name: 'Restore Cache'
description: 'Restores cache from GitHub Releases. Supports both single and multiple cache paths/keys/buckets.'

inputs:
  cache_path:
    description: 'Single path where cache should be restored (use cache_paths for multiple)'
    required: false
    type: string
  cache_paths:
    description: 'Multiline list of paths where caches should be restored (one per line)'
    required: false
    type: string
    default: ""
  cache_key:
    description: 'Primary key to restore cache (use cache_keys for multiple)'
    required: false
    type: string
  cache_keys:
    description: 'Multiline list of primary keys to restore caches (one per line, must match cache_paths)'
    required: false
    type: string
    default: ""
  restore_keys:
    description: 'Multiline list of keys (fallback prefixes)'
    required: false
    type: string
    default: ""
  cache_bucket:
    description: 'Single tag name for the release (use cache_buckets for multiple)'
    required: false
    default: 'general-cache'
  cache_buckets:
    description: 'Multiline list of tag names for releases (one per line, must match cache_paths)'
    required: false
    type: string
    default: ""
  compression_level:
    description: 'Compression level (0-9). 0 = No compression, 1 = Fast, 19 = Best.'
    required: false
    type: number
    default: 6

outputs:
  cache-hit:
    description: 'Returns "true" if at least one cache was found and restored, "false" otherwise'
    value: ${{ steps.restore.outputs.cache-hit }}
  cache-key:
    description: 'The key that matched (primary or fallback). Multiple entries joined by comma.'
    value: ${{ steps.restore.outputs.cache-key }}
  cache-size:
    description: 'Size of restored cache in human-readable format. Multiple entries joined by comma.'
    value: ${{ steps.restore.outputs.cache-size }}

runs:
  using: 'composite'
  steps:
    - name: Detect and Download Cache
      id: restore
      shell: bash
      working-directory: ${{ github.workspace }}
      env:
        TARGET_REPO: "${{ github.repository }}"
      run: |
        set -euo pipefail

        # Validate and parse inputs
        PATHS_INPUT="${{ inputs.cache_paths }}"
        KEYS_INPUT="${{ inputs.cache_keys }}"
        BUCKETS_INPUT="${{ inputs.cache_buckets }}"
        
        # If multiple inputs are empty, fall back to single inputs
        if [ -z "$PATHS_INPUT" ]; then
          PATHS_INPUT="${{ inputs.cache_path }}"
        fi
        if [ -z "$KEYS_INPUT" ]; then
          KEYS_INPUT="${{ inputs.cache_key }}"
        fi
        if [ -z "$BUCKETS_INPUT" ]; then
          BUCKETS_INPUT="${{ inputs.cache_bucket }}"
        fi
        
        # Split inputs into arrays
        IFS=$'\n' read -rd '' -a PATHS_ARRAY <<<"$PATHS_INPUT" || true
        IFS=$'\n' read -rd '' -a KEYS_ARRAY <<<"$KEYS_INPUT" || true
        IFS=$'\n' read -rd '' -a BUCKETS_ARRAY <<<"$BUCKETS_INPUT" || true
        
        # Validate array lengths match
        if [[ ${#PATHS_ARRAY[@]} -ne ${#KEYS_ARRAY[@]} ]] || [[ ${#PATHS_ARRAY[@]} -ne ${#BUCKETS_ARRAY[@]} ]]; then
          echo "::error::Mismatch in array lengths: paths=${#PATHS_ARRAY[@]}, keys=${#KEYS_ARRAY[@]}, buckets=${#BUCKETS_ARRAY[@]}"
          exit 1
        fi
        
        ARIA2_OPTS=(
          "-x16" "-s16" "-k1M" "-j5" "--file-allocation=none"
          "--header=User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36"
          "--header=Accept: */*"
          "--header=Connection: keep-alive"
        )
        
        ARIA2_OPTS+=("--quiet" "--summary-interval=0" "--console-log-level=error")
        
