Derive Randomization Timing

1library(yamaa)23yamaa_spec("spec.yaml")$run()$output

1import yamaa23yamaa.spec("spec.yaml").run().output

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 1.67"""Create the ADSL randomization dataset from the DM and ODM datasets.89One record per subject with columns:10STUDYID, USUBJID, RANDDT, RANDDY, RANDDTC1112- RANDDT is the subject's randomization date from DM.13- RANDDY is the study day of RANDDT measured from RFSTDTC (the subject's14  reference start date): there is no study day 0, so a randomization on15  RFSTDTC is day 1 and the day before RFSTDTC is day -1. RANDDY has no16  value when RANDDT or RFSTDTC has no value.17- RANDDTC is the collected randomization moment RANDDTTM written as text18  at whole-second precision (date and time joined by T, seconds always19  shown). It is written independently of the collected date, so a20  subject with a collected moment but no collected date still has21  RANDDTC while RANDDT and RANDDY have no value.22"""2324import polars as pl2526INPUT_DIR = "/app/input"27OUTPUT_DIR = "/app/output"282930def check_odm_agrees_with_dm(dm: pl.DataFrame) -> None:31    """Confirm the ODM collected DM values agree with the DM dataset."""32    odm = pl.read_csv(f"{INPUT_DIR}/odm.csv", infer_schema=False)33    items = odm.filter(pl.col("ItemGroupOID") == "IG.DM").pivot(34        values="Value", index="SubjectKey", on="ItemOID"35    )36    for item, column, fmt in (37        ("IT.DM.RFSTDTC", "RFSTDTC", "%Y-%m-%d"),38        ("IT.DM.RANDDT", "RANDDT", "%Y-%m-%d"),39        ("IT.DM.RANDDTTM", "RANDDTTM", "%Y-%m-%d %H:%M:%S"),40    ):41        if item not in items.columns:42            if dm[column].is_not_null().any():43                raise ValueError(f"ODM has no {item} but DM has {column}")44            continue45        collected = items.select(46            pl.col("SubjectKey").alias("USUBJID"),47            pl.col(item).alias("collected"),48        )49        joined = dm.join(collected, on="USUBJID", how="left")50        typed_text = joined[column].dt.strftime(fmt)51        collected_text = joined["collected"]52        if column == "RANDDTTM":53            # Collected moments may carry only minute precision54            # ("2024-03-04T10:30"); normalize to whole seconds.55            collected_text = (56                pl.when(collected_text.str.len_chars() == 16)57                .then(collected_text.str.replace("T", " ") + ":00")58                .otherwise(collected_text.str.replace("T", " "))59            )60        mismatch = joined.filter(61            typed_text.is_not_null()62            & collected_text.is_not_null()63            & (typed_text != collected_text)64        )65        if mismatch.height:66            raise ValueError(67                f"ODM collected {column} disagrees with DM for "68                f"{mismatch['USUBJID'].to_list()}"69            )707172def main() -> None:73    dm = pl.read_parquet(f"{INPUT_DIR}/dm.parquet")74    check_odm_agrees_with_dm(dm)7576    diff_days = (pl.col("RANDDT") - pl.col("RFSTDTC")).dt.total_days()77    adsl = dm.select(78        "STUDYID",79        "USUBJID",80        "RANDDT",81        pl.when(pl.col("RANDDT").is_null() | pl.col("RFSTDTC").is_null())82        .then(None)83        .when(diff_days >= 0)84        .then(diff_days + 1)85        .otherwise(diff_days)86        .alias("RANDDY"),87        pl.col("RANDDTTM").dt.strftime("%Y-%m-%dT%H:%M:%S").alias("RANDDTC"),88    )8990    adsl.write_csv(f"{OUTPUT_DIR}/adsl.csv")919293if __name__ == "__main__":94    main()

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 v0.7.1 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;