feat: 新增 CAD 元素识别模块，支持柱子/文字/墙体候选检测

- 新增 element_detector.py，包含 ColumnDetector、TextExtractor、WallCandidateDetector
- ColumnDetector：通过 INSERT 块名关键词识别柱子候选
- TextExtractor：提取 TEXT/MTEXT 内容、坐标、图层
- WallCandidateDetector：按长度阈值筛选 LINE/LWPOLYLINE，排除标注图层
- 采用 OOP 设计，所有规则可配置，保留原始 handle 追溯
- 输出 element_detection.json 和 element_report.txt
- 更新 main.py 集成元素识别步骤

docs/tasks/task_detail_2026_07_08.md

@@ -15,3 +15,27 @@
        - 布局/图纸空间列表
     4. 运行程序验证，成功解析目标 DXF 文件，输出 20 个实体（9 个 DIMENSION、6 个 MTEXT、5 个 LWPOLYLINE）、13 个块定义、3 个布局。
 - **执行结果**: 程序正常运行，完整输出了 samples/decimal-inch-drawing-sheet-size-ASME-Y14.1-cad-block-dwg.dxf 的全部结构和内容信息。
+
+## 会话 ID: 2
+- [2026-07-08 13:58]
+- **执行原因**: 将 main.py 单文件脚本拆分为模块化结构（layer_analysis.py + entity_analysis.py + main.py），面向第二阶段墙体识别、面积计算等功能扩展。
+- **执行过程**:
+    1. 创建 layer_analysis.py：遍历图层统计实体数量和类型，基于常见中英文命名规则进行角色推断（墙体/门窗/标注/文字/柱/设备/填充等）。
+    2. 创建 entity_analysis.py：实体类型统计、INSERT 按块名聚合（含插入坐标/旋转/缩放）、TEXT/MTEXT 文本内容提取，保留 extract_entity_props() 逻辑。
+    3. 重构 main.py：读取 DXF 后调度两个分析模块，组装输出 output/cad_analysis.json，保留终端摘要打印。
+    4. 创建 README.md 使用说明文档。
+    5. 运行验证：成功解析洗浴中心C-48.dxf（719个实体、12个图层、11种块、71条文本），JSON 输出格式正确。
+- **执行结果**: 模块化拆分完成，代码结构清晰，JSON 输出符合 Task1 规范，为后续墙体识别等功能扩展奠定基础。
+
+## 会话 ID: 3
+- [2026-07-08 14:39]
+- **执行原因**: 实现 Task2 —— CAD 基础构件候选识别 Demo，从 DXF 数据中识别柱子、文字和墙体候选。
+- **执行过程**:
+    1. 创建 element_detector.py，采用 OOP 设计，包含三个独立识别器：
+       - ColumnDetector：通过 INSERT 块名关键词（柱/column/COL）识别柱子候选
+       - TextExtractor：提取所有 TEXT/MTEXT 的内容、坐标、图层
+       - WallCandidateDetector：分析 LINE/LWPOLYLINE，按长度阈值（>=500）筛选，排除标注图层（DIM/TEXT/DEFPOINTS）
+    2. ElementDetector 作为编排器统一调度，所有规则可配置。
+    3. 更新 main.py 集成元素识别步骤，输出 element_detection.json 和 element_report.txt。
+    4. 修复 ezdxf Vec3 切片兼容问题。
+- **执行结果**: 成功识别 6 个柱子候选（柱子01）、71 条文字、231 条墙体候选（179 LINE + 52 LWPOLYLINE），输出文件格式符合 Task2 规范。

experiments/dxf-parser/README.md

@@ -0,0 +1,83 @@
+# DXF 解析工具
+
+解析 DXF 格式的 CAD 图纸文件，输出结构化的图层分析、实体统计、块引用和文本信息。
+
+## 目录结构
+
+```
+experiments/dxf-parser/
+├── main.py              # 入口脚本
+├── layer_analysis.py    # 图层分析模块
+├── entity_analysis.py   # 实体分析模块
+├── requirements.txt     # 依赖
+├── output/              # 输出目录
+│   └── {文件名}_cad_analysis.json
+└── README.md
+```
+
+## 安装依赖
+
+```bash
+cd experiments/dxf-parser
+pip install -r requirements.txt
+```
+
+## 使用方式
+
+1. 将要解析的 DXF 文件放入 `samples/` 目录
+2. 修改 `main.py` 中的 `DXF_FILE` 变量指向目标文件
+3. 运行：
+
+```bash
+python main.py
+```
+
+## 输出说明
+
+终端会打印解析摘要（图层、实体类型统计、块引用、文本内容），同时在 `output/{文件名}_cad_analysis.json` 中生成结构化 JSON：
+
+```json
+{
+  "file_info": {
+    "filename": "洗浴中心C-48.dxf",
+    "dxf_version": "AC1032",
+    "encoding": "gbk",
+    "units": null,
+    "extents": { "min": [...], "max": [...] },
+    "total_entities": 719
+  },
+  "layers": [
+    {
+      "name": "DOOR",
+      "entity_count": 52,
+      "entity_types": ["INSERT", "LINE"],
+      "suggested_role": "门窗"
+    }
+  ],
+  "entities": {
+    "LINE": 321,
+    "TEXT": 71
+  },
+  "blocks": [
+    {
+      "block_name": "A$C6B90F135",
+      "count": 8,
+      "inserts": [...]
+    }
+  ],
+  "texts": [
+    {
+      "handle": "2A",
+      "layer": "TEXT",
+      "type": "TEXT",
+      "content": "办公室",
+      "insert": [150, 300]
+    }
+  ]
+}
+```
+
+## 模块说明
+
+- **layer_analysis.py**：遍历所有图层，统计每层实体数量和类型，基于常见命名规则推断图层角色（仅作建议）
+- **entity_analysis.py**：统计模型空间中各实体类型数量，对 INSERT 实体按块名聚合，提取所有 TEXT/MTEXT 文本内容

