Actual Times Since the First Dose and Since the Reference Dose

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

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

1# Reference solution for the yamaa benchmark adam-adpc-relative-times (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(lubridate, warn.conflicts = FALSE)9library(readr)1011pc <- read_csv(12  "/app/input/pc.csv",13  col_types = cols(PCSEQ = col_integer(), .default = col_character())14)15ex <- read_csv("/app/input/ex.csv", col_types = cols(.default = col_character()))1617# A date without a time is not a moment; a time without seconds counts as18# second zero. Clock readings, no time zone.19moment <- function(x) {20  full <- if_else(nchar(x) == 16, paste0(x, ":00"), x)21  if_else(!is.na(x) & nchar(x) > 10, ymd_hms(full, tz = "UTC", quiet = TRUE), as.POSIXct(NA, tz = "UTC"))22}23stamp <- function(x) if_else(is.na(x), NA_character_, format(x, "%Y-%m-%dT%H:%M:%S"))2425doses <- ex |>26  transmute(USUBJID, DRUG = EXTRT, START = moment(EXSTDTC)) |>27  filter(!is.na(START))2829samples <- pc |>30  mutate(31    ADTMM = moment(PCDTC),32    DRUG = case_when(33      PCTESTCD %in% c("DRGX", "MET1") ~ "DRUG X",34      PCTESTCD == "DRGY" ~ "DRUG Y"35    )36  )3738first <- doses |>39  group_by(USUBJID, DRUG) |>40  summarise(FIRSTM = min(START), .groups = "drop")4142# The latest dose of the drug at or before the sample.43latest <- samples |>44  filter(!is.na(ADTMM)) |>45  inner_join(doses, by = c("USUBJID", "DRUG"), relationship = "many-to-many") |>46  filter(START <= ADTMM) |>47  group_by(USUBJID, PCSEQ) |>48  summarise(LATESTM = max(START), .groups = "drop")4950adpc <- samples |>51  left_join(first, by = c("USUBJID", "DRUG")) |>52  left_join(latest, by = c("USUBJID", "PCSEQ")) |>53  mutate(54    # A sample taken before every dose refers to the first dose.55    REFM = if_else(is.na(ADTMM) | is.na(FIRSTM), as.POSIXct(NA, tz = "UTC"),56                   coalesce(LATESTM, FIRSTM)),57    PARAMCD = PCTESTCD,58    ATPT = PCTPT,59    ADTM = stamp(ADTMM),60    FANLDTM = stamp(FIRSTM),61    PCRFTDTM = stamp(REFM),62    AFRLT = as.numeric(difftime(ADTMM, FIRSTM, units = "hours")),63    ARRLT = as.numeric(difftime(ADTMM, REFM, units = "hours"))64  ) |>65  select(STUDYID, USUBJID, PCSEQ, PARAMCD, ATPT, ADTM, FANLDTM, PCRFTDTM, AFRLT, ARRLT)6667dir.create("/app/output", showWarnings = FALSE)68write_csv(adpc, "/app/output/adpc.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adpc-relative-times (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 1.67import csv8from datetime import datetime910INPUT_DIR = "/app/input"11OUTPUT_DIR = "/app/output"1213# The drug each analyte measures: the parent drug DRGX and its14# metabolite MET1 both belong to "DRUG X"; DRGY belongs to "DRUG Y".15ANALYTE_DRUG = {"DRGX": "DRUG X", "MET1": "DRUG X", "DRGY": "DRUG Y"}161718def parse_datetime(value):19    """Parse an ISO 8601 date-time as written.2021    Returns None when the value carries a date but no time; such a22    time is never filled in.23    """24    if value is None:25        return None26    value = value.strip()27    if "T" not in value:28        return None29    return datetime.fromisoformat(value)303132def main():33    with open(f"{INPUT_DIR}/ex.csv", newline="") as f:34        ex_rows = list(csv.DictReader(f))35    with open(f"{INPUT_DIR}/pc.csv", newline="") as f:36        pc_rows = list(csv.DictReader(f))3738    # Dose start date-times per subject and treatment, earliest first.39    # Each entry keeps the collected text alongside the parsed value.40    doses = {}41    for row in ex_rows:42        dtm = parse_datetime(row["EXSTDTC"])43        if dtm is None:44            continue45        key = (row["USUBJID"], row["EXTRT"])46        doses.setdefault(key, []).append((dtm, row["EXSTDTC"].strip()))47    for subject_doses in doses.values():48        subject_doses.sort(key=lambda item: item[0])4950    out_rows = []51    for row in pc_rows:52        drug = ANALYTE_DRUG[row["PCTESTCD"]]53        subject_doses = doses.get((row["USUBJID"], drug), [])54        adtm = parse_datetime(row["PCDTC"])5556        adtm_text = row["PCDTC"].strip() if adtm is not None else ""57        fanldtm = subject_doses[0][0] if subject_doses else None58        fanldtm_text = subject_doses[0][1] if subject_doses else ""5960        # Reference dose: the latest dose of the analyte's drug starting61        # at or before the sample; a sample before every dose refers to62        # the first dose. No value when ADTM or FANLDTM has no value.63        pcrftdtm = None64        pcrftdtm_text = ""65        if adtm is not None and fanldtm is not None:66            ref = subject_doses[0]67            for dose in subject_doses:68                if dose[0] <= adtm:69                    ref = dose70                else:71                    break72            pcrftdtm, pcrftdtm_text = ref7374        afrlt = ""75        if adtm is not None and fanldtm is not None:76            afrlt = (adtm - fanldtm).total_seconds() / 360077        arrlt = ""78        if adtm is not None and pcrftdtm is not None:79            arrlt = (adtm - pcrftdtm).total_seconds() / 36008081        out_rows.append(82            {83                "STUDYID": row["STUDYID"],84                "USUBJID": row["USUBJID"],85                "PCSEQ": row["PCSEQ"],86                "PARAMCD": row["PCTESTCD"],87                "ATPT": row["PCTPT"],88                "ADTM": adtm_text,89                "FANLDTM": fanldtm_text,90                "PCRFTDTM": pcrftdtm_text,91                "AFRLT": afrlt,92                "ARRLT": arrlt,93            }94        )9596    fieldnames = [97        "STUDYID",98        "USUBJID",99        "PCSEQ",100        "PARAMCD",101        "ATPT",102        "ADTM",103        "FANLDTM",104        "PCRFTDTM",105        "AFRLT",106        "ARRLT",107    ]108    with open(f"{OUTPUT_DIR}/adpc.csv", "w", newline="") as f:109        writer = csv.DictWriter(f, fieldnames=fieldnames)110        writer.writeheader()111        writer.writerows(out_rows)112113114if __name__ == "__main__":115    main()

