Derive analysis period start and end dates in a two-period study

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

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

1# Reference solution for the yamaa benchmark adam-adsl-period-dates (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)1011dm <- read_csv("/app/input/dm.csv", col_types = cols(.default = col_character()))12ex <- read_csv(13  "/app/input/ex.csv",14  col_types = cols(EXSEQ = col_integer(), .default = col_character())15) |>16  mutate(START = ymd(EXSTDTC, quiet = TRUE), END = ymd(EXENDTC, quiet = TRUE))1718first_start <- function(epoch) {19  ex |>20    filter(EPOCH == epoch, !is.na(START)) |>21    group_by(USUBJID) |>22    summarise(FIRST = min(START), .groups = "drop")23}2425# The last exposure end counts every record, with or without a start.26last_end <- ex |>27  filter(!is.na(END)) |>28  group_by(USUBJID) |>29  summarise(LASTEND = max(END), .groups = "drop")3031adsl <- dm |>32  left_join(rename(first_start("TREATMENT 1"), AP01SDTN = FIRST), by = "USUBJID") |>33  left_join(rename(first_start("TREATMENT 2"), AP02SDTN = FIRST), by = "USUBJID") |>34  left_join(last_end, by = "USUBJID") |>35  mutate(36    FOLLOW = LASTEND + days(30),37    # Period one ends the day before period two starts; the last period a38    # subject entered ends 30 days after the last exposure end.39    AP01EDTN = case_when(40      is.na(LASTEND) ~ as.Date(NA),41      !is.na(AP02SDTN) ~ AP02SDTN - days(1),42      TRUE ~ FOLLOW43    ),44    AP02EDTN = if_else(is.na(AP02SDTN), as.Date(NA), FOLLOW),45    AP01SDT = format(AP01SDTN, "%Y-%m-%d"),46    AP01EDT = format(AP01EDTN, "%Y-%m-%d"),47    AP02SDT = format(AP02SDTN, "%Y-%m-%d"),48    AP02EDT = format(AP02EDTN, "%Y-%m-%d")49  ) |>50  select(STUDYID, USUBJID, AP01SDT, AP01EDT, AP02SDT, AP02EDT)5152dir.create("/app/output", showWarnings = FALSE)53write_csv(adsl, "/app/output/adsl.csv", na = "")

1# Reference solution for the yamaa benchmark adam-adsl-period-dates (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"""Build ADSL period dates from DM and EX.89One record per subject (from DM) with the analysis-period dates:10AP01SDT/AP02SDT are the earliest exposure start dates in the11"TREATMENT 1" / "TREATMENT 2" epochs; AP01EDT is the day before12period two started, or 30 days after the last exposure end date when13the subject never entered period two; AP02EDT is 30 days after the14last exposure end date.15"""1617import csv18from datetime import date, timedelta1920INPUT_DIR = "/app/input"21OUTPUT_DIR = "/app/output"222324def parse_date(text):25    """Return a date for an ISO date string, or None when blank."""26    text = (text or "").strip()27    if not text:28        return None29    return date.fromisoformat(text)303132def fmt(value):33    return value.isoformat() if value is not None else ""343536def main():37    with open(f"{INPUT_DIR}/dm.csv", newline="") as handle:38        dm_rows = list(csv.DictReader(handle))39    with open(f"{INPUT_DIR}/ex.csv", newline="") as handle:40        ex_rows = list(csv.DictReader(handle))4142    ex_by_subject = {}43    for row in ex_rows:44        ex_by_subject.setdefault(row["USUBJID"], []).append(row)4546    out_rows = []47    for dm in dm_rows:48        records = ex_by_subject.get(dm["USUBJID"], [])4950        def earliest_start(epoch):51            starts = [52                parse_date(r["EXSTDTC"])53                for r in records54                if r["EPOCH"] == epoch55            ]56            starts = [s for s in starts if s is not None]57            return min(starts) if starts else None5859        ap01sdt = earliest_start("TREATMENT 1")60        ap02sdt = earliest_start("TREATMENT 2")6162        ends = [parse_date(r["EXENDTC"]) for r in records]63        ends = [e for e in ends if e is not None]64        last_end = max(ends) if ends else None6566        entered_period_two = ap02sdt is not None67        if entered_period_two:68            ap01edt = ap02sdt - timedelta(days=1)69            ap02edt = (70                last_end + timedelta(days=30) if last_end is not None else None71            )72        else:73            ap01edt = (74                last_end + timedelta(days=30) if last_end is not None else None75            )76            ap02edt = None7778        out_rows.append(79            {80                "STUDYID": dm["STUDYID"],81                "USUBJID": dm["USUBJID"],82                "AP01SDT": fmt(ap01sdt),83                "AP01EDT": fmt(ap01edt),84                "AP02SDT": fmt(ap02sdt),85                "AP02EDT": fmt(ap02edt),86            }87        )8889    columns = ["STUDYID", "USUBJID", "AP01SDT", "AP01EDT", "AP02SDT", "AP02EDT"]90    with open(f"{OUTPUT_DIR}/adsl.csv", "w", newline="") as handle:91        writer = csv.DictWriter(handle, fieldnames=columns)92        writer.writeheader()93        writer.writerows(out_rows)949596if __name__ == "__main__":97    main()

