Assign Each Vital Sign's Trial Period

1library(yamaa)23yamaa_project("project.yaml")$spec("VS")$run()$output

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

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}$", x)) {21    return(as.POSIXct(NA, tz = "UTC"))22  }23  # A date alone has no time and is compared by date instead.24  if (!grepl("T", x, fixed = TRUE)) {25    as.POSIXct(NA, tz = "UTC")26  } else {27    # Element bounds and collections use ISO datetimes without seconds.28    out <- suppressWarnings(ymd_hm(x, tz = "UTC"))29    if (is.na(out)) {30      out <- suppressWarnings(ymd_hms(x, tz = "UTC"))31    }32    out33  }34}3536se <- read_csv(37  "/app/input/se.csv",38  col_types = cols(.default = col_character())39)40vs_raw <- read_csv(41  "/app/input/vs_raw.csv",42  col_types = cols(VSSEQ = col_integer(), .default = col_character())43)4445se <- se |>46  mutate(47    start = vapply(SESTDTC, function(x) as.numeric(parse_moment(x)), numeric(1)),48    end = vapply(SEENDTC, function(x) as.numeric(parse_moment(x)), numeric(1))49  )5051vs_raw <- vs_raw |>52  mutate(53    moment = vapply(VSDTC, function(x) as.numeric(parse_moment(x)), numeric(1)),54    day = if_else(55      grepl("^[0-9]{4}-[0-9]{2}-[0-9]{2}$", coalesce(VSDTC, "")),56      suppressWarnings(as.numeric(ymd(VSDTC))),57      NA_real_58    )59  )6061# The calendar dates of each element's start and end, in days.62se <- se |>63  mutate(64    start_day = as.numeric(as.Date(as.POSIXct(start, origin = "1970-01-01", tz = "UTC"))),65    end_day = as.numeric(as.Date(as.POSIXct(end, origin = "1970-01-01", tz = "UTC")))66  )6768vs <- vs_raw |>69  mutate(70    DOMAIN = "VS",71    EPOCH = pmap_chr(72      list(USUBJID, moment, day),73      function(usubjid, coll, coll_day) {74        # A date-time is placed by moment, a date alone by calendar date;75        # when two elements match, the later one wins.76        if (!is.na(coll)) {77          hits <- se |> filter(USUBJID == usubjid, start <= coll, coll <= end)78        } else if (!is.na(coll_day)) {79          hits <- se |>80            filter(USUBJID == usubjid, start_day <= coll_day, coll_day <= end_day)81        } else {82          return(NA_character_)83        }84        hits <- hits |> arrange(start)85        if (nrow(hits) == 0) {86          return(NA_character_)87        }88        hits$EPOCH[nrow(hits)]89      },90      .progress = FALSE91    )92  ) |>93  arrange(USUBJID, VSSEQ) |>94  select(DOMAIN, STUDYID, USUBJID, VSSEQ, VSTESTCD, VSORRES, VSDTC, EPOCH)9596dir.create("/app/output", showWarnings = FALSE)97write_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 1.67"""Build the SDTM VS dataset, deriving EPOCH from the subject's SE elements.89Reads SE and VS_RAW from /app/input and writes /app/output/vs.csv with one10record per vital signs measurement.1112EPOCH is the trial period of the subject's element in progress at the13collection date-time: the element whose start is on or before the collection14and whose end is on or after it. A date-only collection is compared by date,15with no time assumed: the element whose start date is on or before it and16whose end date is on or after it. A reading that falls in two elements17belongs to the later one. EPOCH is blank when the reading has no date or18only a partial one, or when it falls outside every element.19"""2021import csv22import re23from datetime import date, datetime2425INPUT_DIR = "/app/input"26OUTPUT_DIR = "/app/output"2728FULL_DATETIME = re.compile(r"^\d{4}-\d{2}-\d{2}T\d{2}:\d{2}(:\d{2})?$")29FULL_DATE = re.compile(r"^\d{4}-\d{2}-\d{2}$")303132def parse_element_bound(text):33    """Parse an SE start/end date-time (always fully specified)."""34    return datetime.fromisoformat(text)353637def parse_collection(text):38    """Return the collection date-time or date, or None if missing or partial.3940    A full date-time parses to a datetime; a full date alone parses to a41    date, with no time assumed. Anything else (empty, year-month, year42    only) yields None.43    """44    if not text:45        return None46    text = text.strip()47    if FULL_DATETIME.match(text):48        return datetime.fromisoformat(text)49    if FULL_DATE.match(text):50        return date.fromisoformat(text)51    return None525354def covers(element, collection):55    """Whether the element is in progress at the collection: by moment for a56    date-time, by calendar date for a date alone."""57    if isinstance(collection, datetime):58        return element["start"] <= collection <= element["end"]59    return element["start"].date() <= collection <= element["end"].date()606162def derive_epoch(collection, elements):63    """Return the EPOCH of the element in progress at the collection.6465    When two elements cover the collection, at a shared boundary instant66    or on a shared boundary date, the later element (the one starting at67    the boundary) wins.68    """69    if collection is None:70        return ""71    matches = [element for element in elements if covers(element, collection)]72    if not matches:73        return ""74    later = max(matches, key=lambda element: element["start"])75    return later["epoch"]767778def main():79    with open(f"{INPUT_DIR}/se.csv", newline="") as handle:80        se_rows = list(csv.DictReader(handle))8182    elements_by_subject = {}83    for row in se_rows:84        key = (row["STUDYID"], row["USUBJID"])85        elements_by_subject.setdefault(key, []).append(86            {87                "epoch": row["EPOCH"],88                "start": parse_element_bound(row["SESTDTC"]),89                "end": parse_element_bound(row["SEENDTC"]),90            }91        )9293    with open(f"{INPUT_DIR}/vs_raw.csv", newline="") as handle:94        vs_rows = list(csv.DictReader(handle))9596    out_rows = []97    for row in vs_rows:98        key = (row["STUDYID"], row["USUBJID"])99        elements = elements_by_subject.get(key, [])100        collection = parse_collection(row["VSDTC"])101        out_rows.append(102            {103                "DOMAIN": "VS",104                "STUDYID": row["STUDYID"],105                "USUBJID": row["USUBJID"],106                "VSSEQ": row["VSSEQ"],107                "VSTESTCD": row["VSTESTCD"],108                "VSORRES": row["VSORRES"],109                "VSDTC": row["VSDTC"],110                "EPOCH": derive_epoch(collection, elements),111            }112        )113114    fieldnames = [115        "DOMAIN",116        "STUDYID",117        "USUBJID",118        "VSSEQ",119        "VSTESTCD",120        "VSORRES",121        "VSDTC",122        "EPOCH",123    ]124    with open(f"{OUTPUT_DIR}/vs.csv", "w", newline="") as handle:125        writer = csv.DictWriter(handle, fieldnames=fieldnames)126        writer.writeheader()127        writer.writerows(out_rows)128129130if __name__ == "__main__":131    main()

