apply_patch.sh

#!/usr/bin/env bash
# =============================================================================
# apply_patch.sh — EQ-Bench3 日本語パッチ適用スクリプト
#
# 使用方法:
#   git clone --depth=1 https://github.com/EQ-bench/eqbench3.git
#   cd eqbench3
#   bash /path/to/apply_patch.sh
#
# このスクリプトは data/ と同じディレクトリに置いて実行してください。
#
# 行う変更:
#   [Fix-D1] data/ 以下の全日本語ファイルを配置（英語版は .en.bak に退避）
#   [Fix-B1] core/benchmark.py — STANDARD_ALLOWED_RUBRIC_CRITERIA を動的読込みに変更
#   [Fix-C1] core/conversation.py — セクション正規表現に日本語見出しを追加
# =============================================================================

set -euo pipefail

SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
EQBENCH_DIR="${1:-$(pwd)}"

echo ""
echo "======================================================"
echo "  EQ-Bench3 日本語版パッチ適用"
echo "  対象: ${EQBENCH_DIR}"
echo "======================================================"
echo ""

# ── ディレクトリ確認 ──────────────────────────────────────
if [ ! -f "${EQBENCH_DIR}/eqbench3.py" ]; then
    echo "[ERROR] eqbench3.py が見つかりません: ${EQBENCH_DIR}"
    echo "  EQ-Bench3 のルートディレクトリで実行してください。"
    exit 1
fi

cd "${EQBENCH_DIR}"

# ── [Fix-D1] データファイルの配置 ─────────────────────────
echo "[Fix-D1] 日本語データファイルを data/ に配置..."
DATA_SRC="${SCRIPT_DIR}/data"
DATA_DEST="${EQBENCH_DIR}/data"
mkdir -p "${DATA_DEST}"

FILES=(
    "scenario_prompts_ja.txt|scenario_prompts.txt"
    "scenario_notes_ja.txt|scenario_notes.txt"
    "scenario_master_prompt_ja.txt|scenario_master_prompt.txt"
    "scenario_master_prompt_message_drafting_ja.txt|scenario_master_prompt_message_drafting.txt"
    "scenario_master_prompt_analysis_ja.txt|scenario_master_prompt_analysis.txt"
    "rubric_scoring_prompt_ja.txt|rubric_scoring_prompt.txt"
    "rubric_scoring_prompt_analysis_ja.txt|rubric_scoring_prompt_analysis.txt"
    "rubric_scoring_criteria_ja.txt|rubric_scoring_criteria.txt"
    "rubric_scoring_criteria_analysis_ja.txt|rubric_scoring_criteria_analysis.txt"
    "pairwise_prompt_eqbench3_ja.txt|pairwise_prompt_eqbench3.txt"
    "pairwise_prompt_eqbench3_analysis_ja.txt|pairwise_prompt_eqbench3_analysis.txt"
    "debrief_prompt_ja.txt|debrief_prompt.txt"
)

OK_COUNT=0
FAIL_COUNT=0

for pair in "${FILES[@]}"; do
    SRC_FILE="${pair%%|*}"
    DEST_FILE="${pair##*|}"
    SRC_PATH="${DATA_SRC}/${SRC_FILE}"
    DEST_PATH="${DATA_DEST}/${DEST_FILE}"

    if [ ! -f "${SRC_PATH}" ]; then
        echo "  [WARN] ソースファイルが見つかりません: ${SRC_FILE}"
        FAIL_COUNT=$((FAIL_COUNT + 1))
        continue
    fi

    if [ -f "${DEST_PATH}" ] && [ ! -f "${DEST_PATH}.en.bak" ]; then
        cp "${DEST_PATH}" "${DEST_PATH}.en.bak"
        echo "  [BACKUP] ${DEST_FILE}.en.bak"
    fi

    cp "${SRC_PATH}" "${DEST_PATH}"
    echo "  [OK] ${DEST_FILE}"
    OK_COUNT=$((OK_COUNT + 1))
done

echo "  配置完了: ${OK_COUNT}/${#FILES[@]} ファイル"
[ "${FAIL_COUNT}" -gt 0 ] && echo "  [WARN] ${FAIL_COUNT} ファイルが見つかりませんでした"

# ── [Fix-B1] benchmark.py パッチ ──────────────────────────
echo ""
echo "[Fix-B1] core/benchmark.py にパッチ適用中..."
if grep -q "\[JA-PATCH Fix-B1\]" "${EQBENCH_DIR}/core/benchmark.py"; then
    echo "  [SKIP] Fix-B1 は既に適用済みです"
else
    cp "${EQBENCH_DIR}/core/benchmark.py" "${EQBENCH_DIR}/core/benchmark.py.orig"
    patch -p1 < "${SCRIPT_DIR}/data/benchmark.patch"
    echo "  [OK] Fix-B1 適用完了"
fi

# ── [Fix-C1] conversation.py パッチ ───────────────────────
echo ""
echo "[Fix-C1] core/conversation.py にパッチ適用中..."
if grep -q "\[JA-PATCH Fix-C1\]" "${EQBENCH_DIR}/core/conversation.py"; then
    echo "  [SKIP] Fix-C1 は既に適用済みです"
else
    cp "${EQBENCH_DIR}/core/conversation.py" "${EQBENCH_DIR}/core/conversation.py.orig"
    patch -p1 < "${SCRIPT_DIR}/data/conversation.patch"
    echo "  [OK] Fix-C1 適用完了"
fi

# ── 検証 ──────────────────────────────────────────────────
echo ""
echo "[verify] パッチ適用を検証中..."
ERRORS=0

_check() {
    if grep -q "$1" "$2"; then
        echo "  [OK] $(basename $2) — $3"
    else
        echo "  [ERROR] $(basename $2) — $3"
        ERRORS=$((ERRORS + 1))
    fi
}

_check "{scenario_prompt}"  "${DATA_DEST}/scenario_master_prompt.txt"              "{scenario_prompt} プレースホルダー"
_check "{scenario_prompt}"  "${DATA_DEST}/scenario_master_prompt_message_drafting.txt" "{scenario_prompt} プレースホルダー"
_check "demonstrated_empathy" "${DATA_DEST}/rubric_scoring_criteria.txt"           "英語キー"
_check "{conversation_history_A}" "${DATA_DEST}/pairwise_prompt_eqbench3.txt"      "{conversation_history_A} プレースホルダー"
_check "JA-PATCH Fix-B1"    "${EQBENCH_DIR}/core/benchmark.py"                     "Fix-B1 パッチ確認"
_check "JA-PATCH Fix-C1"    "${EQBENCH_DIR}/core/conversation.py"                  "Fix-C1 パッチ確認"

echo ""
if [ "${ERRORS}" -eq 0 ]; then
    echo "======================================================"
    echo "  ✅ パッチ適用完了 — エラーなし"
    echo "======================================================"
else
    echo "======================================================"
    echo "  ❌ ${ERRORS} 件のエラーが発生しました"
    echo "======================================================"
    exit 1
fi

echo ""
echo "【次のステップ】"
echo "  python eqbench3.py --test-model <model> --judge-model openai/gpt-oss-120b --no-elo --save-interval 1"
echo ""