polars-readstat 0.12.3

Read SAS (sas7bdat), Stata (dta), and SPSS (sav) files with polars

polars_readstat

Polars plugin for SAS (.sas7bdat), Stata (.dta), and SPSS (.sav/.zsav) files.

The Python package wraps the Rust core in polars_readstat_rs and exposes a simple Polars-first API. I have tried to make sure there are no errors or regressions in this release (tested against 178 test files from pandas, pyreadstat, etc.).

The new rust engine is on par or faster than the old for many files, but it's not always faster (at least for SAS data sets), so if it's slower or I missed a bug, you can find info on the prior version and install version 0.11.1 from pypi.

Why use this?

• In project benchmarks, the new Rust-backed engine is typically faster than pandas/pyreadstat on large SAS/Stata files, especially for subset/filter workloads.
• It avoids the older C/C++ toolchain complexity and ships as standard Python wheels.
• API is Polars-first (scan_readstat, read_readstat, write_readstat).

Install

pip install polars-readstat

Core API

1) Lazy scan

import polars as pl
from polars_readstat import scan_readstat

lf = scan_readstat("/path/file.sas7bdat", preserve_order=True)
df = lf.select(["SERIALNO", "AGEP"]).filter(pl.col("AGEP") >= 18).collect()

2) Eager read

from polars_readstat import read_readstat

df = read_readstat("/path/file.dta")

3) Metadata + schema

from polars_readstat import ScanReadstat

reader = ScanReadstat(path="/path/file.sav")
schema = reader.schema
metadata = reader.metadata

4) Write (Stata/SPSS) - EXPERIMENTAL

I can test reading the data back with Stata, as I have access to it, but I don't have access to SPSS. I can make sure my code roundtrips properly and I'll be adding read tests from other packages (pyreadstat and pandas) to make sure they can read the files I create, but I'll need help testing things from others before I'm comfortable with the SPSS code.

from polars_readstat import write_readstat

write_readstat(df, "/path/out.dta", threads=8)
write_readstat(df, "/path/out.sav")

write_readstat supports Stata (dta) and SPSS (sav). SAS writing is not supported.

Tests run

We’ve tried to test this thoroughly:

• Cross-library comparisons on the pyreadstat and pandas test data, checking results against polars-readstat==0.11.1, pyreadstat, and pandas.
• Stata/SPSS read/write roundtrip tests.
• Large-file read/write benchmark runs on real-world data (results below).

If you want to run the same checks locally, helper scripts and tests are in scripts/ and tests/.

Benchmark

For each file, I compared 4 different scenarios: 1) load the full file, 2) load a subset of columns (Subset:True), 3) filter to a subet of rows (Filter: True), 4) load a subset of columns and filter to a subset of rows (Subset:True, Filter: True).

Benchmark context:

• Machine: AMD Ryzen 7 8845HS (16 cores), 14 GiB RAM, Linux Mint 22
• Storage: external SSD
• Last run: August 31, 2025
• Version tested: polars-readstat 0.12 (new Rust engine) against polars-readstat 0.11.1 (prior C++ and C engines) and pandas and pyreadstat
• Method: wall-clock timings via Python time.time()

Compared to Pandas and Pyreadstat (using read_file_multiprocessing for parallel processing in Pyreadstat)

SAS

all times in seconds (speedup relative to pandas in parenthesis below each)

Library	Full File	Subset: True	Filter: True	Subset: True, Filter: True
polars_readstat New rust engine	0.90 (2.3×)	0.07 (29.4×)	1.23 (2.5×)	0.07 (29.9×)
polars_readstat engine="cpp" (fastest for 0.11.1)	1.31 (1.6×)	0.09 (22.9×)	1.56 (1.9×)	0.09 (23.2×)
pandas	2.07	2.06	3.03	2.09
pyreadstat	10.75 (0.2×)	0.46 (4.5×)	11.93 (0.3×)	0.50 (4.2×)

Stata

all times in seconds (speedup relative to pandas in parenthesis below each)

Library	Full File	Subset: True	Filter: True	Subset: True, Filter: True
polars_readstat New rust engine	0.17 (6.7×)	0.12 (9.8×)	0.24 (4.1×)	0.11 (8.7×)
polars_readstat engine="readstat" (the only option for 0.11.1)	1.80 (0.6×)	0.27 (4.4×)	1.31 (0.8×)	0.29 (3.3×)
pandas	1.14	1.18	0.99	0.96
pyreadstat	7.46 (0.2×)	2.18 (0.5×)	7.66 (0.1×)	2.24 (0.4×)

Detailed benchmark notes and dataset descriptions are in BENCHMARKS.md.