Build SDTM subject visits from ODM extracts

1library(yamaa)23sv <- yamaa_domain("spec.yaml")$output4sv

1import yamaa23sv = yamaa.yamaa_domain("spec.yaml").output4sv

1# Reference solution for the yamaa benchmark sdtm-sv-subject-visits (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(readr)10library(tidyr)1112study_day <- function(ref, coll) {13  if (is.na(ref) || is.na(coll)) {14    return(NA_integer_)15  }16  delta <- as.integer(coll - ref)17  if (delta >= 0L) delta + 1L else delta18}1920dm <- read_csv(21  "/app/input/dm.csv",22  col_types = cols(.default = col_character())23)24odm <- read_csv(25  "/app/input/odm.csv",26  col_types = cols(.default = col_character())27)2829ref_map <- setNames(suppressWarnings(ymd(dm$RFSTDTC)), dm$USUBJID)3031wide <- odm |>32  mutate(visit_key = paste(SubjectKey, StudyEventOID, StudyEventRepeatKey, sep = "|")) |>33  select(visit_key, SubjectKey, StudyOID, StudyEventOID, ItemOID, Value) |>34  pivot_wider(names_from = ItemOID, values_from = Value, values_fn = first)3536sv <- wide |>37  rowwise() |>38  mutate(39    dates = list(na.omit(c(IT.VS.VSDTC, IT.LB.LBDTC, IT.EX.EXDTC))),40    SVSTDTC = if (length(dates) == 0) NA_character_ else min(dates),41    SVENDTC = if (length(dates) == 0) NA_character_ else max(dates)42  ) |>43  ungroup() |>44  filter(!is.na(SVSTDTC)) |>45  transmute(46    DOMAIN = "SV",47    STUDYID = StudyOID,48    USUBJID = SubjectKey,49    VISIT = IT.TV.VISIT,50    VISITNUM = suppressWarnings(as.numeric(IT.TV.VISITNUM)),51    VISITDY = suppressWarnings(as.integer(IT.TV.VISITDY)),52    SVSTDTC,53    SVENDTC,54    SVSTDY = vapply(55      seq_len(n()),56      function(i) study_day(ref_map[[USUBJID[i]]], suppressWarnings(ymd(SVSTDTC[i]))),57      integer(1)58    ),59    SVENDY = vapply(60      seq_len(n()),61      function(i) study_day(ref_map[[USUBJID[i]]], suppressWarnings(ymd(SVENDTC[i]))),62      integer(1)63    ),64    TAETORD = suppressWarnings(as.integer(IT.TV.TAETORD)),65    EPOCH = IT.TV.EPOCH,66    SVUPDES = IT.TV.UPDES67  ) |>68  arrange(USUBJID, SVSTDTC) |>69  group_by(USUBJID) |>70  mutate(SVSEQ = row_number()) |>71  ungroup() |>72  arrange(USUBJID, SVSEQ) |>73  select(74    DOMAIN, STUDYID, USUBJID, SVSEQ, VISIT, VISITNUM, VISITDY, SVSTDTC,75    SVENDTC, SVSTDY, SVENDY, TAETORD, EPOCH, SVUPDES76  )7778dir.create("/app/output", showWarnings = FALSE)79write_csv(sv, "/app/output/sv.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-sv-subject-visits (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 date8from pathlib import Path910import polars as pl111213def parse_date(text: str | None) -> date | None:14    if text is None or text == "":15        return None16    try:17        return date.fromisoformat(text[:10])18    except ValueError:19        return None202122def study_day(ref: date | None, coll: date | None) -> int | None:23    if ref is None or coll is None:24        return None25    delta = (coll - ref).days26    return delta + 1 if delta >= 0 else delta272829dm = pl.read_csv("/app/input/dm.csv", infer_schema=False)30odm = pl.read_csv("/app/input/odm.csv", infer_schema=False)3132ref_map = {row["USUBJID"]: parse_date(row["RFSTDTC"]) for row in dm.to_dicts()}33study_map = {row["USUBJID"]: row["STUDYID"] for row in dm.to_dicts()}3435date_items = {"IT.VS.VSDTC", "IT.LB.LBDTC", "IT.EX.EXDTC"}3637visits: dict[tuple[str, str, str], dict] = {}38for row in odm.to_dicts():39    key = (row["SubjectKey"], row["StudyEventOID"], row["StudyEventRepeatKey"])40    entry = visits.setdefault(41        key,42        {43            "subject": row["SubjectKey"],44            "study": row["StudyOID"],45            "tv": {},46            "dates": [],47        },48    )49    item = row["ItemOID"]50    value = row["Value"]51    if item.startswith("IT.TV."):52        entry["tv"][item] = value53    elif item in date_items and value not in (None, ""):54        entry["dates"].append(value)5556records = []57for entry in visits.values():58    tv = entry["tv"]59    dates = sorted(entry["dates"])60    if not dates:61        continue62    start, end = dates[0], dates[-1]63    ref = ref_map.get(entry["subject"])64    records.append(65        {66            "DOMAIN": "SV",67            "STUDYID": entry["study"],68            "USUBJID": entry["subject"],69            "VISIT": tv.get("IT.TV.VISIT"),70            "VISITNUM": float(tv["IT.TV.VISITNUM"]) if tv.get("IT.TV.VISITNUM") not in (None, "") else None,71            "VISITDY": int(tv["IT.TV.VISITDY"]) if tv.get("IT.TV.VISITDY") not in (None, "") else None,72            "SVSTDTC": start,73            "SVENDTC": end,74            "_start": start,75            "SVSTDY": study_day(ref, parse_date(start)),76            "SVENDY": study_day(ref, parse_date(end)),77            "TAETORD": int(tv["IT.TV.TAETORD"]) if tv.get("IT.TV.TAETORD") not in (None, "") else None,78            "EPOCH": tv.get("IT.TV.EPOCH"),79            "SVUPDES": tv.get("IT.TV.UPDES"),80        }81    )8283sv = (84    pl.DataFrame(records)85    .sort(["USUBJID", "_start"])86    .with_columns(SVSEQ=pl.int_range(1, pl.len() + 1).over("USUBJID"))87    .sort(["USUBJID", "SVSEQ"])88    .select(89        "DOMAIN",90        "STUDYID",91        "USUBJID",92        "SVSEQ",93        "VISIT",94        "VISITNUM",95        "VISITDY",96        "SVSTDTC",97        "SVENDTC",98        "SVSTDY",99        "SVENDY",100        "TAETORD",101        "EPOCH",102        "SVUPDES",103    )104)105106Path("/app/output").mkdir(exist_ok=True)107sv.write_csv("/app/output/sv.csv")

