Multiform Labs

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

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

1# Reference solution for the yamaa benchmark sdtm-lb-multiform (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(readr)9library(tidyr)1011odm <- read_csv(12  "/app/input/odm.csv",13  col_types = cols(.default = col_character())14)1516test_meta <- tribble(17  ~short, ~LBTESTCD, ~LBTEST, ~LBCAT, ~LBSPEC, ~LBLOC, ~unit,18  "VITD", "VITD25OH", "25-Hydroxyvitamin D", "CHEMISTRY", "SERUM", NA_character_, "ng/mL",19  "IL13_LES", "IL13", "Interleukin 13 mRNA", "GENE EXPRESSION", "SKIN BIOPSY", "LESIONAL", "CYCLE",20  "IL13_NONLES", "IL13", "Interleukin 13 mRNA", "GENE EXPRESSION", "SKIN BIOPSY", "NON-LESIONAL", "CYCLE",21  "SAL_CAMP", "CAMPPRO", "Cathelicidin Protein", "ANTIMICROBIAL PEPTIDE", "SALIVA", NA_character_, "ng/mL",22  "TS_CAMP_LES", "CAMPPRO", "Cathelicidin Protein", "ANTIMICROBIAL PEPTIDE", "TAPE STRIP", "LESIONAL", "ng/mL",23  "TS_CAMP_NONLES", "CAMPPRO", "Cathelicidin Protein", "ANTIMICROBIAL PEPTIDE", "TAPE STRIP", "NON-LESIONAL", "ng/mL"24)2526wide <- odm |>27  mutate(28    short = sub(".*\\.", "", ItemOID),29    is_date = grepl("DTC$", short)30  )3132dates <- wide |>33  filter(is_date, !is.na(Value) & Value != "") |>34  select(35    StudyOID, SubjectKey, StudyEventOID, ItemGroupOID, ItemGroupRepeatKey,36    LBDTC = Value37  )3839results <- wide |>40  filter(!is_date, short %in% test_meta$short, !is.na(Value) & Value != "") |>41  left_join(dates,42    by = c(43      "StudyOID", "SubjectKey", "StudyEventOID", "ItemGroupOID",44      "ItemGroupRepeatKey"45    )46  ) |>47  left_join(test_meta, by = "short") |>48  mutate(49    event_short = sub(".*\\.", "", StudyEventOID),50    VISIT = case_when(51      event_short == "SCRN" ~ "SCREENING",52      event_short == "BL" ~ "BASELINE",53      event_short == "D21" ~ "DAY 21",54      event_short == "UNSCH" ~ "UNSCHEDULED"55    ),56    VISITNUM = case_when(57      event_short == "SCRN" ~ 1,58      event_short == "BL" ~ 2,59      event_short == "D21" ~ 3,60      event_short == "UNSCH" ~ 2 + as.numeric(ItemGroupRepeatKey) / 100,61      TRUE ~ NA_real_62    ),63    not_done = Value == "NOT DONE",64    LBORRES = if_else(not_done, NA_character_, Value),65    LBORRESU = if_else(not_done, NA_character_, unit),66    LBSTRESC = if_else(not_done, NA_character_, Value),67    LBSTRESN = suppressWarnings(as.numeric(if_else(not_done, NA_character_, Value))),68    LBSTRESU = if_else(not_done, NA_character_, unit),69    LBSTAT = if_else(not_done, "NOT DONE", NA_character_),70    loc_rank = case_when(71      LBLOC == "LESIONAL" ~ 0L,72      LBLOC == "NON-LESIONAL" ~ 1L,73      TRUE ~ 2L74    )75  ) |>76  arrange(SubjectKey, LBDTC, LBTESTCD, LBSPEC, loc_rank) |>77  group_by(SubjectKey) |>78  mutate(LBSEQ = row_number()) |>79  ungroup() |>80  mutate(DOMAIN = "LB") |>81  arrange(SubjectKey, LBSEQ) |>82  select(83    DOMAIN, STUDYID = StudyOID, USUBJID = SubjectKey, LBSEQ, VISIT, VISITNUM,84    LBTESTCD, LBTEST, LBCAT, LBSPEC, LBLOC, LBORRES, LBORRESU, LBSTRESC,85    LBSTRESN, LBSTRESU, LBSTAT, LBDTC86  )8788dir.create("/app/output", showWarnings = FALSE)89write_csv(results, "/app/output/lb.csv", na = "")