experiments/dxf-parser/element_detector.py

@@ -0,0 +1,445 @@
+"""CAD 基础构件候选识别模块 —— 从 DXF 数据中识别柱子、文字和墙体候选。
+
+设计原则：
+- 面向对象，每个识别模块独立
+- 所有规则可配置
+- 保留原始实体 handle 方便追溯
+- 不假设图层命名规范，仅做启发式匹配
+"""
+
+import math
+import json
+from pathlib import Path
+from typing import Optional
+
+import ezdxf
+
+
+# ============================================================
+# 配置常量（可修改）
+# ============================================================
+
+# 柱子识别：INSERT 块名中包含以下关键词之一
+COLUMN_KEYWORDS = ["柱", "column", "COL"]
+
+# 墙体候选：最小线段长度（图纸单位，通常为毫米）
+WALL_MIN_LENGTH = 500
+
+# 墙体候选：排除的图层关键词（标注、文字等非建筑图层）
+WALL_EXCLUDE_LAYERS = ["DIM", "TEXT", "DEFPOINTS", "标注", "NOTE", "PUB_DIM", "AXIS"]
+
+# 墙体候选：优先包含的图层关键词（空列表 = 不限制）
+WALL_INCLUDE_LAYERS = []
+
+
+# ============================================================
+# 柱子识别器
+# ============================================================
+
+class ColumnDetector:
+    """从 INSERT 实体中识别候选柱子。
+
+    规则：块名包含柱/column/COL 等关键词。
+    """
+
+    def __init__(self, keywords: Optional[list[str]] = None):
+        """
+        Args:
+            keywords: 块名匹配关键词列表，默认使用 COLUMN_KEYWORDS
+        """
+        self.keywords = keywords or COLUMN_KEYWORDS
+
+    def detect(self, msp) -> list[dict]:
+        """遍历模型空间，检测柱子候选。
+
+        Args:
+            msp: ezdxf 模型空间对象
+
+        Returns:
+            list[dict]: 柱子候选列表
+        """
+        columns = []
+        for entity in msp:
+            if entity.dxftype() != "INSERT":
+                continue
+            block_name = entity.dxf.name
+            if self._match(block_name):
+                insert = entity.dxf.insert
+                columns.append({
+                    "name": block_name,
+                    "position": {
+                        "x": round(insert[0], 4),
+                        "y": round(insert[1], 4),
+                    },
+                    "source": "INSERT",
+                    "handle": entity.dxf.handle,
+                    "layer": entity.dxf.layer,
+                    "rotation": round(entity.dxf.rotation, 2),
+                    "scale": [
+                        round(entity.dxf.xscale, 2),
+                        round(entity.dxf.yscale, 2),
+                        round(entity.dxf.zscale, 2),
+                    ],
+                })
+        return columns
+
+    def _match(self, block_name: str) -> bool:
+        """检查块名是否匹配柱子关键词。"""
+        upper = block_name.upper()
+        for kw in self.keywords:
+            if kw.upper() in upper or kw in block_name:
+                return True
+        return False
+
+
+# ============================================================
+# 文字提取器
+# ============================================================
+
+class TextExtractor:
+    """提取所有 TEXT / MTEXT 实体的文本内容及其坐标。"""
+
+    def extract(self, msp) -> list[dict]:
+        """提取模型空间中所有文字实体。
+
+        Args:
+            msp: ezdxf 模型空间对象
+
+        Returns:
+            list[dict]: 文字信息列表
+        """
+        texts = []
+        for entity in msp:
+            dxftype = entity.dxftype()
+            if dxftype not in ("TEXT", "MTEXT"):
+                continue
+
+            content = entity.plain_text() if dxftype == "MTEXT" else entity.dxf.text
+            insert = entity.dxf.insert if hasattr(entity.dxf, "insert") else None
+
+            item = {
+                "content": content,
+                "type": dxftype,
+                "handle": entity.dxf.handle,
+                "layer": entity.dxf.layer,
+            }
+            if insert is not None:
+                item["x"] = round(insert[0], 4)
+                item["y"] = round(insert[1], 4)
+            if hasattr(entity.dxf, "height"):
+                item["height"] = round(entity.dxf.height, 2)
+
+            texts.append(item)
+        return texts
+
+
+# ============================================================
+# 墙体候选识别器
+# ============================================================
+
+class WallCandidateDetector:
+    """从 LINE / LWPOLYLINE 中识别墙体候选线段。
+
+    规则：
+    1. 线段长度 >= min_length
+    2. 所在图层不是标注类图层
+    3. 可选：限制在主要建筑图层
+    """
+
+    def __init__(
+        self,
+        min_length: float = WALL_MIN_LENGTH,
+        exclude_layers: Optional[list[str]] = None,
+        include_layers: Optional[list[str]] = None,
+    ):
+        """
+        Args:
+            min_length: 最小线段长度阈值
+            exclude_layers: 要排除的图层关键词列表
+            include_layers: 仅包含的图层关键词列表（空 = 不限制）
+        """
+        self.min_length = min_length
+        self.exclude_layers = exclude_layers or WALL_EXCLUDE_LAYERS
+        self.include_layers = include_layers or WALL_INCLUDE_LAYERS
+
+    def detect(self, msp) -> list[dict]:
+        """检测墙体候选线段。"""
+        candidates = []
+        for entity in msp:
+            dxftype = entity.dxftype()
+            layer = entity.dxf.layer
+
+            if self._is_excluded_layer(layer):
+                continue
+
+            if dxftype == "LINE":
+                start = entity.dxf.start
+                end = entity.dxf.end
+                length = math.dist([start[0], start[1]], [end[0], end[1]])
+                if length >= self.min_length:
+                    candidates.append({
+                        "entity_type": "LINE",
+                        "handle": entity.dxf.handle,
+                        "layer": layer,
+                        "length": round(length, 2),
+                        "start": [round(start[0], 4), round(start[1], 4)],
+                        "end": [round(end[0], 4), round(end[1], 4)],
+                    })
+
+            elif dxftype == "LWPOLYLINE":
+                try:
+                    pts = entity.get_points()
+                except Exception:
+                    continue
+                if len(pts) < 2:
+                    continue
+                # 计算总长度和最长段
+                total_length = 0.0
+                max_seg_length = 0.0
+                segments = []
+                for i in range(len(pts) - 1):
+                    seg_len = math.dist([pts[i][0], pts[i][1]], [pts[i + 1][0], pts[i + 1][1]])
+                    total_length += seg_len
+                    if seg_len > max_seg_length:
+                        max_seg_length = seg_len
+                    segments.append({
+                        "start": [round(pts[i][0], 4), round(pts[i][1], 4)],
+                        "end": [round(pts[i + 1][0], 4), round(pts[i + 1][1], 4)],
+                        "length": round(seg_len, 2),
+                    })
+
+                # 至少最长段满足阈值才算候选
+                if max_seg_length >= self.min_length:
+                    candidates.append({
+                        "entity_type": "LWPOLYLINE",
+                        "handle": entity.dxf.handle,
+                        "layer": layer,
+                        "vertex_count": len(pts),
+                        "closed": entity.closed,
+                        "total_length": round(total_length, 2),