1/* opensas reference solution for the yamaa benchmark adam-adsl-period-dates.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 $1024;7  infile "/app/input/dm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024.;9  ORDER=_n_;10run;1112data ex;13  length STUDYID USUBJID EPOCH EXSTDTC EXENDTC $1024;14  infile "/app/input/ex.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;15  input STUDYID : $char1024. USUBJID : $char1024. EXSEQ EPOCH : $char1024. EXSTDTC : $char1024.16        EXENDTC : $char1024.;17  STARTDT=.; ENDDT=.;18  if not missing(EXSTDTC) then STARTDT=input(EXSTDTC, yymmdd10.);19  if not missing(EXENDTC) then ENDDT=input(EXENDTC, yymmdd10.);20run;2122/* The last exposure end counts every record, with or without a start. */23proc sql;24  create table p1 as select USUBJID, min(STARTDT) as P1 from ex where EPOCH='TREATMENT 1' group by USUBJID;25  create table p2 as select USUBJID, min(STARTDT) as P2 from ex where EPOCH='TREATMENT 2' group by USUBJID;26  create table lastend as select USUBJID, max(ENDDT) as LASTEND from ex group by USUBJID;27  create table subject as28  select d.STUDYID, d.USUBJID, d.ORDER, a.P1, b.P2, c.LASTEND29  from dm as d30  left join p1 as a on d.USUBJID=a.USUBJID31  left join p2 as b on d.USUBJID=b.USUBJID32  left join lastend as c on d.USUBJID=c.USUBJID33  order by d.ORDER;34quit;3536/* Period one ends the day before period two starts; the last period a37   subject entered ends 30 days after the last exposure end. */38data final;39  length AP01SDT AP01EDT AP02SDT AP02EDT $10;40  set subject;41  AP01SDT=''; AP01EDT=''; AP02SDT=''; AP02EDT='';42  if not missing(P1) then AP01SDT=put(P1, yymmdd10.);43  if not missing(P2) then AP02SDT=put(P2, yymmdd10.);44  if not missing(LASTEND) then do;45    if not missing(P2) then do;46      AP01EDT=put(P2-1, yymmdd10.);47      AP02EDT=put(LASTEND+30, yymmdd10.);48    end;49    else AP01EDT=put(LASTEND+30, yymmdd10.);50  end;51  keep STUDYID USUBJID AP01SDT AP01EDT AP02SDT AP02EDT;52run;5354proc sql;55  create table out as select STUDYID, USUBJID, AP01SDT, AP01EDT, AP02SDT, AP02EDT from final;56quit;57proc export data=out outfile="/app/output/adsl.csv" dbms=csv replace; run;