1# Reference solution for the yamaa benchmark sdtm-lb-multiform (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 pathlib import Path89import polars as pl101112def _present(value) -> bool:13    return value is not None and str(value) != ""141516odm = pl.read_csv("/app/input/odm.csv", infer_schema=False)1718# Each form occurrence keeps its own date; lesional comes before19# non-lesional from the same form.20test_meta = {21    "VITD": ("VITD25OH", "25-Hydroxyvitamin D", "CHEMISTRY", "SERUM", None, "ng/mL"),22    "IL13_LES": ("IL13", "Interleukin 13 mRNA", "GENE EXPRESSION", "SKIN BIOPSY", "LESIONAL", "CYCLE"),23    "IL13_NONLES": ("IL13", "Interleukin 13 mRNA", "GENE EXPRESSION", "SKIN BIOPSY", "NON-LESIONAL", "CYCLE"),24    "SAL_CAMP": ("CAMPPRO", "Cathelicidin Protein", "ANTIMICROBIAL PEPTIDE", "SALIVA", None, "ng/mL"),25    "TS_CAMP_LES": ("CAMPPRO", "Cathelicidin Protein", "ANTIMICROBIAL PEPTIDE", "TAPE STRIP", "LESIONAL", "ng/mL"),26    "TS_CAMP_NONLES": ("CAMPPRO", "Cathelicidin Protein", "ANTIMICROBIAL PEPTIDE", "TAPE STRIP", "NON-LESIONAL", "ng/mL"),27}28visit_meta = {29    "SCRN": ("SCREENING", 1.0),30    "BL": ("BASELINE", 2.0),31    "D21": ("DAY 21", 3.0),32}333435def _result_key(itemoid: str) -> str | None:36    short = itemoid.split(".")[-1]37    return short if short in test_meta else None383940groups: dict[tuple, dict] = {}41for row in odm.to_dicts():42    key = (43        str(row["StudyOID"]),44        str(row["SubjectKey"]),45        str(row["StudyEventOID"]),46        str(row["ItemGroupOID"]),47        str(row["ItemGroupRepeatKey"]),48    )49    groups.setdefault(key, {})[str(row["ItemOID"])] = row["Value"]5051rows = []52for (studyid, subj, event_oid, _group_oid, group_rep), values in groups.items():53    event_short = event_oid.split(".")[-1]54    # The collection date from the date item on the same form.55    lbdtc = None56    for oid, val in values.items():57        if oid.split(".")[-1].endswith("DTC") and _present(val):58            lbdtc = str(val).strip()59    if event_short == "UNSCH":60        visit, visitnum = "UNSCHEDULED", 2 + float(group_rep) / 10061    else:62        visit, visitnum = visit_meta.get(event_short, (None, None))63    for oid, val in values.items():64        rkey = _result_key(oid)65        if rkey is None:66            continue67        text = str(val) if _present(val) else ""68        # An item with no reported value produces no record.69        if text == "":70            continue71        testcd, test, cat, spec, loc, unit = test_meta[rkey]72        if text == "NOT DONE":73            rows.append(74                {75                    "DOMAIN": "LB",76                    "STUDYID": studyid,77                    "USUBJID": subj,78                    "VISIT": visit,79                    "VISITNUM": visitnum,80                    "LBTESTCD": testcd,81                    "LBTEST": test,82                    "LBCAT": cat,83                    "LBSPEC": spec,84                    "LBLOC": loc,85                    "LBORRES": None,86                    "LBORRESU": None,87                    "LBSTRESC": None,88                    "LBSTRESN": None,89                    "LBSTRESU": None,90                    "LBSTAT": "NOT DONE",91                    "LBDTC": lbdtc,92                }93            )94        else:95            try:96                number = float(text)97            except ValueError:98                number = None99            rows.append(100                {101                    "DOMAIN": "LB",102                    "STUDYID": studyid,103                    "USUBJID": subj,104                    "VISIT": visit,105                    "VISITNUM": visitnum,106                    "LBTESTCD": testcd,107                    "LBTEST": test,108                    "LBCAT": cat,109                    "LBSPEC": spec,110                    "LBLOC": loc,111                    "LBORRES": text,112                    "LBORRESU": unit,113                    "LBSTRESC": text,114                    "LBSTRESN": number,115                    "LBSTRESU": unit,116                    "LBSTAT": None,117                    "LBDTC": lbdtc,118                }119            )120121loc_rank = {"LESIONAL": 0, "NON-LESIONAL": 1, None: 2}122rows.sort(123    key=lambda r: (124        r["USUBJID"],125        r["LBDTC"] or "",126        r["LBTESTCD"],127        r["LBSPEC"],128        loc_rank[r["LBLOC"]],129    )130)131seq_by_subject: dict[str, int] = {}132for row in rows:133    seq_by_subject[row["USUBJID"]] = seq_by_subject.get(row["USUBJID"], 0) + 1134    row["LBSEQ"] = str(seq_by_subject[row["USUBJID"]])135136lb = (137    pl.DataFrame(138        rows,139        schema={140            "DOMAIN": pl.String,141            "STUDYID": pl.String,142            "USUBJID": pl.String,143            "LBSEQ": pl.String,144            "VISIT": pl.String,145            "VISITNUM": pl.Float64,146            "LBTESTCD": pl.String,147            "LBTEST": pl.String,148            "LBCAT": pl.String,149            "LBSPEC": pl.String,150            "LBLOC": pl.String,151            "LBORRES": pl.String,152            "LBORRESU": pl.String,153            "LBSTRESC": pl.String,154            "LBSTRESN": pl.Float64,155            "LBSTRESU": pl.String,156            "LBSTAT": pl.String,157            "LBDTC": pl.String,158        },159    )160    .with_columns(161        pl.col("LBSEQ").cast(pl.Int64).alias("_seq"),162        pl.col("VISITNUM").cast(pl.Float64).alias("_visitnum"),163    )164    .sort(["USUBJID", "_seq"])165    .select(166        "DOMAIN", "STUDYID", "USUBJID", "LBSEQ", "VISIT", "VISITNUM",167        "LBTESTCD", "LBTEST", "LBCAT", "LBSPEC", "LBLOC", "LBORRES",168        "LBORRESU", "LBSTRESC", "LBSTRESN", "LBSTRESU", "LBSTAT", "LBDTC",169    )170)171172Path("/app/output").mkdir(exist_ok=True)173lb.write_csv("/app/output/lb.csv")

