[ARROW-3205] [R] Minimum working example round-tripping a data frame from R to plasma to pandas

• Type: New Feature
• Status: Closed
• Priority: Minor
• Resolution: Abandoned
• Component/s: R
• External issue URL: https://github.com/apache/arrow/issues/19550
• Assignee: Unassigned
• Reporter: James Lamb
• Created: 10/Sep/18 04:12
• Updated: 11/Jan/23 07:25
• Resolved: 25/Sep/20 23:16

Description

I see tremendous opportunity for interoperability between Python and R (two popular languages for data scientists) using Arrow as an interchange format.

To make this concrete and get developers in those languages interested, I think it would be valuable to create a minimum working example of writing an R data frame into plasma and reading it back up into pandas in a separate Python process, and vice versa.

I could, for example, envision reading a CSV up into a data.table in R to do some cleaning and feature engineering, writing that object to plasma, then kicking off multiple parallel Python processes to search a space of models. This could demonstrate the benefits of replacing "load this dataset from a file 50 times" with "read off this range of memory in plasma".

 

I believe pretty strongly that a tangible example like this would meaningfully improve the R community's interest in and engagement with the Arrow project.