Once-Measured Carryforward

1library(yamaa)23advs <- yamaa_domain("spec.yaml")$output4advs

1import yamaa23advs = yamaa.yamaa_domain("spec.yaml").output4advs

1# Reference solution for the yamaa benchmark adam-advs-carryforward (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)1011plan <- read_csv(12  "/app/input/plan.csv",13  col_types = cols(ASEQ = col_integer(), ADT = col_date(), .default = col_character())14)15vs <- read_csv(16  "/app/input/vs.csv",17  col_types = cols(18    VSSEQ = col_integer(), VSDTC = col_date(), VSSTRESN = col_double(),19    .default = col_character()20  )21)22adsl <- read_csv(23  "/app/input/adsl.csv",24  col_types = cols(TRTSDT = col_date(), .default = col_character())25)2627# The record collected for a planned measurement: same subject, test code28# as the parameter, and collection date as the planned date.29collected <- plan |>30  left_join(31    vs |> select(STUDYID, USUBJID, VSSEQ, VSTESTCD, VSDTC, AVALCOL = VSSTRESN),32    by = c("STUDYID", "USUBJID", "PARAMCD" = "VSTESTCD", "ADT" = "VSDTC")33  )3435# The collected value, otherwise the most recent earlier value for the36# same subject and parameter, in planned order.37carried <- collected |>38  arrange(STUDYID, USUBJID, PARAMCD, ADT, ASEQ) |>39  group_by(STUDYID, USUBJID, PARAMCD) |>40  fill(AVALCOL, .direction = "down") |>41  mutate(AVAL = AVALCOL) |>42  ungroup()4344# The latest height on or before treatment start, repeated on every45# record of the subject.46heights <- vs |>47  filter(VSTESTCD %in% "HEIGHT", !is.na(VSSTRESN)) |>48  left_join(adsl |> select(STUDYID, USUBJID, TRTSDT), by = c("STUDYID", "USUBJID")) |>49  filter(!is.na(VSDTC), VSDTC <= TRTSDT) |>50  arrange(STUDYID, USUBJID, VSDTC, VSSEQ) |>51  group_by(STUDYID, USUBJID) |>52  summarise(HEIGHTBL = last(VSSTRESN), .groups = "drop")5354advs <- carried |>55  left_join(adsl |> select(STUDYID, USUBJID, TRTSDT), by = c("STUDYID", "USUBJID")) |>56  left_join(heights, by = c("STUDYID", "USUBJID")) |>57  arrange(USUBJID, ASEQ) |>58  select(STUDYID, USUBJID, ASEQ, VSSEQ, PARAMCD, ADT, AVAL, TRTSDT, HEIGHTBL)5960dir.create("/app/output", showWarnings = FALSE)61write_csv(advs, "/app/output/advs.csv", na = "")

