Derive Randomization Timing

1library(yamaa)23adsl <- yamaa_domain("spec.yaml")$output4adsl

1import yamaa23adsl = yamaa.yamaa_domain("spec.yaml").output4adsl

1# Reference solution for the yamaa benchmark adam-adsl-randomization (R track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67library(dplyr, warn.conflicts = FALSE)8library(readr)910dm <- nanoparquet::read_parquet("/app/input/dm.parquet")1112# Study day from the reference start: the reference day itself is day 1,13# later dates count forward inclusively, earlier dates count backward with14# no day 0, so the day before is day -1.15adsl <- dm |>16  mutate(17    .diff = as.integer(RANDDT - RFSTDTC),18    RANDDY = case_when(19      is.na(.diff) ~ NA_integer_,20      .diff >= 0L ~ .diff + 1L,21      .default = .diff22    ),23    # The collected moment as ISO 8601 text with a T, at whole-second24    # precision, written from what was collected even when no date was.25    RANDDTC = format(RANDDTTM, "%Y-%m-%dT%H:%M:%S", tz = "UTC")26  ) |>27  select(STUDYID, USUBJID, RANDDT, RANDDY, RANDDTC) |>28  arrange(USUBJID)2930dir.create("/app/output", showWarnings = FALSE)31write_csv(adsl, "/app/output/adsl.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adsl-randomization (Python track).2#3# Written from the benchmark's full prompt and its inputs alone, as an agent4# would write it. Harbor's oracle agent runs it to check that the prompt can5# be solved and that the grader scores a correct answer 1.67from pathlib import Path89import polars as pl1011dm = pl.read_parquet("/app/input/dm.parquet")1213# Study day from the reference start: the reference day itself is day 1,14# later dates count forward inclusively, earlier dates count backward with15# no day 0, so the day before is day -1.16diff = (pl.col("RANDDT") - pl.col("RFSTDTC")).dt.total_days()17adsl = (18    dm.with_columns(19        RANDDY=pl.when(diff.is_null())20        .then(None)21        .when(diff >= 0)22        .then(diff + 1)23        .otherwise(diff),24        # The collected moment as ISO 8601 text with a T, at whole-second25        # precision, written from what was collected even when no date was.26        RANDDTC=pl.col("RANDDTTM").dt.strftime("%Y-%m-%dT%H:%M:%S"),27    )28    .select("STUDYID", "USUBJID", "RANDDT", "RANDDY", "RANDDTC")29    .sort("USUBJID")30)3132Path("/app/output").mkdir(parents=True, exist_ok=True)33adsl.write_csv("/app/output/adsl.csv", null_value="")

1/* opensas reference solution for the yamaa benchmark adam-adsl-randomization.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45/* opensas v0.6.6 has no Parquet reader. The runner must stage dm.parquet6   as dm.csv without deriving any values, preserving ISO dates/timestamps. */7data dm;8  length STUDYID USUBJID $1024 RANDDT RFSTDTC $10 RANDDTTM $32;9  infile '/app/input/dm.csv' dsd dlm=',' firstobs=2 truncover;10  input STUDYID : $char1024. USUBJID : $char1024. RFSTDTC : $char10. RANDDT : $char10. RANDDTTM : $char32.;11run;12data result;13  length RANDDTC $19;14  set dm;15  if not missing(RANDDT) and not missing(RFSTDTC) then do;16    RANDDY=input(RANDDT,yymmdd10.)-input(RFSTDTC,yymmdd10.);17    if RANDDY>=0 then RANDDY=RANDDY+1;18  end;19  if not missing(RANDDTTM) then do;20    RANDDTC=translate(substr(RANDDTTM,1,19),'T',' ');21    if lengthn(RANDDTTM)=16 then RANDDTC=cats(RANDDTTM,':00');22  end;23run;2425proc sql;26  create table final as select STUDYID, USUBJID, RANDDT, RANDDY, RANDDTC from result;27quit;28proc export data=final outfile="/app/output/adsl.csv" dbms=csv replace; run;