R on Snowflake: The Piste guide for R users
R users shouldn't need a separate R server or a Python rewrite to work with Snowflake. The Piste Guide to R and Snowflake is an open guide that shows you how to connect from RStudio, Posit, VS Code, or Snowflake Workspace Notebooks, query data with familiar dplyr and DBI code, and take a model from features through training to the registry.
We built it around two open-source packages, skiLift and skiPatrol. Start with the Snowflake basics or jump straight to the chapter you need. The guide lives on GitHub, so you can follow along, copy the code, and report issues as you go.
WHAT'S INSIDE
- Snowflake platform primer. Warehouses, containers, and ML services, explained for R users who are new to Snowflake.
- R from your IDE. Set up RStudio, Posit, or VS Code to work with Snowflake, including RStudio Server on Snowpark Container Services.
- skiLift. Connect and query with DBI, dbplyr, and the Snowflake SQL API.
- skiPatrol. Run MLOps from R: Feature Store, training data, and Model Registry.
- Appendices. Setup, configuration reference, troubleshooting, a production checklist, package reference, and a glossary.
WHO IS THIS GUIDE FOR
- R users and data scientists who want to work on Snowflake without leaving R.
- Teams that use both R and Python and want to share features, datasets, and models on one platform.
- Platform and IT teams who support R users on Snowflake.
PREREQUISITES
- A Snowflake account
- R 4.2 or later
- Python 3.10 or later for local skiPatrol use.
- You should be comfortable writing R and reading SQL.
Frequently asked questions
What is the guide to R and Snowflake?
The Piste Guide to R and Snowflake is an end-to-end guide for R users who want to run data analysis, modeling, and ML workflows on Snowflake. It covers Snowflake basics, working from your own IDE, connecting and querying with skiLift, and building ML workflows with skiPatrol.
Can I use R on Snowflake without writing Python?
Yes. skiPatrol wraps Snowflake's Python ML SDK for you, so you work in R. You still need Python 3.10 or later installed for local use. In Workspace Notebooks, R runs in cells that use the %%R magic.
What are skiLift and skiPatrol?
skiLift is a DBI driver for Snowflake that works with dbplyr and Arrow. skiPatrol is an R interface to Snowflake's ML platform, including Feature Store and Model Registry. Both are open source.
Which tools can I use to work with R on Snowflake?
You can work from RStudio, Posit, or VS Code on your laptop, or from Snowflake Workspace Notebooks in Snowsight. The guide's introduction helps you choose a starting path.
What do I need to get started?
A Snowflake account, R 4.2 or later, and Python 3.10 or later for local skiPatrol use. For Workspace Notebooks, you also need a role with warehouse usage and notebook execution.
Are skiLift and skiPatrol officially supported by Snowflake?
No. They are community-maintained open-source packages, not officially supported Snowflake product offerings, and their APIs may change. skiPatrol relies on Snowflake's supported snowflake-ml-python SDK.
Can I deploy R models on Snowflake?
The guide covers taking a model from features through training to the Model Registry. Registered R models are served on Snowflake through Snowpark Container Services, so you train in R and serve in Snowflake.