#!/usr/bin/env bash
# Fetch the POJ-104 C++ source corpus from Microsoft's CodeXGLUE repo.
#
# Layout afterwards:
#   bench/corpora/poj-104/dataset/
#     1/1.txt 2.txt ...   # 104 algorithm folders, each containing ~500 .txt
#     2/...
#     ...
#
# CodeXGLUE ships the files as `.txt` (the original submission archive used
# .txt to bypass file-type filtering). We rename them to `.cpp` in place so
# the standard run_bench.py glob picks them up.

set -euo pipefail

HERE="$(cd "$(dirname "$0")" && pwd)"
cd "$HERE"

if [[ -d dataset ]]; then
  echo "POJ-104 already fetched at $HERE/dataset; skipping clone."
  exit 0
fi

echo "Cloning CodeXGLUE (sparse: POJ-104 only)..."
TMPDIR="$(mktemp -d)"
trap 'rm -rf "$TMPDIR"' EXIT
git clone --depth 1 --filter=blob:none --sparse \
  https://github.com/microsoft/CodeXGLUE.git "$TMPDIR/codex"
( cd "$TMPDIR/codex" && \
  git sparse-checkout set Code-Code/Clone-detection-POJ-104/dataset )

mv "$TMPDIR/codex/Code-Code/Clone-detection-POJ-104/dataset" "$HERE/dataset-raw"

echo "Renaming .txt → .cpp so run_bench's *.cpp glob finds them..."
( cd "$HERE/dataset-raw" && find . -name '*.txt' -print0 \
    | xargs -0 -I{} sh -c 'mv "{}" "${1%.txt}.cpp"' _ {} )

mv "$HERE/dataset-raw" "$HERE/dataset"

count=$(find "$HERE/dataset" -name '*.cpp' | wc -l)
echo "POJ-104 ready: $count .cpp files in $HERE/dataset"
echo ""
echo "Next:"
echo "  python3 bench/gold/_generate_poj104.py"
echo "  python3 bench/run_bench.py --corpus $HERE/dataset --name poj-104 \\"
echo "      --tools fold,clang-tidy --detectors equiv --processes 4"
echo "  python3 bench/score.py --run-dir bench/results/<latest> \\"
echo "      --gold bench/gold/poj-104.json"