1/* opensas reference solution for the yamaa benchmark sdtm-lb-multiform.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data odm;6  length StudyOID MetaDataVersionOID SubjectKey StudyEventOID StudyEventRepeatKey ItemGroupOID ItemGroupRepeatKey ItemOID Value $1024;7  infile "/app/input/odm.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input StudyOID : $char1024. MetaDataVersionOID : $char1024. SubjectKey : $char1024. StudyEventOID : $char1024. StudyEventRepeatKey : $char1024. ItemGroupOID : $char1024. ItemGroupRepeatKey : $char1024. ItemOID : $char1024. Value : $char1024.;9run;1011data dateitems; set odm; if scan(ItemOID,-1,'.') in ('LBDTC','BXDTC','SALDTC','TSDTC'); run;12proc sql; create table dates as select StudyOID,SubjectKey,StudyEventOID,StudyEventRepeatKey,ItemGroupOID,ItemGroupRepeatKey,max(Value) as LBDTC from dateitems group by StudyOID,SubjectKey,StudyEventOID,StudyEventRepeatKey,ItemGroupOID,ItemGroupRepeatKey;13create table joined as select a.*,b.LBDTC from odm 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 and a.ItemGroupOID=b.ItemGroupOID and a.ItemGroupRepeatKey=b.ItemGroupRepeatKey; quit;14data prepared; set joined; length DOMAIN $2 STUDYID USUBJID VISIT LBTESTCD LBTEST LBCAT LBSPEC LBLOC LBORRES LBORRESU LBSTRESC LBSTRESU LBSTAT KEY EVENT $1024;15KEY=scan(ItemOID,-1,'.'); EVENT=scan(StudyEventOID,-1,'.'); if not missing(Value) and KEY in ('VITD','IL13_LES','IL13_NONLES','SAL_CAMP','TS_CAMP_LES','TS_CAMP_NONLES');16DOMAIN='LB'; STUDYID=StudyOID; USUBJID=SubjectKey; LBORRES=Value; LBSTRESC=Value; LBSTRESN=.; LBORRESU='ng/mL'; LBLOC=''; LBSTAT=''; LOCORD=2;17if Value ne 'NOT DONE' then LBSTRESN=input(Value,?? best32.);18if KEY='VITD' then do; LBTESTCD='VITD25OH'; LBTEST='25-Hydroxyvitamin D'; LBCAT='CHEMISTRY'; LBSPEC='SERUM'; end;19else if KEY in ('IL13_LES','IL13_NONLES') then do; LBTESTCD='IL13'; LBTEST='Interleukin 13 mRNA'; LBCAT='GENE EXPRESSION'; LBSPEC='SKIN BIOPSY'; LBORRESU='CYCLE'; end;20else do; LBTESTCD='CAMPPRO'; LBTEST='Cathelicidin Protein'; LBCAT='ANTIMICROBIAL PEPTIDE'; if KEY='SAL_CAMP' then LBSPEC='SALIVA'; else LBSPEC='TAPE STRIP'; end;21if KEY in ('IL13_LES','TS_CAMP_LES') then do; LBLOC='LESIONAL'; LOCORD=0; end;22if KEY in ('IL13_NONLES','TS_CAMP_NONLES') then do; LBLOC='NON-LESIONAL'; LOCORD=1; end;23VISITNUM=.; if EVENT='SCRN' then do; VISIT='SCREENING'; VISITNUM=1; end; if EVENT='BL' then do; VISIT='BASELINE'; VISITNUM=2; end; if EVENT='D21' then do; VISIT='DAY 21'; VISITNUM=3; end; if EVENT='UNSCH' then do; VISIT='UNSCHEDULED'; VISITNUM=2+input(ItemGroupRepeatKey,best32.)/100; end;24LBSTRESU=LBORRESU; if Value='NOT DONE' then do; LBORRES=''; LBORRESU=''; LBSTRESC=''; LBSTRESN=.; LBSTRESU=''; LBSTAT='NOT DONE'; end; run;25proc sort data=prepared; by STUDYID USUBJID LBDTC LBTESTCD LBSPEC LOCORD; run;26data result; set prepared; by STUDYID USUBJID; retain LBSEQ;27  if first.USUBJID then LBSEQ=0;28  LBSEQ+1;29run;3031proc sql;32  create table final as select DOMAIN, STUDYID, USUBJID, LBSEQ, VISIT, VISITNUM, LBTESTCD, LBTEST, LBCAT, LBSPEC, LBLOC, LBORRES, LBORRESU, LBSTRESC, LBSTRESN, LBSTRESU, LBSTAT, LBDTC from result;33quit;34proc export data=final outfile="/app/output/lb.csv" dbms=csv replace; run;