Assign Each Vital Sign's Trial Period

1library(yamaa)23vs <- yamaa_domain("spec.yaml")$output4vs

1import yamaa23vs = yamaa.yamaa_domain("spec.yaml").output4vs

1# Reference solution for the yamaa benchmark sdtm-vs-epoch-from-subject-elements (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)9library(purrr)10library(readr)1112parse_moment <- function(x) {13  if (is.na(x) || x == "") {14    return(as.POSIXct(NA, tz = "UTC"))15  }16  # A partial date such as YYYY-MM has no day, so it cannot be placed.17  if (nchar(x) == 7 && substr(x, 5, 5) == "-") {18    return(as.POSIXct(NA, tz = "UTC"))19  }20  if (nchar(x) == 4 && grepl("^[0-9]{4}$", x)) {21    return(as.POSIXct(NA, tz = "UTC"))22  }23  if (!grepl("T", x, fixed = TRUE)) {24    suppressWarnings(ymd_hms(paste0(x, "T00:00:00"), tz = "UTC"))25  } else {26    # Element bounds and collections use ISO datetimes without seconds.27    out <- suppressWarnings(ymd_hm(x, tz = "UTC"))28    if (is.na(out)) {29      out <- suppressWarnings(ymd_hms(x, tz = "UTC"))30    }31    out32  }33}3435se <- read_csv(36  "/app/input/se.csv",37  col_types = cols(.default = col_character())38)39vs_raw <- read_csv(40  "/app/input/vs_raw.csv",41  col_types = cols(VSSEQ = col_integer(), .default = col_character())42)4344se <- se |>45  mutate(46    start = vapply(SESTDTC, function(x) as.numeric(parse_moment(x)), numeric(1)),47    end = vapply(SEENDTC, function(x) as.numeric(parse_moment(x)), numeric(1))48  )4950vs_raw <- vs_raw |>51  mutate(moment = vapply(VSDTC, function(x) as.numeric(parse_moment(x)), numeric(1)))5253vs <- vs_raw |>54  mutate(55    DOMAIN = "VS",56    EPOCH = pmap_chr(57      list(USUBJID, moment),58      function(usubjid, coll) {59        if (is.na(coll)) {60          return(NA_character_)61        }62        hits <- se |>63          filter(USUBJID == usubjid, start <= coll, coll <= end) |>64          arrange(start)65        if (nrow(hits) == 0) {66          return(NA_character_)67        }68        hits$EPOCH[nrow(hits)]69      },70      .progress = FALSE71    )72  ) |>73  arrange(USUBJID, VSSEQ) |>74  select(DOMAIN, STUDYID, USUBJID, VSSEQ, VSTESTCD, VSORRES, VSDTC, EPOCH)7576dir.create("/app/output", showWarnings = FALSE)77write_csv(vs, "/app/output/vs.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-vs-epoch-from-subject-elements (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 1.67from datetime import datetime8from pathlib import Path910import polars as pl111213def parse_moment(text: str | None) -> datetime | None:14    if text is None or text == "":15        return None16    # A partial date such as YYYY-MM has no day, so it cannot be placed in17    # an element.18    if len(text) == 7 and text[4] == "-":19        return None20    if len(text) == 4 and text.isdigit():21        return None22    try:23        # A date-only collection reads as the start of that day.24        if "T" not in text:25            return datetime.fromisoformat(text + "T00:00:00")26        return datetime.fromisoformat(text)27    except ValueError:28        return None293031se = pl.read_csv("/app/input/se.csv", infer_schema=False)32vs_raw = pl.read_csv("/app/input/vs_raw.csv", infer_schema=False).with_columns(33    pl.col("VSSEQ").cast(pl.Int64)34)3536elements: dict[str, list[dict]] = {}37for row in se.to_dicts():38    elements.setdefault(row["USUBJID"], []).append(39        {40            "EPOCH": row["EPOCH"],41            "start": parse_moment(row["SESTDTC"]),42            "end": parse_moment(row["SEENDTC"]),43        }44    )4546rows = []47for row in vs_raw.to_dicts():48    coll = parse_moment(row["VSDTC"])49    # Only a full date or datetime can fall in an element; empty and50    # partial dates stay empty, as do readings outside every element.51    epoch = None52    if coll is not None:53        hits = [54            e55            for e in elements.get(row["USUBJID"], [])56            if e["start"] is not None57            and e["end"] is not None58            and e["start"] <= coll <= e["end"]59        ]60        if hits:61            hits.sort(key=lambda e: e["start"])62            # A reading exactly at a shared boundary belongs to the later63            # element.64            epoch = hits[-1]["EPOCH"]65    rows.append(66        {67            "DOMAIN": "VS",68            "STUDYID": row["STUDYID"],69            "USUBJID": row["USUBJID"],70            "VSSEQ": row["VSSEQ"],71            "VSTESTCD": row["VSTESTCD"],72            "VSORRES": row["VSORRES"],73            "VSDTC": row["VSDTC"],74            "EPOCH": epoch,75        }76    )7778vs = pl.DataFrame(rows).sort(["USUBJID", "VSSEQ"]).select(79    "DOMAIN", "STUDYID", "USUBJID", "VSSEQ", "VSTESTCD", "VSORRES", "VSDTC", "EPOCH"80)8182Path("/app/output").mkdir(exist_ok=True)83vs.write_csv("/app/output/vs.csv")

