SAS to Iceberg

Architecture

SAS in. Iceberg out.

Deterministic parsers read the SAS estate and emit native Iceberg code — not a proprietary lakehouse lock-in.

SAS programs → SAS2PY parser → Iceberg tables

SAS

Base SAS DATA step / macros

DI Studio Jobs + mappings

EG / EM Projects + flows

Viya / CAS CASL + actions

SAS2PY Parser

Deterministic parse AI optional

Row-level parity Before cutover

Table emit Create + write

Job emit Spark or Trino

Iceberg

Iceberg Open tables

Spark / Trino Your engine

Object store S3 / ADLS / GCS

Git DDL in review

AI is an optional add-on, off by default — the conversion runs end to end without it, air-gapped if your estate requires it.

Why Iceberg

The table format should outlive the engine

SAS datasets are a private format

Iceberg is an open spec. Spark, Trino, Flink, and Snowflake can all read what we write.

IMAGE copies are not time travel

Iceberg snapshots replace six generations of SAS backup datasets.

A dump to Parquet is not a modernization

The DATA step has to become SQL or Spark. The table format is the contract.

Parser output

SAS dataset to an Iceberg table

A SAS DATA step write — emitted as an Iceberg CREATE + INSERT.

SAS

/* SAS */
data dw.gold;
  set dw.txn;
  if amount > 1000;
run;

→

SAS2PY
converts

Spark + Iceberg

# SAS dataset → Iceberg
(spark.table("txn")
      .filter("amount > 1000")
      .writeTo("dw.gold")
      .using("iceberg")
      .createOrReplace())

The SAS library becomes a catalog namespace. Snapshots replace IMAGE copies.

Coverage

SAS to Iceberg — artifact mapping

SAS	Iceberg	Notes
SAS dataset	Iceberg table	Open format
DATA step / SQL	Spark or Trino SQL	Engine of your choice
IMAGE copies	Snapshots	Time travel
SAS indexes	Partition specs	Pruning

Validation

Every conversion validated to row-level parity

SAS output compared to Iceberg output — row by row, column by column. Differences flagged before sign-off.

See how Data Matching works →

6,400

jobs

5.4M

LOC

$9.2M

saved

3

catalogs

Manufacturer: 6,400 SAS jobs to Iceberg on three clouds

Three SAS Grids, one Iceberg spec: Glue+S3, ADLS+Hive Metastore, BigLake+Dataproc. Same modules, three catalog adapters.

Read the case study →

← SAS2PY Home SAS → PySpark SAS → AWS Case studies

Convert SAS programs to Iceberg tables