Derive absolute differential records

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

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

1# Reference solution for the yamaa benchmark adam-adlb-lymphocytes (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)910raw <- read_csv(11  "/app/input/lb.csv",12  col_types = cols(LBSTRESN = col_double(), .default = col_character())13) |>14  rename(PARAMCD = LBTESTCD, AVAL = LBSTRESN, PARAM = LBTEST)1516# Keep every collected record unchanged, with DTYPE empty.17collected <- raw |>18  select(USUBJID, PARAMCD, AVAL, PARAM, VISIT) |>19  mutate(DTYPE = NA_character_)2021# Differential fraction code -> absolute code and parameter name.22differentials <- list(23  LYMLE = c("LYMPH", "Lymphocytes Abs (10^9/L)"),24  NEUTLE = c("NEUT", "Neutrophils Abs (10^9/L)"),25  MONOLE = c("MONO", "Monocytes Abs (10^9/L)"),26  EOSLE = c("EOS", "Eosinophils Abs (10^9/L)"),27  BASOLE = c("BASO", "Basophils Abs (10^9/L)")28)2930wbc <- raw |>31  filter(PARAMCD == "WBC") |>32  select(USUBJID, VISIT, WBCVAL = AVAL)3334# One absolute record per subject and visit for each differential35# fraction present with a WBC count and no absolute record yet.36calc <- bind_rows(lapply(names(differentials), function(frac_code) {37  abs_code <- differentials[[frac_code]][1]38  abs_param <- differentials[[frac_code]][2]39  frac <- raw |>40    filter(PARAMCD == frac_code) |>41    select(USUBJID, VISIT, FRACVAL = AVAL)42  existing <- raw |>43    filter(PARAMCD == abs_code) |>44    distinct(USUBJID, VISIT) |>45    mutate(HAS_ABS = TRUE)46  wbc |>47    inner_join(frac, by = c("USUBJID", "VISIT")) |>48    left_join(existing, by = c("USUBJID", "VISIT")) |>49    filter(!is.na(WBCVAL), !is.na(FRACVAL), is.na(HAS_ABS)) |>50    mutate(51      PARAMCD = abs_code,52      AVAL = WBCVAL * FRACVAL,53      PARAM = abs_param,54      DTYPE = "CALCULATION"55    ) |>56    select(USUBJID, PARAMCD, AVAL, PARAM, VISIT, DTYPE)57}))5859adlb <- bind_rows(collected, calc)6061dir.create("/app/output", showWarnings = FALSE)62write_csv(adlb, "/app/output/adlb.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adlb-lymphocytes (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 ADLB from LB: collected records plus calculated absolute8differential counts (WBC count times the matching differential fraction)9where the absolute parameter is not already present."""1011import csv12import os1314INPUT_DIR = "/app/input"15OUTPUT_DIR = "/app/output"1617# differential fraction test code -> (absolute parameter code, PARAM name)18DIFFERENTIALS = [19    ("LYMLE", "LYMPH", "Lymphocytes Abs (10^9/L)"),20    ("NEUTLE", "NEUT", "Neutrophils Abs (10^9/L)"),21    ("MONOLE", "MONO", "Monocytes Abs (10^9/L)"),22    ("EOSLE", "EOS", "Eosinophils Abs (10^9/L)"),23    ("BASOLE", "BASO", "Basophils Abs (10^9/L)"),24]252627def format_number(value):28    text = f"{value:.10f}".rstrip("0").rstrip(".")29    return text if text not in ("", "-0") else "0"303132def main():33    with open(os.path.join(INPUT_DIR, "lb.csv"), newline="") as handle:34        lb_rows = list(csv.DictReader(handle))3536    collected = []37    wbc = {}        # (subject, visit) -> WBC count (float)38    fractions = {}  # (subject, visit, fraction code) -> fraction (float)39    existing = set()  # (subject, visit, paramcd) among collected records40    group_order = []  # (subject, visit) in first-appearance order4142    for row in lb_rows:43        subject = row["USUBJID"]44        visit = row["VISIT"]45        testcd = row["LBTESTCD"]46        key = (subject, visit)47        if key not in group_order:48            group_order.append(key)49        value_text = row["LBSTRESN"].strip()50        value = float(value_text) if value_text else None5152        collected.append(53            {54                "USUBJID": subject,55                "PARAMCD": testcd,56                "AVAL": value_text,57                "PARAM": row["LBTEST"],58                "VISIT": visit,59                "DTYPE": "",60            }61        )62        existing.add((subject, visit, testcd))63        if value is None:64            continue65        if testcd == "WBC":66            wbc.setdefault(key, value)67        elif any(testcd == frac for frac, _, _ in DIFFERENTIALS):68            fractions.setdefault((subject, visit, testcd), value)6970    added = []71    for subject, visit in group_order:72        key = (subject, visit)73        if key not in wbc:74            continue75        for frac_code, abs_code, abs_name in DIFFERENTIALS:76            frac_key = (subject, visit, frac_code)77            if frac_key not in fractions:78                continue79            if (subject, visit, abs_code) in existing:80                continue81            added.append(82                {83                    "USUBJID": subject,84                    "PARAMCD": abs_code,85                    "AVAL": format_number(wbc[key] * fractions[frac_key]),86                    "PARAM": abs_name,87                    "VISIT": visit,88                    "DTYPE": "CALCULATION",89                }90            )9192    os.makedirs(OUTPUT_DIR, exist_ok=True)93    fieldnames = ["USUBJID", "PARAMCD", "AVAL", "PARAM", "VISIT", "DTYPE"]94    with open(95        os.path.join(OUTPUT_DIR, "adlb.csv"), "w", newline=""96    ) as handle:97        writer = csv.DictWriter(handle, fieldnames=fieldnames)98        writer.writeheader()99        writer.writerows(collected + added)100101102if __name__ == "__main__":103    main()