+                        "max_segment_length": round(max_seg_length, 2),
+                        "segments": [s for s in segments if s["length"] >= self.min_length],
+                        "points": [[round(p[0], 4), round(p[1], 4)] for p in pts],
+                    })
+
+        return candidates
+
+    def _is_excluded_layer(self, layer_name: str) -> bool:
+        """检查图层是否属于排除范围（标注类图层）。"""
+        upper = layer_name.upper()
+        for kw in self.exclude_layers:
+            if kw.upper() in upper or kw in layer_name:
+                return True
+        # 如果配置了 include 列表，则仅包含匹配的图层
+        if self.include_layers:
+            for kw in self.include_layers:
+                if kw.upper() in upper or kw in layer_name:
+                    return False
+            return True  # 不在 include 列表中则排除
+        return False
+
+
+# ============================================================
+# 主检测器（编排）
+# ============================================================
+
+class ElementDetector:
+    """编排各识别器，输出统一的检测结果。"""
+
+    def __init__(
+        self,
+        column_keywords: Optional[list[str]] = None,
+        wall_min_length: float = WALL_MIN_LENGTH,
+        wall_exclude_layers: Optional[list[str]] = None,
+        wall_include_layers: Optional[list[str]] = None,
+    ):
+        self.column_detector = ColumnDetector(keywords=column_keywords)
+        self.text_extractor = TextExtractor()
+        self.wall_detector = WallCandidateDetector(
+            min_length=wall_min_length,
+            exclude_layers=wall_exclude_layers,
+            include_layers=wall_include_layers,
+        )
+
+    def detect_all(self, doc) -> dict:
+        """执行全部识别，返回结构化结果。
+
+        Args:
+            doc: ezdxf Drawing 对象
+
+        Returns:
+            dict: {"columns": [...], "texts": [...], "wall_candidates": [...]}
+        """
+        msp = doc.modelspace()
+        return {
+            "columns": self.column_detector.detect(msp),
+            "texts": self.text_extractor.extract(msp),
+            "wall_candidates": self.wall_detector.detect(msp),
+        }
+
+
+# ============================================================
+# 报告生成
+# ============================================================
+
+def generate_report(result: dict) -> str:
+    """根据检测结果生成纯文本统计报告。
+
+    Args:
+        result: detect_all() 的返回值
+
+    Returns:
+        str: 格式化报告文本
+    """
+    lines = []
+    lines.append("=" * 50)
+    lines.append("  CAD 元素分析报告")
+    lines.append("=" * 50)
+    lines.append("")
+
+    # 柱子
+    columns = result["columns"]
+    lines.append(f"【柱子候选】")
+    lines.append(f"  数量: {len(columns)}")
+    if columns:
+        # 按块名分组统计
+        block_counts = {}
+        for col in columns:
+            name = col["name"]
+            block_counts[name] = block_counts.get(name, 0) + 1
+        for name, count in sorted(block_counts.items()):
+            lines.append(f"    - {name}: {count} 个")
+    lines.append("")
+
+    # 文字
+    texts = result["texts"]
+    lines.append(f"【文字】")
+    lines.append(f"  数量: {len(texts)}")
+    if texts:
+        # 按图层分组统计
+        layer_counts = {}
+        for t in texts:
+            ly = t["layer"]
+            layer_counts[ly] = layer_counts.get(ly, 0) + 1
+        for ly, count in sorted(layer_counts.items()):
+            lines.append(f"    - 图层 {ly}: {count} 条")
+    lines.append("")
+
+    # 墙体候选
+    walls = result["wall_candidates"]
+    lines.append(f"【墙体候选】")
+    lines.append(f"  数量: {len(walls)}")
+    if walls:
+        # 按类型统计
+        type_counts = {}
+        layer_counts = {}
+        lengths = []
+        for w in walls:
+            etype = w["entity_type"]
+            type_counts[etype] = type_counts.get(etype, 0) + 1
+            ly = w["layer"]
+            layer_counts[ly] = layer_counts.get(ly, 0) + 1
+            # 获取长度
+            if etype == "LINE":
+                lengths.append(w["length"])
+            elif etype == "LWPOLYLINE":
+                lengths.append(w["max_segment_length"])
+
+        lines.append("  按类型:")
+        for etype, count in sorted(type_counts.items()):
+            lines.append(f"    - {etype}: {count} 条")
+
+        lines.append("  按图层:")
+        for ly, count in sorted(layer_counts.items(), key=lambda x: -x[1]):
+            lines.append(f"    - 图层 {ly}: {count} 条")
+
+        if lengths:
+            lines.append(f"  最长段: {max(lengths):.2f}")
+            lines.append(f"  最短段: {min(lengths):.2f}")
+            lines.append(f"  平均长: {sum(lengths) / len(lengths):.2f}")
+
+    lines.append("")
+    lines.append("=" * 50)
+    lines.append("  报告结束")
+    lines.append("=" * 50)
+    return "\n".join(lines)
+
+
+# ============================================================
+# 入口
+# ============================================================
+
+def run_detection(
+    dxf_path: str,
+    output_dir: str,
+    column_keywords: Optional[list[str]] = None,
+    wall_min_length: float = WALL_MIN_LENGTH,
+    wall_exclude_layers: Optional[list[str]] = None,
+    wall_include_layers: Optional[list[str]] = None,
+) -> dict:
+    """执行元素检测并输出 JSON 和报告文件。
+
+    Args:
+        dxf_path: DXF 文件路径
+        output_dir: 输出目录路径
+        column_keywords: 柱子识别关键词
+        wall_min_length: 墙体最小长度阈值
+        wall_exclude_layers: 墙体排除图层关键词
+        wall_include_layers: 墙体限定图层关键词
+
+    Returns:
+        dict: 检测结果
+    """
+    doc = ezdxf.readfile(dxf_path)
+    filename = Path(dxf_path).stem
+
+    detector = ElementDetector(
+        column_keywords=column_keywords,
+        wall_min_length=wall_min_length,
+        wall_exclude_layers=wall_exclude_layers,
+        wall_include_layers=wall_include_layers,
+    )
+    result = detector.detect_all(doc)
+
+    out_dir = Path(output_dir)
+    out_dir.mkdir(parents=True, exist_ok=True)
+
+    # 输出 JSON
+    json_path = out_dir / f"{filename}_element_detection.json"
+    json_path.write_text(
+        json.dumps(result, ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    print(f"检测结果已保存到: {json_path}")
+
+    # 输出报告
+    report_path = out_dir / f"{filename}_element_report.txt"
+    report = generate_report(result)
+    report_path.write_text(report, encoding="utf-8")
+    print(f"分析报告已保存到: {report_path}")
+
+    # 打印摘要
+    print(f"\n  柱子候选 : {len(result['columns'])} 个")
+    print(f"  文字     : {len(result['texts'])} 条")
+    print(f"  墙体候选 : {len(result['wall_candidates'])} 条")
+    print(report)
+
+    return result
+
+
+# ============================================================
+# 独立运行入口
+# ============================================================
+
+if __name__ == "__main__":
+    import sys
+    BASE_DIR = Path(__file__).parent.parent.parent
+    SAMPLE_DIR = BASE_DIR / "samples"
+    OUTPUT_DIR = Path(__file__).parent / "output"
+
+    dxf_file = SAMPLE_DIR / "洗浴中心C-48.dxf"
+    if not dxf_file.exists():
+        print(f"错误：文件不存在 -> {dxf_file}")
+        sys.exit(1)
+
+    run_detection(str(dxf_file), str(OUTPUT_DIR))

