and
derive new products from it. We strongly encourage people to
contribute their classes to the effort.
Java
The project uses Java as the
primary development tool. Its reflectance facilities make it an easy
language in which to develop an interpreter. Its unparalleled
popularity means that there is an enormous amount of research related
to it, including many third-party tools, libraries, etc. Perhaps most
importantly, it provides a structured development environment that
allows us to engineer extensible and well-formed implementation that
others can easily adapt. The core support for graphics, GUIs, network
capabilities, database connectivity, CORBA/RMI,
persistence/serialization, etc., coupled with the extensive
documentation and acceptance make it an almost ideal language for this
project.
Status
As with most projects, getting a solid foundation is
difficult especially with multiple participants
separated geographically.
However, we have built a complete interpreter
that supports many of the language features in S, R and Xlisp-Stat.
The language is a superset of Java, providing
- full access to public Java classes, methods and fields;
- databases/work areas for storing variables
with support for optional type declarations,
constant/read-only variables, persistence,
archive support,...;
- functions including closures and tail-recursive functions,
optional argument types, variable number of arguments,
missing arguments;
- multiple evaluators in separate threads;
- simple and extensible subsetting operations providing
easier access to common statistical operations;
- dynamic class resolution;
- dynamic array initialization;
- sequence operation;
Additionally, there is support for
- expression and functions as
event listener actions/callbacks for arbitrary
event sources. This allows easy scripting
of GUIs.
- dynamic CORBA method invocation (DII),
and server implementation of arbitrary
Java objects (DSI), conversion of Java classes
to IDL;
- extensible data scanners for reading different
file formats (e.g. tables, spreadsheets, XML-formatted
data, etc.)
- both console/terminal and GUI access to the environment;
- scripted startup and customization via options/properties;
More Information
More information can be found on
the primary site
as we get time to add it.
Duncan Temple Lang
<duncan@research.bell-labs.com>
Last modified: Wed Dec 16 09:22:23 EST 1998