通常問題を60番まで追加し、問題文と解答例を整備

Author: YusukeKato

backend/tests/integration/README.md

@@ -85,11 +85,14 @@ runnerとjudgeには起動時検証済みの同じ不変problem repositoryを注
 期待出力、正解画像がrequestごとのfile再読込なしで利用される経路を検証します。
 legacy `yaml_data/`とのsemantic一致と意図した改訂の範囲は、
 [問題データ仕様](../../../problems/README.md)に従いnon-Docker testで別途検証します。
-`test_reservation_problem.py`は未受付者を求める問題の参照解答と2つの別解を実行し、
-集合差の逆転・重複の残存・部分一致による除外を不正解として区別します。
+`test_reservation_problem.py`は未受付者を求める問題の参照解答と3つの別解を実行し、
+集合差の逆転・重複の残存・部分一致による除外・初出判定前の予約抽出を不正解として区別します。
 `test_remaining_problem_solutions.py`は通常21〜51・練習問題の改訂に関係する別解と誤答を検証します。
 画像5問ではMIFF中間画像を使う別解の正解判定と、色・寸法・図形端点の誤りの不正解判定を確認します。
-全93問の参照解答は上記の全問題回帰で確認します。
+現在登録されている全問題の参照解答は上記の全問題回帰で確認します。
+`test_new_standard_solutions.py`は通常53〜60の参照解答と別解、および初出優先、差の逆転、
+境界の部分一致、空欄を失う分割、大小文字の未統一、結合キーの重複除去、
+タブの固定幅置換・元のスペースの変更、終了時刻を含む区間、指定された集計時刻の欠落の誤答を実sandboxで区別します。
 
 `test_revised_standard_solutions.py`は改訂した通常問題3・5〜20の参照解答・別解と、
 累積の欠落、丸め違い、改行の全削除、探索範囲の不足、記録や空白の破壊、置換漏れに加え、