1/* opensas reference solution for the yamaa benchmark adam-adpc-relative-times.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data pc;6  length STUDYID USUBJID PCTESTCD PCTEST PCTPT PCDTC $1024;7  infile "/app/input/pc.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. PCSEQ PCTESTCD : $char1024. PCTEST : $char1024.9        PCTPT : $char1024. PCDTC : $char1024.;10run;1112data ex;13  length STUDYID USUBJID EXTRT EXDOSU EXSTDTC $1024;14  infile "/app/input/ex.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;15  input STUDYID : $char1024. USUBJID : $char1024. EXSEQ EXTRT : $char1024. EXDOSE EXDOSU : $char1024.16        EXSTDTC : $char1024.;17run;1819/* A date without a time is not a moment; a time without seconds counts as20   second zero. Clock readings, no time zone. */21data doses;22  set ex;23  length DRUG $1024;24  DRUG=EXTRT;25  if lengthn(EXSTDTC)>10 then START=input(cats(EXSTDTC, ifc(lengthn(EXSTDTC)=16, ':00', '')), e8601dt.);26  if not missing(START);27  keep USUBJID DRUG START;28run;2930data samples;31  set pc;32  length DRUG $1024;33  if PCTESTCD in ('DRGX', 'MET1') then DRUG='DRUG X';34  else if PCTESTCD='DRGY' then DRUG='DRUG Y';35  if lengthn(PCDTC)>10 then ADTMN=input(cats(PCDTC, ifc(lengthn(PCDTC)=16, ':00', '')), e8601dt.);36run;3738proc sql;39  create table first as select USUBJID, DRUG, min(START) as FIRSTN40    from doses group by USUBJID, DRUG;41  create table latest as select s.USUBJID, s.PCSEQ, max(d.START) as LATESTN42    from samples as s inner join doses as d on s.USUBJID=d.USUBJID and s.DRUG=d.DRUG43    where not missing(s.ADTMN) and d.START <= s.ADTMN44    group by s.USUBJID, s.PCSEQ;45  create table joined as select s.*, f.FIRSTN, l.LATESTN46    from samples as s47    left join first as f on s.USUBJID=f.USUBJID and s.DRUG=f.DRUG48    left join latest as l on s.USUBJID=l.USUBJID and s.PCSEQ=l.PCSEQ49    order by s.USUBJID, s.PCSEQ;50quit;5152data result;53  length PARAMCD ATPT $1024 ADTM FANLDTM PCRFTDTM $19;54  set joined;55  PARAMCD=PCTESTCD; ATPT=PCTPT;56  ADTM=''; FANLDTM=''; PCRFTDTM='';57  /* A sample taken before every dose refers to the first dose. */58  REFN=.;59  if not missing(ADTMN) and not missing(FIRSTN) then REFN=coalesce(LATESTN, FIRSTN);60  if not missing(ADTMN) then ADTM=put(ADTMN, e8601dt19.);61  if not missing(FIRSTN) then FANLDTM=put(FIRSTN, e8601dt19.);62  if not missing(REFN) then PCRFTDTM=put(REFN, e8601dt19.);63  AFRLT=.; ARRLT=.;64  format AFRLT ARRLT best32.;65  if not missing(ADTMN) and not missing(FIRSTN) then AFRLT=(ADTMN-FIRSTN)/3600;66  if not missing(ADTMN) and not missing(REFN) then ARRLT=(ADTMN-REFN)/3600;67run;6869proc sql;70  create table final as select STUDYID, USUBJID, PCSEQ, PARAMCD, ATPT, ADTM, FANLDTM, PCRFTDTM, AFRLT, ARRLT71    from result;72quit;73proc export data=final outfile="/app/output/adpc.csv" dbms=csv replace; run;