        # Collect results
        OVERALL_HIT=false
        MATCHED_KEYS=()
        CACHE_SIZES=()
        
        # Process each cache entry
        for ((idx=0; idx<${#PATHS_ARRAY[@]}; idx++)); do
          CACHE_PATH="${PATHS_ARRAY[$idx]}"
          CACHE_KEY="${KEYS_ARRAY[$idx]}"
          CACHE_BUCKET="${BUCKETS_ARRAY[$idx]}"
          
          [ -z "$CACHE_PATH" ] && continue
          
          TAG_NAME="$CACHE_BUCKET"
          BASE_URL="https://github.com/${TARGET_REPO}/releases/download/$TAG_NAME"
          ASSETS_URL="https://github.com/${TARGET_REPO}/releases/expanded_assets/$TAG_NAME"
          
          echo "::group:: [$((idx+1))/${#PATHS_ARRAY[@]}] Cache: $CACHE_BUCKET - Fetching Asset List"
          HTTP_RESPONSE=$(curl -sL -A "Mozilla/5.0" -w "%{http_code}" "$ASSETS_URL" -o assets.html || echo "404")
          
          if [ "$HTTP_RESPONSE" -eq 200 ]; then
            ALL_ASSETS=$(grep -oP "download/$TAG_NAME/\K[^\"' ]+" assets.html | sort -u || echo "")
          else
            echo "Status $HTTP_RESPONSE: Failed to fetch assets from scraper endpoint."
            ALL_ASSETS=""
          fi
          rm -f assets.html
          
          if [ -z "$ALL_ASSETS" ]; then
            echo "No assets found in bucket '$TAG_NAME'. Skipping."
            echo "::endgroup::"
            continue
          fi
          
          echo "Successfully fetched asset list. Has $(echo "$ALL_ASSETS" | wc -l) assets."
          echo "::endgroup::"
          
          SEARCH_LIST=$(printf "%s\n%s" "$CACHE_KEY" "${{ inputs.restore_keys }}" | sed '/^$/d')
          
          FOUND=false
          while read -r KEY; do
            [ -z "$KEY" ] && continue
            echo "::group:: [$((idx+1))/${#PATHS_ARRAY[@]}] Searching Prefix: $KEY"
            
            MATCH=$(echo "$ALL_ASSETS" | grep -E "^cache-$KEY.*\.(tzst|tar)(\.part[a-z]{2})?$" | sort -r | head -n 1 || true)
            
            if [ -z "$MATCH" ]; then
                echo "Not found."
                echo "::endgroup::"
                continue
            fi
            
            BASE_FILENAME=$(echo "$MATCH" | sed -E 's/\.(tzst|tar)(\.part[a-z]{2})?$//')
            EXT=$(echo "$MATCH" | grep -oP '\\.(tzst|tar)' | head -n 1)
            
            [[ "$MATCH" == *".part"* ]] && IS_SPLIT=true || IS_SPLIT=false
            [ "$EXT" = ".tzst" ] && COMPRESSED=true || COMPRESSED=false
            FOUND=true
            
            echo "✅ Hit! Restoring $BASE_FILENAME$EXT"
            echo "::endgroup::"
            
            echo "::group:: [$((idx+1))/${#PATHS_ARRAY[@]}] Downloading Cache"
            
            download_asset() {
              local filename="$1"
              local url="$2"
              local max_retries=3
              local attempt=1
              
              while [ $attempt -le $max_retries ]; do
                echo "  Attempt $attempt: Downloading $filename..."
                # Prefer gh release download (reliable and atomic) when available
                rm -f "$filename"
                if command -v gh >/dev/null 2>&1; then
                  if timeout 10m gh release download "$TAG_NAME" --repo "$TARGET_REPO" --pattern "$filename" -D . >/dev/null 2>&1; then
                    # gh writes the file with the same name into the current directory
                    if [ -s "$filename" ]; then
                      return 0
                    fi
                  fi
                fi

