Upcoming feature: Arrow file format support

[lwhite1]

In a couple weeks I will be starting a new gig as an open source dev working on Apache Arrow so i thought I would provide Arrow support in Tablesaw.

If you are not familiar with Arrow, it provides a high-performance, cross-platform 'backend' for dataframes, and a standard way to exchange data between data science tools and projects. The project was created by Wes McKinney, who also started the Pandas project. Arrow is widely supported as a data interchange standard so it would provide another way for Tablesaw to exchange data with projects like Dremio and Apache Spark, as well as other dataframes like Pandas and the R dataframe.

While Arrow is at its best when used as an in-memory data structure, initially, at least, Tablesaw support will be limited to reading and writing files in the Arrow format. Arrow's data model is based on a memory-mapped binary structure that would require an extensive rewrite to implement within Tablesaw, so that's off the table - at least for now.

Arrow IO will be implemented as a separate sub-project, like our excel support, so there should be no additional dependencies and few if any changes to Tablesaw core or other subprojects.