sasonr

Run SAS® programs on an R engine.

sasonr parses SAS source code (DATA step, macro language, PROC SQL and a growing set of procedures) into a faithful abstract syntax tree with a hand-written lexer, recursive-descent parser and Pratt expression parser, then executes it in R: DATA steps are compiled into tidyverse code, PROC SQL runs on DuckDB. The goal is end-to-end results identical to SAS 9.4.

Three ways to use it:

1. Run SAS on R - execute SAS programs without a SAS licence.
2. Deterministic SAS to R migration - sas_translate() produces R code without any language model, and the result is verified against SAS output.
3. R / SAS double programming - one programmer writes R, the other writes SAS and runs it with sasonr on the same platform.

Design rules:

• Deterministic: no language model is involved at run time.
• Fail loud: any language feature that has not been verified against real SAS output raises a sasonr_unsupported condition with the SAS line number instead of guessing.
• Verified against SAS: features are accepted only after their test cases match output produced by SAS 9.4 (datasets, variable attributes, log diagnostics and listing).

Status: early development, not on CRAN.

# pak::pak("kaipingyang/sasonr")
library(sasonr)
res <- sas_run("data a; do i = 1 to 3; x = i**2; output; end; run;")
sas_translate("data b; set a; if x > 1; run;")   # show generated R code

SAS is a registered trademark of SAS Institute Inc. This project is not affiliated with or endorsed by SAS Institute Inc.