backend/tests/integration/test_new_standard_solutions.py

@@ -0,0 +1,124 @@
+"""通常53〜60の別解と典型的な誤答を、実sandboxとjudgeで検査する。"""
+
+import asyncio
+import os
+from pathlib import Path
+
+import pytest
+
+from soj_backend.judge import JudgeVerdict, ShellgeiJudge
+from soj_runner.container_manager import ContainerManager
+from soj_runner.run_shellgei import ShellgeiDockerClient
+from soj_shared.models.execution import ExecutionStatus
+from soj_shared.problem_repository import build_problem_repository
+
+
+pytestmark = [
+    pytest.mark.docker,
+    pytest.mark.skipif(
+        os.getenv("SOJ_RUN_DOCKER_TESTS") != "1",
+        reason="explicit isolated-host opt-in is required",
+    ),
+]
+PROBLEMS = Path(__file__).resolve().parents[3] / "problems"
+CASES = [
+    (
+        53,
+        "awk '{last[$1]=$0} END{for(k in last)print last[k]}' input.txt | sort",
+        "awk '!seen[$1]++' input.txt | sort",
+    ),
+    (
+        54,
+        "awk 'NR>1{print $1,$2-prev} {prev=$2}' input.txt",
+        "awk 'NR>1{print $1,prev-$2} {prev=$2}' input.txt",
+    ),
+    (
+        55,
+        'awk \'$0=="END"{inside=0} inside{print} $0=="BEGIN"{inside=1}\' input.txt',
+        "sed -n '/BEGIN/,/END/{/BEGIN/d;/END/d;p;}' input.txt",
+    ),
+    (
+        56,
+        "awk 'BEGIN{FS=OFS=\"\\t\"} {print $1,$3}' input.txt",
+        "awk 'BEGIN{OFS=\"\\t\"} {print $1,$3}' input.txt",
+    ),
+    (
+        57,
+        "awk -F. '{n[tolower($NF)]++} END{for(e in n)print e,n[e]}' input.txt | sort",
+        "sed 's/.*\\.//' input.txt | sort | uniq -c | awk '{print $2,$1}'",
+    ),
+    (
+        58,
+        'awk \'$1=="team"{people[$2,$3]=1} $1=="file"{files[$2,$3]=1} '
+        "END{for(p in people){split(p,a,SUBSEP);for(f in files){split(f,b,SUBSEP);"
+        "if(a[1]==b[1])print a[1],a[2],b[2]}}}' input.txt | sort",
+        "join <(sed -n 's/^team //p' input.txt | sort -u -k1,1) "
+        "<(sed -n 's/^file //p' input.txt | sort) | sort",
+    ),
+    (
+        59,
+        "awk '{col=0;for(i=1;i<=length;i++){c=substr($0,i,1);"
+        'if(c=="\\t"){n=4-col%4;printf "%s",substr(">>>>",1,n);col+=n}'
+        'else{printf "%s",c;col++}}print ""}\' input.txt',
+        "sed 's/\t/>>>>/g' input.txt",
+    ),
+    (
+        60,
+        "awk '{s[NR]=$2;e[NR]=$3} END{for(t=540;t<=660;t+=10){"
+        'k=sprintf("%02d:%02d",t/60,t%60);n=0;'
+        "for(i=1;i<=NR;i++)if(s[i]<=k&&k<e[i])n++;print k,n}}' input.txt",
+        "awk '{s[NR]=$2;e[NR]=$3} END{for(t=540;t<=660;t+=10){"
+        'k=sprintf("%02d:%02d",t/60,t%60);n=0;'
+        "for(i=1;i<=NR;i++)if(s[i]<=k&&k<=e[i])n++;print k,n}}' input.txt",
+    ),
+]
+
+
+def test_new_standard_solutions_and_mistakes() -> None:
+    # 各問の参照解答・別解は正解、初出優先・空白分割・閉区間等の誤解は不正解になる。
+    repository = build_problem_repository(
+        PROBLEMS / "v3", PROBLEMS / "image", PROBLEMS / "v3/manifest.json"
+    )
+    manager = ContainerManager(pool_size=1)
+    client = ShellgeiDockerClient(
+        container_manager=manager, max_concurrent=1, problem_repository=repository
+    )
+    judge = ShellgeiJudge(repository)
+    try:
+        manager.initialize_pool()
+        for number, alternative, mistake in CASES:
+            problem_id = f"STANDARD-{number:08}"
+            reference = repository.require(problem_id).definition.reference_solution
+            commands = [
+                (reference, True),
+                (alternative, True),
+                (mistake, False),
+            ]
+            if number == 59:
+                # 元からあるスペースまで>に変換する誤答を区別する。
+                commands.append(("expand -t 4 input.txt | tr ' ' '>'", False))
+            if number == 60:
+                # 記録に現れる時刻だけの出力と、11:00を落とす出力も区別する。
+                commands.extend(
+                    [
+                        (
+                            "awk '{print $2,1;print $3,-1}' input.txt | sort -k1,1 | "
+                            "awk 'NR>1&&$1!=t{print t,n} {t=$1;n+=$2} END{print t,n}'",
+                            False,
+                        ),
+                        (reference + " | head -n 12", False),
+                    ]
+                )
+            for command, accepted in commands:
+                execution = asyncio.run(client.run_with_timeout(command, problem_id))
+                context = (problem_id, command, execution)
+                assert execution.status is ExecutionStatus.COMPLETED, context
+                assert execution.exit_code == 0, context
+                assert execution.stderr == "", context
+                expected = (
+                    JudgeVerdict.ACCEPTED if accepted else JudgeVerdict.WRONG_ANSWER
+                )
+                assert judge.judge(execution, problem_id).verdict is expected, context
+    finally:
+        manager.shutdown_pool()
+        client.close()

backend/tests/integration/test_reservation_problem.py

@@ -26,12 +26,13 @@
 
 
 def test_reservation_solutions_and_common_mistakes_in_real_sandboxes() -> None:
-    # 参照解答・別解を受理し、逆の差・重複残存・部分一致による除外を実sandboxで区別する。
+    # 参照解答・別解を受理し、逆の差・重複残存・部分一致・初出判定前の予約抽出を区別する。
     repository = build_problem_repository(
         PROBLEMS / "v3", PROBLEMS / "image", PROBLEMS / "v3/manifest.json"
     )
     cases = [
         (repository.require(PROBLEM_ID).definition.reference_solution, True),
+        (f"comm -23 <({RESERVED} | sort -u) <({ARRIVED} | sort -u)", True),
         (
             'awk \'$1=="arrived"{a[$2]=1} $1=="reserved"{r[$2]=1} '
             "END{for(n in r)if(!(n in a))print n}' input.txt | sort",
@@ -47,6 +48,11 @@ def test_reservation_solutions_and_common_mistakes_in_real_sandboxes() -> None:
             False,
         ),
         (f"grep -Fv -f <({ARRIVED}) <({RESERVED} | sort -u)", False),
+        (
+            "cat input.txt | awk '{print $2,$1}' | sort -u | "
+            "awk '$2==\"reserved\" && !seen[$1]++{print $1}'",
+            False,
+        ),
     ]
     manager = ContainerManager(pool_size=1)
     client = ShellgeiDockerClient(

