CardScoreDataServer/app/api/rating_report.py

from fastapi import APIRouter, HTTPException, Depends, Query
from mysql.connector.pooling import PooledMySQLConnection

import io
from app.core.minio_client import minio_client
from typing import List, Dict, Any, Optional
from PIL import Image

from app.core.logger import get_logger
from app.core.config import settings
from app.core.database_loader import get_db_connection
from app.crud import crud_card
from app.utils.scheme import ImageType

logger = get_logger(__name__)
router = APIRouter()


def _get_active_json(image_data: Any) -> Optional[Dict]:
    """获取有效的json数据，优先 modified_json"""
    if not image_data:
        return None

    # image_data 可能是 Pydantic 对象或 字典，做兼容处理
    if hasattr(image_data, "modified_json"):
        mj = image_data.modified_json
        dj = image_data.detection_json
    else:
        mj = image_data.get("modified_json")
        dj = image_data.get("detection_json")

    # 注意：根据 schema.py，这里读出来已经是 dict 了，不需要 json.loads
    # 如果数据库里存的是 null，读出来是 None
    if mj:
        return mj
    return dj


def _crop_defect_image(original_image_path_str: str, min_rect: List, output_filename: str) -> str:
    """
    通过 MinIO 切割缺陷图片为正方形并保存
    """
    try:
        # ★ 将进来的全路径 URL 剥离为相对路径 (如 /Data/xxx.jpg) 供 MinIO 读取
        rel_path = original_image_path_str.replace(settings.DATA_HOST_URL, "")
        rel_path = "/" + rel_path.lstrip('/\\')
        object_name = f"{settings.MINIO_BASE_PREFIX}{rel_path}"

        # 1. 从 MinIO 获取原图字节
        try:
            response = minio_client.get_object(settings.MINIO_BUCKET, object_name)
            image_bytes = response.read()
            response.close()
            response.release_conn()
        except Exception as e:
            logger.warning(f"从MinIO获取原图失败: {object_name} -> {e}")
            return ""

        # 2. 在内存中用 PIL 切图
        with Image.open(io.BytesIO(image_bytes)) as img:
            img_w, img_h = img.size
            center_x, center_y = min_rect[0]
            rect_w, rect_h = min_rect[1]
            side_length = max(max(rect_w, rect_h) * 1.5, 100)
            half_side = side_length / 2

            left, top = max(0, center_x - half_side), max(0, center_y - half_side)
            right, bottom = min(img_w, center_x + half_side), min(img_h, center_y + half_side)

            cropped_img = img.crop((left, top, right, bottom))

            # 3. 将切割后的图存入内存流，并上传到 MinIO
            out_bytes = io.BytesIO()
            cropped_img.save(out_bytes, format="JPEG", quality=95)
            out_bytes.seek(0)

            out_rel_path = f"/DefectImage/{output_filename}"
            out_object_name = f"{settings.MINIO_BASE_PREFIX}{out_rel_path}"

            minio_client.put_object(
                settings.MINIO_BUCKET,
                out_object_name,
                out_bytes,
                len(out_bytes.getvalue()),
                content_type="image/jpeg"
            )

            return settings.get_full_url(out_rel_path)

    except Exception as e:
        logger.error(f"切割并上传图片失败: {e}")
        return ""


@router.get("/generate", status_code=200, summary="生成评级报告数据")
def generate_rating_report(
        cardNo: str,
        db_conn: PooledMySQLConnection = Depends(get_db_connection)
):
    if not cardNo or not cardNo.strip():
        raise HTTPException(status_code=400, detail="cardNo 不能为空")

    # 根据cardNo 查询id
    try:
        with db_conn.cursor(buffered=True) as cursor:
            query_sql = f"SELECT id FROM {settings.DB_CARD_TABLE_NAME} WHERE cardNo = %s LIMIT 1"
            cursor.execute(query_sql, (cardNo,))
            row = cursor.fetchone()
    except Exception as e:
        logger.error(f"创建卡牌失败: {e}")
        raise HTTPException(status_code=500, detail="数据库查询失败。")

    if not row:
        raise HTTPException(
            status_code=404,
            detail=f"未找到卡牌编号为 {cardNo} 的相关记录"
        )

    card_id = row[0]

    top_n_defects = 3
    """
    根据 Card ID 生成评级报告 JSON
    """
    # 1. 获取卡片详情 (复用 Crud 逻辑，确保能拿到所有图片)
    card_data = crud_card.get_card_with_details(db_conn, card_id)
    if not card_data:
        raise HTTPException(status_code=404, detail="未找到该卡片信息")

    # 初始化返回结构
    response_data = {
        "backImageUrl": "",
        "frontImageUrl": "",
        "cardNo": cardNo,
        "centerBack": "",
        "centerFront": "",
        "measureLength": 0.0,
        "measureWidth": 0.0,
        "cornerBackNum": 0,
        "sideBackNum": 0,
        "surfaceBackNum": 0,
        "cornerFrontNum": 0,
        "sideFrontNum": 0,
        "surfaceFrontNum": 0,
        "scoreThreshold": float(card_data.get("detection_score") or 0),
        "evaluateNo": str(card_data.get("id")),
        "defectDetailList": []
    }

