Run SAS programs in Databricks with Jenner

Jenner is an independent engine for the SAS language. It runs your DATA steps, PROCs and macros inside your Databricks workspace, on the tables you already have, so a move to Databricks does not have to start with a rewrite.

Can SAS code run in Databricks?

Yes, with Jenner. Install the Jenner package from a Unity Catalog volume, load it in a notebook, and put SAS code in a %%jenner cell next to your Python and SQL cells. Name the tables the program reads with --in and the tables it writes with --out. The same program can also run as a task in a Databricks job.

The language does not need translating. Most programs do need their file and library paths pointed at Databricks locations, because a program that reads /sasdata/sales has to be pointed at a volume or a table instead.

What runs, and where

Jenner is one engine. In Databricks it runs on your serverless compute, under your Unity Catalog permissions.

In a notebook

Put a SAS program in a %%jenner cell. The listing and the log appear below the cell, and WORK data sets carry over from one cell to the next for the rest of the session.

As a job

Run a .sas file as a task in a Databricks job with the jenner-run entry point, on the schedules you already use. Each run keeps its program, log and listing in a run folder on a volume.

On your Unity Catalog tables

--in reads a table, or a temporary view of a DataFrame, as a SAS data set. --out writes a data set to a Delta table: it replaces the rows, appends them, or refuses if the table exists. Columns keep the table's types, and nothing is cast silently.

Output on your volumes

A permanent SAS library is kept as Parquet files on a Unity Catalog volume. Each run's program, log, listing and output files are copied to the user's own storage on a volume, so they outlast the serverless machine.

DATA steps, PROC SQL, macros and 100 procedures

It is the same engine that runs on Linux, macOS and Windows, packaged for Databricks. Procedures that need R cannot run on serverless compute, which has no R; they stop with a clear error.

Your data stays in your workspace

Jenner runs on your Databricks compute and reads your tables under your Unity Catalog permissions, so your data stays in your workspace.

A private volume for each user

By default each user gets a volume of their own for their WORK library, their JENUSER library and the logs, listings and files of their runs. Only that user, your metastore admins and the owners of the catalog and schema can read it.

A shared folder, if you choose it

An administrator can choose one shared volume with a folder for each user instead. Every Jenner user can then read every other user's files, and the Jenner App says so before the switch.

The Jenner App handles the licence

A workspace administrator installs the Jenner App in Databricks. It shows the licence status, installs and renews the licence key on a volume, and sends a monthly usage report. To buy, the administrator opens the Jenner website from the App and pays by card.

What the App sends to Jenner

Your workspace ID and host, the cloud, the App version, the licence serial number, monthly counts of active users with salted hashes of their identities, and the administrator's email address at checkout. It does not send your tables or your SAS programs.

Jenner Studio is coming to Databricks. It is not available there yet.

Frequently Asked Questions

Install the Jenner package with %pip install from the Unity Catalog volume your administrator set up, run %load_ext jenner_dbx, and start a cell with %%jenner. Name input tables with --in and output tables with --out. The cell shows the listing and the log, and a failed program stops a Run all at that cell.

Yes. The Jenner package has a jenner-run entry point for a Python wheel task. Give it the .sas file with --program and the tables with --in and --out, and schedule the job as you schedule any other. Each run's program, log and listing are kept in a run folder on a volume.

No. Jenner is an independent implementation of the SAS language. It does not contain, wrap or call SAS software, so nothing in your workspace needs a SAS licence to run it.

Jenner runs on your Databricks compute and reads and writes your tables there, under your Unity Catalog permissions. The Jenner App sends us only what it needs for the licence and the bill: your workspace ID and host, the cloud, the App version, the licence serial number, monthly counts of active users with salted hashes of their identities, and the administrator's email address at checkout.

By default each user has a private volume, which only that user, your metastore admins and the owners of the catalog and schema can read. An administrator can switch to one shared volume with a folder per user, which every Jenner user can read; the Jenner App shows who can read the files before it switches.

The language does not need translating. Programs usually still need their paths changed, because a program that reads /sasdata/sales has to be pointed at a volume or a table instead. A program that calls operating-system commands, or needs software that serverless compute does not have, needs more work. Try a representative program before you plan around it.

Procedures that need R, such as GAMPL, cannot run on serverless compute, which has no R. Procedures backed by Python need Python 3.12, which means serverless environment 3 or later. We have tested on AWS; Azure, Google Cloud and classic clusters have not been tested.

Through the Jenner App. An administrator buys from the App, which opens the Jenner website, and pays by card. The App then installs the licence and renews it. Your Databricks compute, including the compute the App uses, is billed by Databricks as usual.

Not yet. Jenner Studio is coming to Databricks. Today Jenner runs there in notebooks and jobs, and the Jenner App handles the licence.

Often that is right. Spark on several workers is faster than Jenner on large data, and if your SAS programs are few and simple, a rewrite may be the better answer. The case for running them as they are is strongest when there are many, they are validated, and rewriting means re-validating rather than merely re-coding.

No. Jenner runs in Databricks, but this page is not a statement of partnership with, or endorsement by, Databricks.

Get Jenner for your Databricks workspace

The Jenner listing on Databricks Marketplace is on its way. Ask for early access and we will get back to you.

Request early access

Ask about Jenner on Databricks

Tell us roughly how many SAS programs you run, how often, and which procedures matter most.