experiments/dxf-parser/entity_analysis.py

@@ -0,0 +1,148 @@
+"""实体分析模块 —— 统计模型空间中各实体类型的数量、INSERT 块聚合、文本提取。"""
+
+
+def _extract_entity_props(entity):
+    """提取单个实体的关键属性为 JSON 友好格式。"""
+    dxftype = entity.dxftype()
+    props = {}
+
+    if dxftype == "LINE":
+        props["start"] = list(entity.dxf.start)
+        props["end"] = list(entity.dxf.end)
+    elif dxftype == "CIRCLE":
+        props["center"] = list(entity.dxf.center)
+        props["radius"] = round(entity.dxf.radius, 4)
+    elif dxftype == "ARC":
+        props["center"] = list(entity.dxf.center)
+        props["radius"] = round(entity.dxf.radius, 4)
+        props["start_angle"] = round(entity.dxf.start_angle, 1)
+        props["end_angle"] = round(entity.dxf.end_angle, 1)
+    elif dxftype == "LWPOLYLINE":
+        pts = entity.get_points()
+        props["vertex_count"] = len(entity)
+        props["closed"] = entity.closed
+        props["points"] = [[round(p[0], 4), round(p[1], 4)] for p in pts]
+    elif dxftype == "POLYLINE":
+        verts = list(entity.vertices())
+        props["vertex_count"] = len(verts)
+        props["points"] = [[round(v.dxf.location[0], 4), round(v.dxf.location[1], 4)] for v in verts]
+    elif dxftype == "INSERT":
+        props["block_name"] = entity.dxf.name
+        props["insert"] = list(entity.dxf.insert)
+        props["scale"] = [round(entity.dxf.xscale, 2), round(entity.dxf.yscale, 2), round(entity.dxf.zscale, 2)]
+        props["rotation"] = round(entity.dxf.rotation, 2)
+    elif dxftype in ("MTEXT", "TEXT"):
+        text = entity.plain_text() if dxftype == "MTEXT" else entity.dxf.text
+        props["text"] = text
+        if hasattr(entity.dxf, "insert"):
+            props["insert"] = list(entity.dxf.insert)
+        if hasattr(entity.dxf, "height"):
+            props["height"] = round(entity.dxf.height, 2)
+    elif dxftype == "DIMENSION":
+        props["text"] = entity.dxf.text
+        if hasattr(entity.dxf, "dimtype"):
+            props["dimtype"] = entity.dxf.dimtype
+    elif dxftype == "HATCH":
+        props["pattern"] = entity.dxf.pattern_name
+    elif dxftype in ("SOLID", "TRACE"):
+        for vi in range(4):
+            try:
+                v = getattr(entity.dxf, f"vtx{vi}")
+                props[f"vtx{vi}"] = list(v) if v else None
+            except Exception:
+                break
+    elif dxftype == "POINT":
+        props["location"] = list(entity.dxf.location)
+    elif dxftype == "ELLIPSE":
+        props["center"] = list(entity.dxf.center)
+        props["major_axis"] = list(entity.dxf.major_axis)
+        props["ratio"] = round(entity.dxf.ratio, 4)
+    elif dxftype == "SPLINE":
+        props["fit_points"] = [list(p) for p in entity.fit_points]
+    elif dxftype == "3DFACE":
+        for vi in range(4):
+            try:
+                v = getattr(entity.dxf, f"vtx{vi}")
+                props[f"vtx{vi}"] = list(v) if v else None
+            except Exception:
+                break
+    else:
+        for attr in entity.dxf.all_existing_dxf_attribs():
+            if attr in ("handle", "owner", "layer", "linetype", "color", "lineweight"):
+                continue
+            val = getattr(entity.dxf, attr)
+            if val is not None:
+                props[attr] = list(val) if hasattr(val, "__iter__") and not isinstance(val, str) else val
+
+    return props
+
+
+def analyze_entities(doc) -> dict:
+    """分析模型空间中的所有实体。
+
+    Args:
+        doc: ezdxf Drawing 对象
+
+    Returns:
+        dict: {
+            "total_entities": int,
+            "type_counts": dict,        # {dxftype: count}
+            "blocks": list[dict],        # INSERT 按块名聚合
+            "texts": list[dict],         # TEXT/MTEXT 文本列表
+        }
+    """
+    msp = doc.modelspace()
+    type_counts = {}
+
+    # INSERT 按块名聚合
+    block_map = {}  # block_name -> list of insert_info
+    texts = []
+
+    for entity in msp:
+        dxftype = entity.dxftype()
+        type_counts[dxftype] = type_counts.get(dxftype, 0) + 1
+
+        if dxftype == "INSERT":
+            block_name = entity.dxf.name
+            insert_info = {
+                "handle": entity.dxf.handle,
+                "layer": entity.dxf.layer,
+                "insert_point": list(entity.dxf.insert),
+                "rotation": round(entity.dxf.rotation, 2),
+                "scale": [
+                    round(entity.dxf.xscale, 2),
+                    round(entity.dxf.yscale, 2),
+                    round(entity.dxf.zscale, 2),
+                ],
+            }
+            if block_name not in block_map:
+                block_map[block_name] = []
+            block_map[block_name].append(insert_info)
+
+        elif dxftype in ("TEXT", "MTEXT"):
+            content = entity.plain_text() if dxftype == "MTEXT" else entity.dxf.text
+            text_info = {
+                "handle": entity.dxf.handle,
+                "layer": entity.dxf.layer,
+                "type": dxftype,
+                "content": content,
+            }
+            if hasattr(entity.dxf, "insert"):
+                text_info["insert"] = list(entity.dxf.insert)
+            texts.append(text_info)
+
+    # 块聚合结果
+    blocks = []
+    for block_name, inserts in sorted(block_map.items()):
+        blocks.append({
+            "block_name": block_name,
+            "count": len(inserts),
+            "inserts": inserts,
+        })
+
+    return {
+        "total_entities": sum(type_counts.values()),
+        "type_counts": type_counts,
+        "blocks": blocks,
+        "texts": texts,
+    }