                # Fallback to aria2c
                if timeout 10m aria2c "${ARIA2_OPTS[@]}" --retry-wait=10 --max-tries=10 -o "$filename" "$url"; then
                  # ensure file is non-empty
                  if [ -s "$filename" ]; then
                    return 0
                  else
                    echo "  ⚠️ Downloaded file is empty. Retrying..."
                    rm -f "$filename"
                  fi
                fi

                echo "  ⚠️ Download failed or timed out. Retrying in 5s..."
                sleep 5
                attempt=$((attempt + 1))
              done
              return 1
            }
            
            if [ "$IS_SPLIT" = "true" ]; then
              for part in {a..z}{a..z}; do
                PART_NAME="$BASE_FILENAME$EXT.part$part"
                if echo "$ALL_ASSETS" | grep -q "^$PART_NAME$"; then
                  if ! download_asset "$PART_NAME" "$BASE_URL/$PART_NAME"; then
                    echo "::error::Failed to download $PART_NAME after multiple retries."
                    exit 1
                  fi
                else break; fi
              done
            else
              if ! download_asset "$BASE_FILENAME$EXT" "$BASE_URL/$BASE_FILENAME$EXT"; then
                echo "::error::Failed to download $BASE_FILENAME$EXT after multiple retries."
                exit 1
              fi
            fi
            echo "::endgroup::"
            
            echo "::group:: [$((idx+1))/${#PATHS_ARRAY[@]}] Extracting Cache"
            mkdir -p "$CACHE_PATH"
            FINAL_ARCHIVE="$BASE_FILENAME$EXT"
            EXTRACT_DIR="$(dirname "$CACHE_PATH")"
            
            if [ "$IS_SPLIT" = "true" ]; then
              PARTS=$(ls "$FINAL_ARCHIVE.part"* | sort)
              if [ "$COMPRESSED" = "true" ]; then
                cat $PARTS | tar -I "zstd -d -T0" -xf - -C "$EXTRACT_DIR"
              else
                cat $PARTS | tar -xf - -C "$EXTRACT_DIR"
              fi
              rm -f $PARTS
            else
              if [ "$COMPRESSED" = "true" ]; then
                tar -I "zstd -d -T0" -xf "$FINAL_ARCHIVE" -C "$EXTRACT_DIR"
              else
                tar -xf "$FINAL_ARCHIVE" -C "$EXTRACT_DIR"
              fi
              rm -f "$FINAL_ARCHIVE"
            fi
            echo "✅ Restore complete."
            echo "::endgroup::"
            
            # Track results for this entry
            OVERALL_HIT=true
            MATCHED_KEYS+=("$KEY")
            CACHE_SIZE=$(du -sh "$CACHE_PATH" 2>/dev/null | awk '{print $1}' || echo "unknown")
            CACHE_SIZES+=("$CACHE_SIZE")
            # Provide a concise summary of restored path (top-level + one level deep)
            echo "--- Cache restore summary for: $CACHE_PATH ---"
            if [ -e "$CACHE_PATH" ]; then
              du -sh "$CACHE_PATH" || true
              du -h --max-depth=1 "$CACHE_PATH" 2>/dev/null | sort -hr | head -n 10 || true
              echo "Showing one-level-deep details for top entries (up to 5 each):"
              while read -r _entry; do
                entry_path=$(echo "$_entry" | awk '{print $2}')
                [ -z "$entry_path" ] && continue
                if [ "$entry_path" = "$CACHE_PATH" ]; then
                  continue
                fi
                echo "-> $entry_path :"
                du -h --max-depth=1 "$entry_path" 2>/dev/null | sort -hr | head -n 5 || true
              done < <(du -h --max-depth=1 "$CACHE_PATH" 2>/dev/null | sort -hr | head -n 10)
            fi
            break
          done <<< "$SEARCH_LIST"
          
          if [ "$FOUND" = "false" ]; then
              echo "⚠️ No cache matches found for bucket '$CACHE_BUCKET'. Proceeding with fresh run."
              MATCHED_KEYS+=("none")
              CACHE_SIZES+=("0B")
          fi
        done
        
        # Output results
        if [ "$OVERALL_HIT" = "true" ]; then
          echo "cache-hit=true" >> $GITHUB_OUTPUT
        else
          echo "cache-hit=false" >> $GITHUB_OUTPUT
        fi
        echo "cache-key=$(IFS=,; echo "${MATCHED_KEYS[*]}")" >> $GITHUB_OUTPUT
        echo "cache-size=$(IFS=,; echo "${CACHE_SIZES[*]}")" >> $GITHUB_OUTPUT