backend/tests/test_new_standard_problems.py

@@ -0,0 +1,85 @@
+"""通常53〜60の期待値を、参照シェル解とは独立した処理で検証する。"""
+
+from collections import Counter
+from pathlib import Path
+
+import pytest
+from PIL import Image
+
+from soj_shared.problem_schema import load_problem_definition
+
+
+PROBLEMS = Path(__file__).resolve().parents[2] / "problems"
+
+
+@pytest.mark.parametrize("number", range(53, 61))
+def test_new_standard_problem_expectations(number: int) -> None:
+    # 公開入力から期待値を独立計算し、空欄・区間境界・出力順まで完全一致で検査する。
+    problem_id = f"STANDARD-{number:08}"
+    definition = load_problem_definition(PROBLEMS / "v3" / f"{problem_id}.yaml")
+    assert definition.judge.type == "text"
+    assert definition.execution.exit_code == "zero"
+    assert definition.execution.stderr == "must_be_empty"
+    assert len(definition.execution.fixtures) == 1
+    assert definition.execution.fixtures[0].path == "input.txt"
+    content = definition.execution.fixtures[0].content
+    lines = content.splitlines()
+    result: list[str] = []
+    if number == 53:
+        devices = sorted({line.split()[0] for line in lines})
+        result = [
+            [line for line in lines if line.split()[0] == device][-1]
+            for device in devices
+        ]
+    elif number == 54:
+        measurements = [line.split() for line in lines]
+        result = [
+            f"{current[0]} {int(current[1]) - int(previous[1])}"
+            for previous, current in zip(measurements, measurements[1:])
+        ]
+    elif number == 55:
+        # 境界の行番号から区間を切り出し、参照解答の範囲状態とは別に確認する。
+        starts = [i for i, line in enumerate(lines) if line == "BEGIN"]
+        ends = [i for i, line in enumerate(lines) if line == "END"]
+        assert len(starts) == len(ends)
+        for start, end in zip(starts, ends):
+            assert start < end
+            result.extend(lines[start + 1 : end])
+    elif number == 56:
+        fields = [line.split("\t") for line in lines]
+        assert all(len(row) == 4 for row in fields)
+        result = [f"{row[0]}\t{row[2]}" for row in fields]
+    elif number == 57:
+        counts = Counter(line.rsplit(".", 1)[1].lower() for line in lines)
+        result = [f"{extension} {counts[extension]}" for extension in sorted(counts)]
+    elif number == 58:
+        records = [line.split() for line in lines]
+        result = sorted(
+            f"{team} {person} {document}"
+            for kind, team, person in records
+            for other_kind, other_team, document in records
+            if kind == "team" and other_kind == "file" and team == other_team
+        )
+    elif number == 59:
+        # 元のスペースを1文字幅の印で保護し、expandtabsが追加した空白だけを可視化する。
+        assert "\0" not in content
+        result = [
+            line.replace(" ", "\0").expandtabs(4).replace(" ", ">").replace("\0", " ")
+            for line in lines
+        ]
+    else:
+        intervals = [line.split()[1:] for line in lines]
+        assert all(start < end for start, end in intervals)
+        times = [
+            f"{minute // 60:02}:{minute % 60:02}" for minute in range(540, 661, 10)
+        ]
+        # イベントの累積を使わず、各時刻に区間を直接数える。
+        result = [
+            f"{time} {sum(start <= time < end for start, end in intervals)}"
+            for time in times
+        ]
+    assert definition.judge.expected_output == "\n".join(result) + "\n"
+    with Image.open(PROBLEMS / "image" / f"{problem_id}.jpg") as image:
+        image.load()
+        assert image.format == "JPEG"
+        assert image.convert("RGB").getextrema() == ((255, 255),) * 3