experiments/dxf-parser/layer_analysis.py

@@ -0,0 +1,94 @@
+"""图层分析模块 —— 统计每层实体数量与类型，并基于图层名称进行角色推断。"""
+
+# 角色推断规则（启发式，基于常见中文/英文命名）
+_ROLE_RULES = [
+    (["WALL", "墙", "W", "A-WALL", "A-WALL-"], "墙体"),
+    (["DOOR", "门", "窗", "WINDOW", "D", "M"], "门窗"),
+    (["DIM", "标注", "DEFPOINTS", "PUB_DIM", "PUB_DIM"], "标注"),
+    (["TEXT", "文字", "NOTE"], "文字"),
+    (["COLUMN", "柱", "Z"], "柱"),
+    (["E", "EQUIP", "设备"], "设备"),
+    (["HATCH", "填充", "BH"], "填充"),
+    (["AXIS", "轴", "AXIS"], "轴网"),
+    (["GRID", "网格"], "网格"),
+    (["FURN", "家具", "FURNITURE"], "家具"),
+    (["BEAM", "梁", "B"], "梁"),
+    (["SLAB", "板"], "板"),
+    (["STAIR", "楼梯", "S"], "楼梯"),
+    (["ROOF", "屋顶", "R"], "屋顶"),
+    (["PIPE", "管", "P"], "管道"),
+    (["DUCT", "风管"], "风管"),
+    (["ELEC", "电气", "E-"], "电气"),
+    (["PLUMB", "给排水", "P-"], "给排水"),
+    (["FOUND", "基础", "F"], "基础"),
+    (["SITE", "场地", "总图"], "场地"),
+    (["CEIL", "天花", "吊顶", "C"], "天花"),
+    (["FLOOR", "地面", "FL"], "地面"),
+    (["SECTION", "剖面", "SECT"], "剖面"),
+    (["ELEV", "立面", "ELEVATION"], "立面"),
+    (["0", "A-"], "主体结构"),
+]
+
+
+def _infer_role(layer_name: str) -> str:
+    """根据图层名称推断图层角色（仅作建议）。"""
+    upper_name = layer_name.upper()
+    for keywords, role in _ROLE_RULES:
+        for kw in keywords:
+            if kw in upper_name or kw in layer_name:
+                return role
+    return "其他"
+
+
+def analyze_layers(doc) -> list:
+    """分析所有图层，返回每层的实体统计和角色推断。
+
+    Args:
+        doc: ezdxf Drawing 对象
+
+    Returns:
+        list[dict]: 每层信息，格式为 {
+            "name": str,
+            "entity_count": int,
+            "entity_types": list[str],
+            "suggested_role": str
+        }
+    """
+    msp = doc.modelspace()
+
+    # 按图层聚合实体
+    layer_entities = {}  # layer_name -> list of dxftype
+    for entity in msp:
+        layer_name = entity.dxf.layer
+        if layer_name not in layer_entities:
+            layer_entities[layer_name] = []
+        layer_entities[layer_name].append(entity.dxftype())
+
+    # 也包含没有实体的图层（在 doc.layers 中定义但模型空间无实体）
+    result = []
+    seen_layers = set()
+
+    for layer_name, types in layer_entities.items():
+        seen_layers.add(layer_name)
+        unique_types = sorted(set(types))
+        result.append({
+            "name": layer_name,
+            "entity_count": len(types),
+            "entity_types": unique_types,
+            "suggested_role": _infer_role(layer_name),
+        })
+
+    # 补充空图层（在图层定义中存在但无实体）
+    for layer in doc.layers:
+        name = layer.dxf.name
+        if name not in seen_layers:
+            result.append({
+                "name": name,
+                "entity_count": 0,
+                "entity_types": [],
+                "suggested_role": _infer_role(name),
+            })
+
+    # 按实体数降序排列
+    result.sort(key=lambda x: -x["entity_count"])
+    return result

experiments/dxf-parser/main.py

@@ -1,387 +1,102 @@
-"""DXF 解析测试程序 —— 使用 ezdxf 读取并输出 DXF 文件内容。"""
+"""DXF 解析工具 —— 读取 DXF 文件，输出图层分析和实体分析结果。"""
 
-import io
 import json
-import sys
 from datetime import datetime
 from pathlib import Path
 import ezdxf
 
-# 目录
+from layer_analysis import analyze_layers
+from entity_analysis import analyze_entities
+from element_detector import ElementDetector
+
+# 路径配置
 BASE_DIR = Path(__file__).parent.parent.parent
 SAMPLE_DIR = BASE_DIR / "samples"
 OUTPUT_DIR = Path(__file__).parent / "output"
 
-#DXF_FILE = SAMPLE_DIR / "decimal-inch-drawing-sheet-size-ASME-Y14.1-cad-block-dwg.dxf"
+# 要解析的 DXF 文件
 DXF_FILE = SAMPLE_DIR / "洗浴中心C-48.dxf"
 
-class OutputRecorder:
-    """同时输出到 stdout 和内存缓冲，最终写入文件。"""
 