1# Reference solution for the yamaa benchmark adam-advs-carryforward (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 pl1011plan = pl.read_csv("/app/input/plan.csv", infer_schema=False).with_columns(12    pl.col("ASEQ").cast(pl.Int64, strict=False),13    pl.col("ADT").str.to_date(strict=False),14)15vs = pl.read_csv("/app/input/vs.csv", infer_schema=False).with_columns(16    pl.col("VSSEQ").cast(pl.Int64, strict=False),17    pl.col("VSDTC").str.to_date(strict=False),18    pl.col("VSSTRESN").cast(pl.Float64, strict=False),19)20adsl = pl.read_csv("/app/input/adsl.csv", infer_schema=False).with_columns(21    pl.col("TRTSDT").str.to_date(strict=False)22)2324# The record collected for a planned measurement: same subject, test code25# as the parameter, and collection date as the planned date.26collected = plan.join(27    vs.select("STUDYID", "USUBJID", "VSSEQ", VSTESTCD="VSTESTCD", VSDTC="VSDTC", AVALCOL="VSSTRESN"),28    left_on=["STUDYID", "USUBJID", "PARAMCD", "ADT"],29    right_on=["STUDYID", "USUBJID", "VSTESTCD", "VSDTC"],30    how="left",31)3233# The collected value, otherwise the most recent earlier value for the34# same subject and parameter, in planned order.35carried = (36    collected.sort(["STUDYID", "USUBJID", "PARAMCD", "ADT", "ASEQ"])37    .with_columns(38        AVAL=pl.col("AVALCOL")39        .fill_null(strategy="forward")40        .over(["STUDYID", "USUBJID", "PARAMCD"])41    )42    .with_columns(AVAL=pl.coalesce("AVALCOL", "AVAL"))43)4445# The latest height on or before treatment start, repeated on every46# record of the subject.47heights = (48    vs.filter(49        (pl.col("VSTESTCD") == "HEIGHT") & pl.col("VSSTRESN").is_not_null()50    )51    .join(adsl.select("STUDYID", "USUBJID", "TRTSDT"), on=["STUDYID", "USUBJID"], how="left")52    .filter(pl.col("VSDTC").is_not_null() & (pl.col("VSDTC") <= pl.col("TRTSDT")))53    .sort(["VSDTC", "VSSEQ"])54    .unique(["STUDYID", "USUBJID"], keep="last", maintain_order=True)55    .select("STUDYID", "USUBJID", HEIGHTBL="VSSTRESN")56)5758advs = (59    carried.join(adsl.select("STUDYID", "USUBJID", "TRTSDT"), on=["STUDYID", "USUBJID"], how="left")60    .join(heights, on=["STUDYID", "USUBJID"], how="left")61    .sort(["USUBJID", "ASEQ"])62    .select(63        "STUDYID", "USUBJID", "ASEQ", "VSSEQ", "PARAMCD", "ADT", "AVAL",64        "TRTSDT", "HEIGHTBL",65    )66)6768Path("/app/output").mkdir(parents=True, exist_ok=True)69advs.write_csv("/app/output/advs.csv")

1/* opensas reference solution for the yamaa benchmark adam-advs-carryforward.2   Reads the benchmark inputs and derives the requested output.3   Execute with the existing opensas runtime: opensas /app/output/result.sas. */45data adsl;6  length STUDYID USUBJID TRTSDT $1024;7  infile "/app/input/adsl.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;8  input STUDYID : $char1024. USUBJID : $char1024. TRTSDT : $char1024.;9run;1011data plan;12  length STUDYID USUBJID PARAMCD ADT $1024;13  infile "/app/input/plan.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;14  input STUDYID : $char1024. USUBJID : $char1024. ASEQ PARAMCD : $char1024. ADT : $char1024.;15run;1617data vs;18  length STUDYID USUBJID VSTESTCD VSDTC $1024;19  infile "/app/input/vs.csv" dsd dlm="," firstobs=2 truncover lrecl=32767;20  input STUDYID : $char1024. USUBJID : $char1024. VSSEQ VSTESTCD : $char1024. VSDTC : $char1024. VSSTRESN;21run;2223proc sql;24  create table heights as select a.STUDYID,a.USUBJID,a.VSSTRESN as HEIGHTBL,a.VSDTC,a.VSSEQ from vs as a inner join adsl as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID25    where a.VSTESTCD='HEIGHT' and not missing(a.VSSTRESN) and a.VSDTC ne '' and a.VSDTC<=b.TRTSDT;26quit;27proc sort data=heights; by STUDYID USUBJID descending VSDTC descending VSSEQ; run;28data height; set heights; by STUDYID USUBJID; if first.USUBJID; run;29proc sql;30  create table collected as select a.*,b.VSSEQ,b.VSSTRESN as AVALCOL,c.TRTSDT,d.HEIGHTBL from plan as a31    left join vs as b on a.STUDYID=b.STUDYID and a.USUBJID=b.USUBJID and a.PARAMCD=b.VSTESTCD and a.ADT=b.VSDTC32    left join adsl as c on a.STUDYID=c.STUDYID and a.USUBJID=c.USUBJID left join height as d on a.STUDYID=d.STUDYID and a.USUBJID=d.USUBJID;33quit;34proc sort data=collected; by STUDYID USUBJID PARAMCD ADT ASEQ; run;35data result;36  retain carried; set collected; by STUDYID USUBJID PARAMCD;37  if first.PARAMCD then carried=.;38  if not missing(AVALCOL) then carried=AVALCOL;39  AVAL=carried;40run;4142proc sql;43  create table final as select STUDYID, USUBJID, ASEQ, VSSEQ, PARAMCD, ADT, AVAL, TRTSDT, HEIGHTBL from result;44quit;45proc export data=final outfile="/app/output/advs.csv" dbms=csv replace; run;