1/* opensas reference solution for the yamaa benchmark sdtm-vs-epoch-from-subject-elements.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data se;6  length STUDYID USUBJID ETCD ELEMENT EPOCH SESTDTC SEENDTC $1024;7  infile "/app/input/se.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. ETCD : $char1024. ELEMENT : $char1024. EPOCH : $char1024. SESTDTC : $char1024. SEENDTC : $char1024.;9run;1011data vs_raw;12  length STUDYID USUBJID VSTESTCD VSORRES VSDTC $1024;13  infile "/app/input/vs_raw.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;14  input STUDYID : $char1024. USUBJID : $char1024. VSSEQ VSTESTCD : $char1024. VSORRES : $char1024. VSDTC : $char1024.;15run;1617data collections; set vs_raw; length DTC MOMENT $1024; DTC=VSDTC; MOMENT=''; if lengthn(DTC)=10 then MOMENT=cats(DTC,'T00:00:00'); else if lengthn(DTC)=16 then MOMENT=cats(DTC,':00'); else if lengthn(DTC)>=19 then MOMENT=substr(DTC,1,19); run;18data elements; set se; length DTC MOMENT STARTMOMENT ENDMOMENT $1024; DTC=SESTDTC; MOMENT=''; if lengthn(DTC)=10 then MOMENT=cats(DTC,'T00:00:00'); else if lengthn(DTC)=16 then MOMENT=cats(DTC,':00'); else if lengthn(DTC)>=19 then MOMENT=substr(DTC,1,19); STARTMOMENT=MOMENT; DTC=SEENDTC; MOMENT=''; if lengthn(DTC)=10 then MOMENT=cats(DTC,'T00:00:00'); else if lengthn(DTC)=16 then MOMENT=cats(DTC,':00'); else if lengthn(DTC)>=19 then MOMENT=substr(DTC,1,19); ENDMOMENT=MOMENT; run;19proc sql; create table candidates as select a.*,b.EPOCH,b.STARTMOMENT from collections a left join elements b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID and a.MOMENT>=b.STARTMOMENT and a.MOMENT<=b.ENDMOMENT and not missing(a.MOMENT); quit;20proc sort data=candidates; by STUDYID USUBJID VSSEQ descending STARTMOMENT; run;21data result; length DOMAIN $2; set candidates; DOMAIN='VS'; by STUDYID USUBJID VSSEQ; if first.VSSEQ; run;22proc sql;23  create table final as select DOMAIN, STUDYID, USUBJID, VSSEQ, VSTESTCD, VSORRES, VSDTC, EPOCH from result;24quit;25proc export data=final outfile="/app/output/vs.csv" dbms=csv replace; run;