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;1617/* A date-time is placed by moment. A date alone is compared by date, with no18   time assumed: an element whose start date is on or before it and whose end19   date is on or after it, which is an element starting by the end of that20   day and ending at or after its start. */21data collections; set vs_raw; length DTC KEYLO KEYHI $1024; DTC=VSDTC; KEYLO=''; KEYHI=''; if lengthn(DTC)=10 then do; KEYLO=cats(DTC,'T23:59:59'); KEYHI=cats(DTC,'T00:00:00'); end; else if lengthn(DTC)=16 then do; KEYLO=cats(DTC,':00'); KEYHI=KEYLO; end; else if lengthn(DTC)>=19 then do; KEYLO=substr(DTC,1,19); KEYHI=KEYLO; end; run;22data 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;23proc 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.KEYLO>=b.STARTMOMENT and a.KEYHI<=b.ENDMOMENT and not missing(a.KEYLO); quit;24proc sort data=candidates; by STUDYID USUBJID VSSEQ descending STARTMOMENT; run;25data result; length DOMAIN $2; set candidates; DOMAIN='VS'; by STUDYID USUBJID VSSEQ; if first.VSSEQ; run;26proc sql;27  create table final as select DOMAIN, STUDYID, USUBJID, VSSEQ, VSTESTCD, VSORRES, VSDTC, EPOCH from result;28quit;29proc export data=final outfile="/app/output/vs.csv" dbms=csv replace; run;