-    def __init__(self):
-        self._buffer = io.StringIO()
-        self._terminal = sys.stdout
-
-    def write(self, message):
-        self._terminal.write(message)
-        self._buffer.write(message)
-
-    def flush(self):
-        self._terminal.flush()
-        self._buffer.flush()
-
-    def getvalue(self):
-        return self._buffer.getvalue()
-
-
-def extract_entity_props(entity):
-    """提取单个实体的关键属性为 JSON 友好格式。"""
-    dxftype = entity.dxftype()
-    props = {}
-
-    if dxftype == "LINE":
-        props["start"] = list(entity.dxf.start)
-        props["end"] = list(entity.dxf.end)
-    elif dxftype == "CIRCLE":
-        props["center"] = list(entity.dxf.center)
-        props["radius"] = round(entity.dxf.radius, 4)
-    elif dxftype == "ARC":
-        props["center"] = list(entity.dxf.center)
-        props["radius"] = round(entity.dxf.radius, 4)
-        props["start_angle"] = round(entity.dxf.start_angle, 1)
-        props["end_angle"] = round(entity.dxf.end_angle, 1)
-    elif dxftype == "LWPOLYLINE":
-        pts = entity.get_points()
-        props["vertex_count"] = len(entity)
-        props["closed"] = entity.closed
-        props["points"] = [[round(p[0], 4), round(p[1], 4)] for p in pts]
-    elif dxftype == "POLYLINE":
-        verts = list(entity.vertices())
-        props["vertex_count"] = len(verts)
-        props["points"] = [[round(v.dxf.location[0], 4), round(v.dxf.location[1], 4)] for v in verts]
-    elif dxftype == "INSERT":
-        props["block_name"] = entity.dxf.name
-        props["insert"] = list(entity.dxf.insert)
-        props["scale"] = [round(entity.dxf.xscale, 2), round(entity.dxf.yscale, 2), round(entity.dxf.zscale, 2)]
-        props["rotation"] = round(entity.dxf.rotation, 2)
-    elif dxftype in ("MTEXT", "TEXT"):
-        text = entity.plain_text() if dxftype == "MTEXT" else entity.dxf.text
-        props["text"] = text
-        if hasattr(entity.dxf, "insert"):
-            props["insert"] = list(entity.dxf.insert)
-        if hasattr(entity.dxf, "height"):
-            props["height"] = round(entity.dxf.height, 2)
-    elif dxftype == "DIMENSION":
-        props["text"] = entity.dxf.text
-        if hasattr(entity.dxf, "dimtype"):
-            props["dimtype"] = entity.dxf.dimtype
-    elif dxftype == "HATCH":
-        props["pattern"] = entity.dxf.pattern_name
-    elif dxftype == "SOLID" or dxftype == "TRACE":
-        for vi in range(4):
-            try:
-                v = getattr(entity.dxf, f"vtx{vi}")
-                props[f"vtx{vi}"] = list(v) if v else None
-            except Exception:
-                break
-    elif dxftype == "POINT":
-        props["location"] = list(entity.dxf.location)
-    elif dxftype == "ELLIPSE":
-        props["center"] = list(entity.dxf.center)
-        props["major_axis"] = list(entity.dxf.major_axis)
-        props["ratio"] = round(entity.dxf.ratio, 4)
-    elif dxftype == "SPLINE":
-        props["fit_points"] = [list(p) for p in entity.fit_points]
-    elif dxftype == "3DFACE":
-        for vi in range(4):
-            try:
-                v = getattr(entity.dxf, f"vtx{vi}")
-                props[f"vtx{vi}"] = list(v) if v else None
-            except Exception:
-                break
-    else:
-        for attr in entity.dxf.all_existing_dxf_attribs():
-            if attr in ("handle", "owner", "layer", "linetype", "color", "lineweight"):
-                continue
-            val = getattr(entity.dxf, attr)
-            if val is not None:
-                props[attr] = list(val) if hasattr(val, "__iter__") and not isinstance(val, str) else val
-
-    return props
-
-
-def extract_to_dict(doc: ezdxf.document.Drawing, filename: str, parse_time: str):
-    """提取 DXF 文档全部内容为 JSON 可序列化的 dict。"""
-    result = {
-        "file": filename,
-        "parse_time": parse_time,
-        "header": {
+def build_file_info(doc, filename: str) -> dict:
+    """提取 DXF 文件基本信息。"""
+    info = {
+        "filename": filename,
         "dxf_version": doc.dxfversion,
         "encoding": doc.encoding,
-            "acad_ver": doc.header.get("$ACADVER", None),
-        },
     }
 
     # 单位
     units = doc.header.get("$INSUNITS", None)
-    unit_names = {1: "inches", 2: "feet", 4: "millimeters", 5: "centimeters", 6: "meters"}
-    result["header"]["units"] = unit_names.get(units, str(units)) if units else None
+    info["units"] = units
 
     # 图形范围
     extmin = doc.header.get("$EXTMIN", None)
     extmax = doc.header.get("$EXTMAX", None)
     if extmin and extmax:
-        result["header"]["extents"] = {"min": list(extmin), "max": list(extmax)}
-
-    # 图层
-    layers = []
-    for layer in doc.layers:
-        layers.append({
-            "name": layer.dxf.name,
-            "color": layer.dxf.color,
-            "linetype": layer.dxf.linetype,
-            "lineweight": layer.dxf.lineweight,
-            "on": layer.is_on(),
-            "frozen": layer.is_frozen(),
-            "locked": layer.is_locked(),
-        })
-    result["layers"] = layers
-
-    # 实体
-    entities = []
-    entity_type_counts = {}
-    for entity in doc.modelspace():
-        dxftype = entity.dxftype()
-        entity_type_counts[dxftype] = entity_type_counts.get(dxftype, 0) + 1
-        entities.append({
-            "index": len(entities) + 1,
-            "type": dxftype,
-            "handle": entity.dxf.handle,
-            "layer": entity.dxf.layer,
-            "properties": extract_entity_props(entity),
-        })
-    result["entity_summary"] = entity_type_counts
-    result["entities"] = entities
-
-    # 块定义
-    blocks = []
-    for block in doc.blocks:
-        blocks.append({
-            "name": block.name,
-            "entity_count": len(list(block)),
-        })
-    result["blocks"] = blocks
-
-    # 布局
-    layouts = []
-    for layout in doc.layouts:
-        try:
-            pw = layout.dxf.paper_width if hasattr(layout.dxf, "paper_width") else None
-            ph = layout.dxf.paper_height if hasattr(layout.dxf, "paper_height") else None
-            layouts.append({"name": layout.name, "paper_width": pw, "paper_height": ph})
-        except Exception:
-            layouts.append({"name": layout.name, "paper_width": None, "paper_height": None})
-    result["layouts"] = layouts
-
-    return result
-
-
-def print_header(title: str):
-    print(f"\n{'=' * 70}")
-    print(f"  {title}")
-    print(f"{'=' * 70}")
-
-
-def print_dxf_info(doc: ezdxf.document.Drawing):
-    """打印 DXF 文档基本信息。"""
-    print(f"  DXF 版本       : {doc.dxfversion}")
-    print(f"  编码           : {doc.encoding}")
-    print(f"  创建者         : {doc.header.get('$ACADVER', 'N/A')}")
-
-    # 绘图单位
-    units = doc.header.get("$INSUNITS", None)
-    unit_map = {
-        1: "英寸 (Inches)",
-        2: "英尺 (Feet)",
-        4: "毫米 (Millimeters)",
-        5: "厘米 (Centimeters)",
-        6: "米 (Meters)",
-    }
-    unit_name = unit_map.get(units, f"未知 ({units})") if units else "未指定"
-    print(f"  绘图单位       : {unit_name}")
-
-    # 图形范围
-    extmin = doc.header.get("$EXTMIN", None)
-    extmax = doc.header.get("$EXTMAX", None)
-    if extmin and extmax:
-        print(f"  图形范围       : {extmin} -> {extmax}")
-
-
-def print_layers(doc: ezdxf.document.Drawing):
-    """打印所有图层信息。"""
-    layers = list(doc.layers)
-    if not layers:
-        print("  (无图层)")
-        return
-
-    print(f"  共 {len(layers)} 个图层:")
-    print(f"  {'名称':<30} {'颜色':<8} {'线型':<20} {'线宽':<8} {'状态'}")
-    print(f"  {'-' * 80}")
-    for layer in layers:
-        name = layer.dxf.name
-        color = layer.dxf.color
-        linetype = layer.dxf.linetype
-        # 线宽：-1 表示 DEFAULT
-        lineweight = layer.dxf.lineweight
-        lw_display = "默认" if lineweight == -1 else str(lineweight)
-        flags = []
-        if layer.is_on():
-            flags.append("ON")
+        info["extents"] = {"min": list(extmin), "max": list(extmax)}
     else:
-            flags.append("OFF")
-        if layer.is_frozen():
-            flags.append("FROZEN")
-        if layer.is_locked():
-            flags.append("LOCKED")
-        print(f"  {name:<30} {str(color):<8} {linetype:<20} {lw_display:<8} {', '.join(flags)}")
+        info["extents"] = None
+
+    return info
 
