I discovered the impressive data.table package more than a year ago. In order to learn how to use it, I try to find a solution to some questions I read at mailing lists or at stackoverflow. My last experiment has been inspired by the bigvis package and its associated paper. This package is a proposal for exploratory data analysis of large datasets following the workflow of binning, summarizing and display.
Please note that I am neither trying to mimic the behavior of bigvis nor comparing both packages. It is only an excuse to learn more about data.table and this post shows the code I have used.
The first part of my experiment deals with one-dimensional data:

The second part is more sophisticated. It uses the movie dataset to show how to carry out 2D binning.

Some key points I have learned about data.table:

• := to add, remove or modify by reference (avoids memory overhead since it does not make additional copies)
• .N and .SD symbols for grouping

Still learning!