Read data from Amazon Redshift into Spark DataFrames

What is sparkredshift?

sparkredshipt is an extension for sparklyr to read data from Amazon Redshift into Spark DataFrames. It uses the Spark package spark-redshift to load redshift data into Spark DataFrames.

Installation

sparkredshift requires the sparklyr package to run

Install sparklyr

I recommend the latest stable version of sparklyr available on CRAN

install.packages("sparklyr")

Install sparkredshift

Install the development version of sparkredshift from this Github repo using devtools

library(devtools)
devtools::install_github("emaasit/sparkredshift")

Connecting to Spark

If Spark is not already installed, use the following sparklyr command to install your preferred version of Spark:

library(sparklyr)
spark_install(version = "2.0.0")

The call to library(sparkredshift) will make the sparkredshift functions available on the R search path and will also ensure that the dependencies required by the package are included when we connect to Spark.

library(sparkredshift) 

We can create a Spark connection as follows:

sc <- spark_connect(master = "local")

Reading redshift files

sparkredshift provides the function spark_read_redshift to read redshift data files into Spark DataFrames. It uses a Spark package called spark-redshift. Here's an example.

mtcars_file <- system.file("extdata", "mtcars.redshift", package = "sparkredshift")

mtcars_df <- spark_read_redshift(sc, path = mtcars_file, table = "redshift_table")
mtcars_df

The resulting pointer to a Spark table can be further used in dplyr statements.

library(dplyr)
mtcars_df %>% group_by(cyl) %>%
  summarise(count = n(), avg.mpg = mean(mpg), avg.displacment = mean(disp), avg.horsepower = mean(hp))

Logs & Disconnect

Look at the Spark log from R:

spark_log(sc, n = 100)

Now we disconnect from Spark:

spark_disconnect(sc)

Acknowledgements

Thanks to RStudio for the sparklyr package that provides functionality to create Extensions.