 
-def print_entity_summary(doc: ezdxf.document.Drawing):
-    """统计模型空间中各类型实体数量。"""
-    msp = doc.modelspace()
-    entity_types = {}
-    for entity in msp:
-        dtype = entity.dxftype()
-        entity_types[dtype] = entity_types.get(dtype, 0) + 1
+def print_summary(file_info: dict, layers: list, entities: dict):
+    """终端打印解析摘要。"""
+    print(f"\n{'=' * 60}")
+    print(f"  CAD 文件解析报告")
+    print(f"{'=' * 60}")
+    print(f"  文件     : {file_info['filename']}")
+    print(f"  版本     : {file_info['dxf_version']}")
+    print(f"  编码     : {file_info['encoding']}")
+    print(f"  实体总数 : {entities['total_entities']}")
+    if file_info.get("extents"):
+        print(f"  图形范围 : {file_info['extents']}")
 
-    if not entity_types:
-        print("  (模型空间无实体)")
-        return
+    # 图层摘要
+    print(f"\n  {'─' * 56}")
+    print(f"  图层分析（共 {len(layers)} 个图层）:")
+    print(f"  {'名称':<25} {'实体数':<8} {'角色推断'}")
+    print(f"  {'-' * 50}")
+    for layer in layers[:20]:  # 最多显示前 20 个
+        print(f"  {layer['name']:<25} {layer['entity_count']:<8} {layer['suggested_role']}")
+    if len(layers) > 20:
+        print(f"  ... 还有 {len(layers) - 20} 个图层")
 
-    print(f"  共 {sum(entity_types.values())} 个实体:")
+    # 实体类型统计
+    type_counts = entities["type_counts"]
+    print(f"\n  {'─' * 56}")
+    print(f"  实体类型统计:")
     print(f"  {'类型':<25} {'数量':<8}")
     print(f"  {'-' * 35}")
-    for dtype, count in sorted(entity_types.items(), key=lambda x: -x[1]):
+    for dtype, count in sorted(type_counts.items(), key=lambda x: -x[1]):
         print(f"  {dtype:<25} {count:<8}")
 
+    # 块统计
+    blocks = entities["blocks"]
+    if blocks:
+        print(f"\n  {'─' * 56}")
+        print(f"  块引用统计（共 {len(blocks)} 种块）:")
+        print(f"  {'块名':<35} {'引用次数':<10}")
+        print(f"  {'-' * 47}")
+        for blk in blocks[:15]:
+            print(f"  {blk['block_name']:<35} {blk['count']:<10}")
+        if len(blocks) > 15:
+            print(f"  ... 还有 {len(blocks) - 15} 种块")
 
-def print_entity_details(doc: ezdxf.document.Drawing):
-    """打印模型空间中每个实体的详细信息。"""
-    msp = doc.modelspace()
-    entities = list(msp)
-    if not entities:
-        print("  (模型空间无实体)")
-        return
+    # 文本
+    texts = entities["texts"]
+    if texts:
+        print(f"\n  {'─' * 56}")
+        print(f"  文本内容（共 {len(texts)} 条）:")
+        for t in texts[:10]:
+            print(f"  [{t['layer']}] {t['content'][:60]}")
+        if len(texts) > 10:
+            print(f"  ... 还有 {len(texts) - 10} 条文本")
 
