[OT] Live from the S-Plus 7 Financial Seminar

So this is what it's like to be a blogger...

The focus of the first talk was on their new "large data set"
handling.

They use a "pipeline architecture" to implement "out of memory" data
processing for gigabyte data sets. The way it works, is they store
pointers to the data in RAM but keep the data on disk using an
"efficient binary cache file." This is much faster than swapping.
The system allows the user to define what part, if any, of the data
is stored in RAM.

The user can use SQL on the dataset from the command line.

They also have a nice IDE with a project view window (like a DEPFUN
based view). The IDE uses Eclipse. The IDE has a nice multi-indexed
task list that lets the user click on a task and be sent to the
associated file.

As usual the UI is sl and the graphics are exceptional. They have
a nice new plot called a "hexbin plot" which is an alternative to a
scatterplot.

The interactive graphics using WinJava is very nice and their data
storage, with names for the columns and rows is nice as well.