1/* opensas reference solution for the yamaa benchmark adam-adlb-lymphocytes.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data lb;6  length USUBJID LBTESTCD LBTEST VISIT $1024;7  infile "/app/input/lb.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input USUBJID : $char1024. LBTESTCD : $char1024. LBSTRESN LBTEST : $char1024. VISIT : $char1024.;9run;1011proc sql; create table derived as12select a.USUBJID,'LYMPH' as PARAMCD,a.LBSTRESN*b.LBSTRESN as AVAL,'Lymphocytes Abs (10^9/L)' as PARAM,a.VISIT,'CALCULATION' as DTYPE from lb as a inner join lb as b on a.USUBJID=b.USUBJID and a.VISIT=b.VISIT left join lb as c on a.USUBJID=c.USUBJID and a.VISIT=c.VISIT and c.LBTESTCD='LYMPH' where a.LBTESTCD='LYMLE' and b.LBTESTCD='WBC' and not missing(a.LBSTRESN) and not missing(b.LBSTRESN) and missing(c.LBTESTCD)13union all14select a.USUBJID,'NEUT' as PARAMCD,a.LBSTRESN*b.LBSTRESN as AVAL,'Neutrophils Abs (10^9/L)' as PARAM,a.VISIT,'CALCULATION' as DTYPE from lb as a inner join lb as b on a.USUBJID=b.USUBJID and a.VISIT=b.VISIT left join lb as c on a.USUBJID=c.USUBJID and a.VISIT=c.VISIT and c.LBTESTCD='NEUT' where a.LBTESTCD='NEUTLE' and b.LBTESTCD='WBC' and not missing(a.LBSTRESN) and not missing(b.LBSTRESN) and missing(c.LBTESTCD)15union all16select a.USUBJID,'MONO' as PARAMCD,a.LBSTRESN*b.LBSTRESN as AVAL,'Monocytes Abs (10^9/L)' as PARAM,a.VISIT,'CALCULATION' as DTYPE from lb as a inner join lb as b on a.USUBJID=b.USUBJID and a.VISIT=b.VISIT left join lb as c on a.USUBJID=c.USUBJID and a.VISIT=c.VISIT and c.LBTESTCD='MONO' where a.LBTESTCD='MONOLE' and b.LBTESTCD='WBC' and not missing(a.LBSTRESN) and not missing(b.LBSTRESN) and missing(c.LBTESTCD)17union all18select a.USUBJID,'EOS' as PARAMCD,a.LBSTRESN*b.LBSTRESN as AVAL,'Eosinophils Abs (10^9/L)' as PARAM,a.VISIT,'CALCULATION' as DTYPE from lb as a inner join lb as b on a.USUBJID=b.USUBJID and a.VISIT=b.VISIT left join lb as c on a.USUBJID=c.USUBJID and a.VISIT=c.VISIT and c.LBTESTCD='EOS' where a.LBTESTCD='EOSLE' and b.LBTESTCD='WBC' and not missing(a.LBSTRESN) and not missing(b.LBSTRESN) and missing(c.LBTESTCD)19union all20select a.USUBJID,'BASO' as PARAMCD,a.LBSTRESN*b.LBSTRESN as AVAL,'Basophils Abs (10^9/L)' as PARAM,a.VISIT,'CALCULATION' as DTYPE from lb as a inner join lb as b on a.USUBJID=b.USUBJID and a.VISIT=b.VISIT left join lb as c on a.USUBJID=c.USUBJID and a.VISIT=c.VISIT and c.LBTESTCD='BASO' where a.LBTESTCD='BASOLE' and b.LBTESTCD='WBC' and not missing(a.LBSTRESN) and not missing(b.LBSTRESN) and missing(c.LBTESTCD); quit;21data collected; length PARAMCD $16 PARAM $1024 DTYPE $11; set lb; PARAMCD=LBTESTCD; PARAM=LBTEST; AVAL=LBSTRESN; run;22data result; set collected derived; run;2324proc sql;25  create table final as select USUBJID, PARAMCD, AVAL, PARAM, VISIT, DTYPE from result;26quit;27proc export data=final outfile="/app/output/adlb.csv" dbms=csv replace; run;