-    print(f"  共 {len(entities)} 个实体:")
-    print(f"  {'序号':<6} {'类型':<25} {'句柄':<12} {'图层':<20} {'关键属性'}")
-    print(f"  {'-' * 100}")
-
-    for i, entity in enumerate(entities):
-        dxftype = entity.dxftype()
-        handle = entity.dxf.handle
-        layer = entity.dxf.layer
-
-        # 提取各类型的关键属性
-        key_attrs = []
-        if dxftype == "LINE":
-            key_attrs.append(f"起点={entity.dxf.start}, 终点={entity.dxf.end}")
-        elif dxftype == "CIRCLE":
-            key_attrs.append(f"圆心={entity.dxf.center}, 半径={entity.dxf.radius:.4f}")
-        elif dxftype == "ARC":
-            key_attrs.append(f"圆心={entity.dxf.center}, 半径={entity.dxf.radius:.4f}, 角度={entity.dxf.start_angle:.1f}~{entity.dxf.end_angle:.1f}")
-        elif dxftype == "LWPOLYLINE":
-            count = len(entity)
-            closed = "闭合" if entity.closed else "开放"
-            pts = entity.get_points()
-            first_pt = pts[0] if pts else "N/A"
-            key_attrs.append(f"顶点数={count}, {closed}, 起点={first_pt[:2]}")
-        elif dxftype == "POLYLINE":
-            count = len(list(entity.vertices()))
-            key_attrs.append(f"顶点数={count}")
-        elif dxftype == "INSERT":
-            key_attrs.append(f"块名={entity.dxf.name}, 插入点={entity.dxf.insert}, 缩放={entity.dxf.xscale:.2f},{entity.dxf.yscale:.2f}")
-        elif dxftype == "MTEXT":
-            text = entity.plain_text()[:50]
-            key_attrs.append(f'内容="{text}..."' if len(entity.plain_text()) > 50 else f'内容="{text}"')
-        elif dxftype == "TEXT":
-            text = entity.dxf.text[:50]
-            key_attrs.append(f'内容="{text}..."' if len(entity.dxf.text) > 50 else f'内容="{entity.dxf.text}"')
-        elif dxftype == "DIMENSION":
-            key_attrs.append(f"测量值={entity.dxf.text}")
-        elif dxftype == "HATCH":
-            key_attrs.append(f"图案={entity.dxf.pattern_name}")
-        elif dxftype == "SOLID" or dxftype == "TRACE":
-            key_attrs.append(f"顶点={entity.dxf.vtx0} {entity.dxf.vtx1} {entity.dxf.vtx2} {entity.dxf.vtx3}")
-        elif dxftype == "POINT":
-            key_attrs.append(f"位置={entity.dxf.location}")
-        elif dxftype == "ELLIPSE":
-            key_attrs.append(f"圆心={entity.dxf.center}, 长轴={entity.dxf.major_axis}")
-        elif dxftype == "SPLINE":
-            key_attrs.append(f"拟合点数={len(entity.fit_points)}")
-        elif dxftype == "3DFACE":
-            key_attrs.append(f"顶点={entity.dxf.vtx0} {entity.dxf.vtx1} {entity.dxf.vtx2} {entity.dxf.vtx3}")
-        else:
-            # 通用：列出非默认的 dxf 属性
-            for attr in entity.dxf.all_existing_dxf_attribs():
-                if attr in ("handle", "owner", "layer", "linetype", "color", "lineweight"):
-                    continue
-                val = getattr(entity.dxf, attr)
-                if val is not None:
-                    key_attrs.append(f"{attr}={val}")
-
-        attrs_str = "; ".join(key_attrs[:3]) if key_attrs else "(无额外属性)"
-        print(f"  {i + 1:<6} {dxftype:<25} {handle:<12} {layer:<20} {attrs_str}")
-
-
-def print_blocks(doc: ezdxf.document.Drawing):
-    """打印所有块定义。"""
-    blocks = list(doc.blocks)
-    if not blocks:
-        print("  (无自定义块)")
-        return
-
-    print(f"  共 {len(blocks)} 个块定义:")
-    print(f"  {'名称':<35} {'实体数':<10} {'说明'}")
-    print(f"  {'-' * 70}")
-    for block in blocks:
-        name = block.name
-        count = len(list(block))
-        desc = ""
-        if name.startswith("*"):
-            if name.startswith("*Model_Space"):
-                desc = "(模型空间)"
-            elif name.startswith("*Paper_Space"):
-                desc = "(图纸空间)"
-            elif name.startswith("*D"):
-                desc = "(标注块)"
-            elif name.startswith("*X"):
-                desc = "(外部参照)"
-            elif name.startswith("*U"):
-                desc = "(匿名块)"
-            else:
-                desc = "(特殊块)"
-        print(f"  {name:<35} {count:<10} {desc}")
-
-
-def print_layouts(doc: ezdxf.document.Drawing):
-    """打印所有布局（图纸空间）。"""
-    layouts = list(doc.layouts)
-    if not layouts:
-        print("  (无布局)")
-        return
-
-    print(f"  共 {len(layouts)} 个布局:")
-    print(f"  {'名称':<30} {'尺寸 (宽 x 高)'}")
-    print(f"  {'-' * 60}")
-    for layout in layouts:
-        name = layout.name
-        try:
-            page_w = layout.dxf.paper_width if hasattr(layout.dxf, 'paper_width') else "?"
-            page_h = layout.dxf.paper_height if hasattr(layout.dxf, 'paper_height') else "?"
-            print(f"  {name:<30} {page_w} x {page_h}")
-        except Exception:
-            print(f"  {name:<30} N/A")
+    print(f"\n{'=' * 60}")
+    print(f"  解析完成")
+    print(f"{'=' * 60}\n")
 
 
 def main():
@@ -390,61 +105,67 @@ def main():
         return
 
     OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
-
-    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
     parse_time = datetime.now().strftime("%Y-%m-%d %H:%M:%S")
-    output_txt = OUTPUT_DIR / f"{DXF_FILE.stem}_{timestamp}.txt"
-    output_json = OUTPUT_DIR / f"{DXF_FILE.stem}_{timestamp}.json"
 
-    recorder = OutputRecorder()
-    sys.stdout = recorder
-
-    try:
     print(f"解析文件: {DXF_FILE.name}")
-        print(f"路径: {DXF_FILE}")
-        print(f"输出文件: {output_txt.relative_to(BASE_DIR)}")
-        print(f"JSON 文件: {output_json.relative_to(BASE_DIR)}")
     print(f"解析时间: {parse_time}")
 
+    # 读取 DXF
     doc = ezdxf.readfile(str(DXF_FILE))
 
-        # 1. 基本信息
-        print_header("一、文档基本信息")
-        print_dxf_info(doc)
+    # 文件基本信息
+    file_info = build_file_info(doc, DXF_FILE.name)
 
-        # 2. 图层
-        print_header("二、图层信息")
-        print_layers(doc)
+    # 图层分析
+    layers = analyze_layers(doc)
 
-        # 3. 实体统计
-        print_header("三、模型空间实体统计")
-        print_entity_summary(doc)
+    # 实体分析
+    entities = analyze_entities(doc)
 
-        # 4. 实体详情
-        print_header("四、模型空间实体详情")
-        print_entity_details(doc)
+    # 添加实体总数到文件信息
+    file_info["total_entities"] = entities["total_entities"]
 
-        # 5. 块定义
-        print_header("五、块定义")
-        print_blocks(doc)
+    # 终端打印摘要
+    print_summary(file_info, layers, entities)
 
-        # 6. 布局
-        print_header("六、布局（图纸空间）")
-        print_layouts(doc)
+    # 组装并输出 JSON
+    output = {
+        "file_info": file_info,
+        "layers": layers,
+        "entities": entities["type_counts"],
+        "blocks": entities["blocks"],
+        "texts": entities["texts"],
+    }
 
-        print_header("解析完成")
-    finally:
-        sys.stdout = recorder._terminal
-
-    # 写入文本文件
-    output_txt.write_text(recorder.getvalue(), encoding="utf-8")
-    print(f"\n结果已保存到: {output_txt}")
-
-    # 写入 JSON 文件
-    json_data = extract_to_dict(doc, DXF_FILE.name, parse_time)
-    output_json.write_text(json.dumps(json_data, ensure_ascii=False, indent=2), encoding="utf-8")
+    output_json = OUTPUT_DIR / f"{DXF_FILE.stem}_cad_analysis.json"
+    output_json.write_text(
+        json.dumps(output, ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
     print(f"JSON 已保存到: {output_json}")
 
+    # ========== 第二阶段：元素识别 ==========
+    print(f"\n{'─' * 56}")
+    print(f"  开始元素识别...")
+    detector = ElementDetector()
+    detection_result = detector.detect_all(doc)
+
+    # 输出元素检测 JSON
+    detection_json = OUTPUT_DIR / f"{DXF_FILE.stem}_element_detection.json"
+    detection_json.write_text(
+        json.dumps(detection_result, ensure_ascii=False, indent=2),
+        encoding="utf-8",
+    )
+    print(f"元素检测结果已保存到: {detection_json}")
+
+    # 输出统计报告
+    from element_detector import generate_report
+    report = generate_report(detection_result)
+    report_path = OUTPUT_DIR / f"{DXF_FILE.stem}_element_report.txt"
+    report_path.write_text(report, encoding="utf-8")
+    print(f"分析报告已保存到: {report_path}")
+    print(report)
+
 
 if __name__ == "__main__":
     main()