Build Status R Package for WebHDFS REST API


In this package, most code is same with original rwebhdfs package. But I modify the "http" address to "https" address and change token to delagation token for user to be easier to use.

Additional function added: read_all() to allow users to load whole directory files into variable

This R package provides access to HDFS via WebHDFS REST API. For more information, please see:

Hadoop Configuration

Ensure that WebHDFS is enabled in the hdfs-site.xml


How to Use

More exmaples will arrive in the function help pages but for now, here's a brief guide on how to use rwebhdfs


I'm recommend HDP 2.0 for quick demo and testing:

Create your webhdfs object(token should be a long string of your delegation token)

WebHDFS is a S3 object and can be created using

hdfs <- webhdfs("localhost", 50070, "hue",token="your delegation token")

List the files under you home directory

dir_stat(hdfs, "")

Creates an empty file named "test" and get its information

write_file(hdfs, "test")
file_stat(hdfs, "test")

Reading whole directory files into one single variable

data <- read_all(hdfs, "dirPath")

Write local file onto HDFS and see what we just wrote

foo <- tempfile()
writeLines("foobar", foo)
write_file(hdfs, "foo", foo)
read_file(hdfs, "foo")

Creates a directory and move our file in it

mkdir(hdfs, "bar")
rename_file(hdfs, "foo", "bar/foo")

Finally delete the test file and folder

delete_file(hdfs, "test")
delete_file(hdfs, "bar", recursive=TRUE)

How to Install

rwebhdfs is not on CRAN yet. I plan to play with it in a couple Hadoop projects before submission to CRAN. So that I can decide if all functions are intuitive and well designed.

To get latest version on Github:



webhdfs has been implemented as a S3 object and all common FileSystem related functions are coded as S3 methods. Since R provides some basic FileSystem functions like list.files,, read.*, write.* and etc, I try to name my functions in a similar logic but easy to find using auto-completion when actually typing. So you will find functions like write_file, file_stat, rename_file and etc.

It seems that in Hadoop itself, WebHDFS has been implemented as a subclass of FileSystem and there are a lot of others like FTP, S3 and (regular) HDFS that extend to this interface. I think it would be awesome if we do the same in R so data can be fetched/stored in a more transparent way from different FileSystem.

Discussion is more than welcomed on design decisions and choice on OO System. I have zero experience on OO programing in R and chose S3 based on the suggestions here:


Both Kerberos and delegation token security are implemented. Use the securityON flag in webhdfs constructor to enable security, if in addition token is also supplied then delegation token will be used, otherwise Kerberos is assumed. However, I have not tested this feature just yet. Please report any issues you see.

linz1112/rwebhdfs-fix documentation built on May 21, 2019, 6:39 a.m.