    # 临时列表用于收集所有缺陷，最后排序取 Top N
    all_defects_collected = []

    # 遍历图片寻找 Front Ring 和 Back Ring
    images = card_data.get("images", [])

    # 辅助字典：defect_type 到 统计字段 的映射
    defect_map_keys = {
        "front_ring": {
            "corner": "cornerFrontNum",
            "edge": "sideFrontNum",
            "face": "surfaceFrontNum"
        },
        "back_ring": {
            "corner": "cornerBackNum",
            "edge": "sideBackNum",
            "face": "surfaceBackNum"
        }
    }

    for img in images:
        img_type = img.image_type

        # 只处理环光图
        if img_type not in ["front_ring", "back_ring"]:
            continue

        # 设置主图 URL
        if img_type == "front_ring":
            response_data["frontImageUrl"] = img.image_path
        elif img_type == "back_ring":
            response_data["backImageUrl"] = img.image_path

        # 获取有效 JSON
        json_data = _get_active_json(img)
        if not json_data or "result" not in json_data:
            continue

        result_node = json_data["result"]

        # 1. 处理居中 (Center)
        center_inf = result_node.get("center_result", {}).get("box_result", {}).get("center_inference", {})
        if center_inf:
            # 格式: L/R=47/53, T/B=51/49 (取整)
            # center_inference 包含 center_left, center_right, center_top, center_bottom
            c_str = (
                f"L/R={int(round(center_inf.get('center_left', 0)))}/{int(round(center_inf.get('center_right', 0)))}, "
                f"T/B={int(round(center_inf.get('center_top', 0)))}/{int(round(center_inf.get('center_bottom', 0)))}"
            )
            if img_type == "front_ring":
                response_data["centerFront"] = c_str
                # 2. 处理尺寸 (仅从正面取，或者只要有就取) - mm 转 cm，除以 10，保留2位
                rw_mm = center_inf.get("real_width_mm", 0)
                rh_mm = center_inf.get("real_height_mm", 0)
                response_data["measureWidth"] = round(rw_mm / 10.0, 2)
                response_data["measureLength"] = round(rh_mm / 10.0, 2)
            else:
                response_data["centerBack"] = c_str

        # 2. 处理缺陷 (Defects)
        defects = result_node.get("defect_result", {}).get("defects", [])

        for defect in defects:
            # 过滤 edit_type == 'del'
            if defect.get("edit_type") == "del":
                continue

            d_type = defect.get("defect_type", "")  # corner, edge, face
            d_label = defect.get("label", "")  # scratch, wear, etc.

            # 统计数量
            count_key = defect_map_keys.get(img_type, {}).get(d_type)
            if count_key:
                response_data[count_key] += 1

            # 收集详细信息用于 Top N 列表
            # 需要保存：缺陷对象本身，图片路径，正反面标识
            side_str = "FRONT" if img_type == "front_ring" else "BACK"

            all_defects_collected.append({
                "defect_data": defect,
                "image_path": img.image_path,
                "side": side_str,
                "area": defect.get("actual_area", 0)
            })

    # 3. 处理 defectDetailList (Top N 切图)
    # 按实际面积从大到小排序
    all_defects_collected.sort(key=lambda x: x["area"], reverse=True)

    top_defects = all_defects_collected[:top_n_defects]

    final_defect_list = []
    for idx, item in enumerate(top_defects, start=1):
        defect = item["defect_data"]
        side = item["side"]
        original_img_path = item["image_path"]

        # 构造 ID
        d_id = idx  # 1, 2, 3

        # 构造文件名: {card_id}_{seq_id}.jpg
        filename = f"{card_id}_{d_id}.jpg"

        # 执行切图
        min_rect = defect.get("min_rect")
        defect_img_url = ""
        location_str = ""

        if min_rect and len(min_rect) == 3:
            # 切图并保存
            defect_img_url = _crop_defect_image(original_img_path, min_rect, filename)

            # 计算 Location (中心坐标)
            # min_rect[0] 是 [x, y]
            cx, cy = min_rect[0]
            location_str = f"{int(cx)},{int(cy)}"

        # 构造 Type 字符串: defect_type + label (大写)
        # 例如: defect_type="edge", label="wear" -> "EDGE WEAR"
        d_type_raw = defect.get("defect_type", "")
        d_label_raw = defect.get("label", "")
        type_str = f"{d_type_raw.upper()} {d_label_raw.upper()}".strip()

        final_defect_list.append({
            "id": d_id,
            "side": side,
            "location": location_str,
            "type": type_str,
            "defectImgUrl": defect_img_url
        })

    response_data["defectDetailList"] = final_defect_list

    return response_data