1/* opensas reference solution for the yamaa benchmark sdtm-sv-subject-visits.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 RFSTDTC $1024;7  infile "/app/input/dm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. RFSTDTC : $char1024.;9run;1011data odm;12  length StudyOID MetaDataVersionOID SubjectKey StudyEventOID StudyEventRepeatKey ItemGroupOID ItemGroupRepeatKey ItemOID Value $1024;13  infile "/app/input/odm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;14  input StudyOID : $char1024. MetaDataVersionOID : $char1024. SubjectKey : $char1024. StudyEventOID : $char1024. StudyEventRepeatKey : $char1024. ItemGroupOID : $char1024. ItemGroupRepeatKey : $char1024. ItemOID : $char1024. Value : $char1024.;15run;1617proc sort data=odm out=ordered; by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey; run;18data visits;19  length VISIT VNUM VDAY ORD EPOCH SVUPDES $1024;20  retain VISIT VNUM VDAY ORD EPOCH SVUPDES;21  set ordered;2223  by StudyOID SubjectKey StudyEventOID StudyEventRepeatKey;24  if first.StudyEventRepeatKey then do;25    VISIT=''; VNUM=''; VDAY=''; ORD=''; EPOCH=''; SVUPDES='';26  end;27    if ItemOID='IT.TV.VISIT' then VISIT=Value;28  if ItemOID='IT.TV.VISITNUM' then VNUM=Value;29  if ItemOID='IT.TV.VISITDY' then VDAY=Value;30  if ItemOID='IT.TV.TAETORD' then ORD=Value;31  if ItemOID='IT.TV.EPOCH' then EPOCH=Value;32  if ItemOID='IT.TV.UPDES' then SVUPDES=Value;33  if last.StudyEventRepeatKey then output;34run;35proc sql;36create table dates as select StudyOID,SubjectKey,StudyEventOID,StudyEventRepeatKey,min(Value) as SVSTDTC,max(Value) as SVENDTC from odm where ItemOID in ('IT.VS.VSDTC','IT.LB.LBDTC','IT.EX.EXDTC') and not missing(Value) group by StudyOID,SubjectKey,StudyEventOID,StudyEventRepeatKey;37create table joined as select a.*,b.SVSTDTC,b.SVENDTC,c.RFSTDTC from visits a inner join dates b on a.StudyOID=b.StudyOID and a.SubjectKey=b.SubjectKey and a.StudyEventOID=b.StudyEventOID and a.StudyEventRepeatKey=b.StudyEventRepeatKey left join dm c on a.StudyOID=c.STUDYID and a.SubjectKey=c.USUBJID; quit;38data prepared; set joined; length DOMAIN $2 STUDYID USUBJID $1024; DOMAIN='SV'; STUDYID=StudyOID; USUBJID=SubjectKey; VISITNUM=input(VNUM,best32.); VISITDY=input(VDAY,best32.); TAETORD=input(ORD,best32.); SVSTDY=.; SVENDY=.; if not missing(RFSTDTC) then do; SVSTDY=input(SVSTDTC,yymmdd10.)-input(RFSTDTC,yymmdd10.); if SVSTDY>=0 then SVSTDY+1; SVENDY=input(SVENDTC,yymmdd10.)-input(RFSTDTC,yymmdd10.); if SVENDY>=0 then SVENDY+1; end; run;39proc sort data=prepared; by STUDYID USUBJID SVSTDTC StudyEventOID StudyEventRepeatKey; run;40data result; set prepared; by STUDYID USUBJID; retain SVSEQ UNSCH;41if first.USUBJID then do; SVSEQ=0; UNSCH=0; end; SVSEQ+1;42if StudyEventOID='UNSCH' then do; UNSCH+1; VISITNUM=99+UNSCH/10; VISITDY=.; TAETORD=.; end; else SVUPDES=''; run;43proc sql;44  create table final as select DOMAIN, STUDYID, USUBJID, SVSEQ, VISIT, VISITNUM, VISITDY, SVSTDTC, SVENDTC, SVSTDY, SVENDY, TAETORD, EPOCH, SVUPDES from result;45quit;46proc export data=final outfile="/app/output/sv.csv" dbms=csv replace; run;