Derive planned period treatments and sequence from the trial design

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

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

1# Reference solution for the yamaa benchmark2# adam-adsl-crossover-sequence (R track).3#4# Written from the benchmark's full prompt and its inputs alone, as an agent5# would write it. Harbor's oracle agent runs it to check that the prompt can6# be solved and that the grader scores a correct answer 1.78library(dplyr, warn.conflicts = FALSE)9library(readr)1011dm <- read_csv(12  "/app/input/dm.csv",13  col_types = cols(.default = col_character())14)15ta <- read_csv(16  "/app/input/ta.csv",17  col_types = cols(TAETORD = col_integer(), .default = col_character())18)19te <- read_csv(20  "/app/input/te.csv",21  col_types = cols(.default = col_character())22)2324# The treatment elements of each arm, numbered in planned order: only the25# elements in an epoch whose name begins with TREATMENT are counted.26ranked <- ta |>27  filter(startsWith(EPOCH, "TREATMENT")) |>28  arrange(STUDYID, ARMCD, TAETORD) |>29  group_by(STUDYID, ARMCD) |>30  mutate(TRTNUM = row_number()) |>31  ungroup() |>32  left_join(33    te |> select(STUDYID, ETCD, ELEMENT),34    by = c("STUDYID", "ETCD")35  )3637period1 <- ranked |>38  filter(TRTNUM == 1L) |>39  transmute(STUDYID, ARMCD, TRT01P = ELEMENT)40period2 <- ranked |>41  filter(TRTNUM == 2L) |>42  transmute(STUDYID, ARMCD, TRT02P = ELEMENT)4344adsl <- dm |>45  left_join(period1, by = c("STUDYID", "ARMCD")) |>46  left_join(period2, by = c("STUDYID", "ARMCD")) |>47  mutate(48    TRTSEQP = if_else(49      !is.na(TRT01P) & !is.na(TRT02P),50      paste(TRT01P, TRT02P, sep = "/"),51      NA_character_52    )53  ) |>54  select(STUDYID, USUBJID, TRT01P, TRT02P, TRTSEQP)5556dir.create("/app/output", showWarnings = FALSE)57write_csv(adsl, "/app/output/adsl.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adsl-crossover-sequence (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 ADSL planned treatment sequence from DM, TA and TE.89One record per subject in DM. TRT01P/TRT02P are the names (from TE) of the10first and second treatment elements of the subject's planned arm (DM ARMCD),11taking the arm's elements from TA in TAETORD order and counting only12elements whose epoch name begins with "TREATMENT". TRTSEQP joins the two13with "/". Subjects with no planned arm code get no values.14"""1516import csv17import os1819INPUT_DIR = "/app/input"20OUTPUT_DIR = "/app/output"212223def read_csv(name):24    with open(os.path.join(INPUT_DIR, name), newline="", encoding="utf-8") as f:25        return list(csv.DictReader(f))262728def main():29    dm = read_csv("dm.csv")30    ta = read_csv("ta.csv")31    te = read_csv("te.csv")3233    # Element code -> element name, per study.34    element_name = {}35    for row in te:36        element_name[(row["STUDYID"], row["ETCD"])] = row["ELEMENT"]3738    # Planned arm -> ordered treatment element codes, per study.39    arm_elements = {}40    for row in ta:41        if not row["EPOCH"].startswith("TREATMENT"):42            continue43        key = (row["STUDYID"], row["ARMCD"])44        arm_elements.setdefault(key, []).append(45            (int(row["TAETORD"]), row["ETCD"])46        )47    for key in arm_elements:48        arm_elements[key] = [49            etcd for _, etcd in sorted(arm_elements[key])50        ]5152    out_rows = []53    for subj in dm:54        studyid = subj["STUDYID"]55        armcd = subj.get("ARMCD", "")56        trt01p = trt02p = trtseqp = ""57        if armcd:58            codes = arm_elements.get((studyid, armcd), [])59            names = [element_name.get((studyid, c), "") for c in codes]60            if len(names) >= 1:61                trt01p = names[0]62            if len(names) >= 2:63                trt02p = names[1]64            parts = [n for n in (trt01p, trt02p) if n]65            if parts:66                trtseqp = "/".join(parts)67        out_rows.append(68            {69                "STUDYID": studyid,70                "USUBJID": subj["USUBJID"],71                "TRT01P": trt01p,72                "TRT02P": trt02p,73                "TRTSEQP": trtseqp,74            }75        )7677    os.makedirs(OUTPUT_DIR, exist_ok=True)78    fieldnames = ["STUDYID", "USUBJID", "TRT01P", "TRT02P", "TRTSEQP"]79    with open(80        os.path.join(OUTPUT_DIR, "adsl.csv"), "w", newline="", encoding="utf-8"81    ) as f:82        writer = csv.DictWriter(f, fieldnames=fieldnames)83        writer.writeheader()84        writer.writerows(out_rows)858687if __name__ == "__main__":88    main()

1/* opensas reference solution for the yamaa benchmark adam-adsl-crossover-sequence.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data dm;6  length STUDYID USUBJID ARMCD ARM ACTARMCD ACTARM ARMNRS $1024;7  infile "/app/input/dm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. ARMCD : $char1024. ARM : $char1024. ACTARMCD : $char1024. ACTARM : $char1024. ARMNRS : $char1024.;9run;1011data ta;12  length STUDYID ARMCD ARM ETCD EPOCH $1024;13  infile "/app/input/ta.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;14  input STUDYID : $char1024. ARMCD : $char1024. ARM : $char1024. TAETORD ETCD : $char1024. EPOCH : $char1024.;15run;1617data te;18  length STUDYID ETCD ELEMENT TESTRL TEDUR $1024;19  infile "/app/input/te.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;20  input STUDYID : $char1024. ETCD : $char1024. ELEMENT : $char1024. TESTRL : $char1024. TEDUR : $char1024.;21run;2223/* Only elements in an epoch whose name begins with TREATMENT count, taken24   in planned order within each arm. */25data treatment;26  set ta;27  where substr(EPOCH, 1, 9) = 'TREATMENT';28run;29proc sort data=treatment;30  by STUDYID ARMCD TAETORD;31run;32data ranked;33  set treatment;34  by STUDYID ARMCD;35  if first.ARMCD then TRTNUM = 0;36  TRTNUM + 1;37run;3839proc sql;40  create table periods as41    select r.STUDYID, r.ARMCD, r.TRTNUM, t.ELEMENT42    from ranked as r43    left join te as t on r.STUDYID = t.STUDYID and r.ETCD = t.ETCD;44  create table period1 as45    select STUDYID, ARMCD, ELEMENT as TRT01P from periods where TRTNUM = 1;46  create table period2 as47    select STUDYID, ARMCD, ELEMENT as TRT02P from periods where TRTNUM = 2;48  create table joined as49    select a.STUDYID, a.USUBJID, b.TRT01P, c.TRT02P50    from dm as a51    left join period1 as b on a.STUDYID = b.STUDYID and a.ARMCD = b.ARMCD52    left join period2 as c on a.STUDYID = c.STUDYID and a.ARMCD = c.ARMCD;53quit;5455data result;56  length TRT01P TRT02P TRTSEQP $1024;57  set joined;58  if not missing(TRT01P) and not missing(TRT02P) then59    TRTSEQP = trim(TRT01P) || '/' || trim(TRT02P);60run;6162proc sql;63  create table final as select STUDYID, USUBJID, TRT01P, TRT02P, TRTSEQP from result;64quit;65proc export data=final outfile="/app/output/adsl.csv" dbms=csv replace; run;