dsh-engineering-ui 0.0.0-stage → 1.5.0

package/engineering/tools/drawing_input.py

@@ -0,0 +1,314 @@
+"""Prepare drawing pages and embedded PDF text for visual interpretation.
+
+Text is evidence, not confirmed dimensions. No measurements are inferred.
+"""
+import argparse
+from collections import Counter
+import json
+from pathlib import Path
+import re
+import shutil
+import subprocess
+import sys
+import tempfile
+
+
+def json_value(value):
+    if value is None or isinstance(value, (str, int, float, bool)):
+        return value
+    try:
+        return [json_value(item) for item in value]
+    except TypeError:
+        return str(value)
+
+
+def converted_dwg_tags(stream, warnings):
+    from ezdxf.lldxf.tagger import ascii_tags_loader
+    from ezdxf.lldxf.tags import group_tags
+
+    section = None
+    for tags in group_tags(ascii_tags_loader(stream)):
+        if tags[0].value == "SECTION":
+            section = tags.get_first_value(2, None)
+        if (section == "OBJECTS" and tags[0].value == "DICTIONARY"
+                and tags.get_first_value(330, None) == "0"):
+            removed = 0
+            index = len(tags) - 1
+            while index > 0:
+                if (tags[index - 1].code == 3 and tags[index - 1].value == ""
+                        and tags[index].code in (350, 360) and tags[index].value == "0"):
+                    del tags[index - 1:index + 1]
+                    removed += 1
+                    index -= 1
+                index -= 1
+            if removed:
+                warnings.append("DWG 转换结果的根管理字典含 %d 个空名字、空指针条目，读取时已忽略；"
+                                "其余标签及原始转换文件保持不变。仍需对照原图核验转换完整性。" % removed)
+        yield from tags
+
+
+def read_dxf(source, result, output_dir, preview_limit):
+    try:
+        import ezdxf
+        from ezdxf.lldxf.encoding import decode_dxf_unicode
+    except ImportError as exc:
+        raise RuntimeError("读取 DXF 需要安装 ezdxf；现有工程的 ac_validate.py 也使用该库") from exc
+    try:
+        from ezdxf.lldxf.validator import is_binary_dxf_file
+
+        if "converted_dxf" in result and not is_binary_dxf_file(source):
+            from ezdxf.document import Drawing
+            from ezdxf.filemanagement import dxf_file_info
+
+            with source.open(encoding=dxf_file_info(source).encoding, errors="surrogateescape") as stream:
+                doc = Drawing.load(converted_dwg_tags(stream, result["warnings"]))
+            doc.filename = str(source)
+        else:
+            doc = ezdxf.readfile(str(source))
+    except (ezdxf.DXFError, ValueError) as exc:
+        raise RuntimeError("无法解析 DXF，请从原 CAD 导出标准 DXF 或 PDF: " + str(exc)) from exc
+    layouts = list(doc.layouts)
+    blocks = [block for block in doc.blocks if not block.block_record.is_any_layout]
+    entities = []
+    spaces = [(layout, "layout") for layout in layouts] + [(block, "block_definition") for block in blocks]
+    for space, space_kind in spaces:
+        for entity in space:
+            kind = entity.dxftype()
+            item = {"type": kind, "handle": entity.dxf.get("handle"),
+                    "layer": entity.dxf.get("layer"),
+                    "space": space.name, "source_space": space_kind,
+                    "attributes": {k: json_value(v) for k, v in entity.dxf.all_existing_dxf_attribs().items()}}
+            if kind == "LWPOLYLINE":
+                item["points_xyseb"] = [list(point) for point in entity.get_points("xyseb")]
+            if kind == "POLYLINE":
+                item["vertices"] = [{k: json_value(v) for k, v in p.dxf.all_existing_dxf_attribs().items()} for p in entity.vertices]
+            if kind == "SPLINE":
+                item["control_points"] = [list(p) for p in entity.control_points]
+                item["fit_points"] = [list(p) for p in entity.fit_points]
+                item["knots"] = [float(k) for k in entity.knots]
+                item["weights"] = [float(w) for w in entity.weights]
+            if kind == "MTEXT":
+                item["raw_text"] = entity.text
+            if kind in ("TEXT", "MTEXT"):
+                item["plain_text"] = entity.plain_text()
+            if kind == "DIMENSION":
+                geometry = entity.get_geometry_block()
+                item["rendered_text"] = []
+                for part in geometry or []:
+                    if part.dxftype() not in ("TEXT", "MTEXT"):
+                        continue
+                    layer = doc.layers.get(part.dxf.layer) if part.dxf.layer in doc.layers else None
+                    if part.dxf.get("invisible") or layer and (layer.is_off() or layer.is_frozen()):
+                        continue
+                    item["rendered_text"].append(part.plain_text())
+                if entity.dxf.get("dimstyle", "STANDARD") in doc.dimstyles:
+                    item["dimension_style_overrides"] = {
+                        k: json_value(v) for k, v in entity.override().dimstyle_attribs.items()}
+                else:
+                    result["warnings"].append("标注 %s 的样式 %s 缺失，无法核对显示规则。" %
+                                              (entity.dxf.handle, entity.dxf.dimstyle))
+                if geometry is None:
+                    result["warnings"].append("标注 %s（%s）没有显示图形，预览可能缺少该尺寸；不会根据定义点补画。" %
+                                              (entity.dxf.handle, space.name))
+            if kind == "INSERT":
+                item["attached_attributes"] = [{k: json_value(v) for k, v in attribute.dxf.all_existing_dxf_attribs().items()}
+                                               for attribute in entity.attribs]
+                if doc.blocks.get(entity.dxf.name) is None:
+                    result["warnings"].append("块引用 %s 的块 %s 缺失，图形无法完整显示。" %
+                                              (entity.dxf.handle, entity.dxf.name))
+            entities.append(item)
+    for block in blocks:
+        if block.block.dxf.flags & 12:
+            result["warnings"].append("外部引用 %s 未加载：%s；预览不包含该文件的图形。" %
+                                      (block.name, block.block.dxf.get("xref_path", "")))
+    output_dir = Path(output_dir).resolve()
+    output_dir.mkdir(parents=True, exist_ok=True)
+    evidence_path = output_dir / "dxf-entities.json"
+    styles = [{k: json_value(v) for k, v in style.dxf.all_existing_dxf_attribs().items()} for style in doc.dimstyles]
+    with evidence_path.open("x", encoding="utf-8") as evidence:
+        json.dump({"entities": entities, "dimension_styles": styles,
+                   "block_definitions": [{k: json_value(v) for k, v in block.block.dxf.all_existing_dxf_attribs().items()}
+                                         for block in blocks],
+                   "source_warnings": result["warnings"]}, evidence, ensure_ascii=False)
+    if len(result["warnings"]) > 8:
+        result["warnings"] = result["warnings"][:7] + [
+            "其余 %d 条源数据缺失告警见 entity_evidence 的 source_warnings；不能视为完整图纸。" % (len(result["warnings"]) - 7)]
+    previews = []
+    preview_chars = 0
+    for entity in entities:
+        if entity["source_space"] != "layout" or entity["type"] not in ("DIMENSION", "TEXT", "MTEXT"):
+            continue
+        layer = doc.layers.get(entity["layer"]) if entity["layer"] in doc.layers else None
+        if entity["attributes"].get("invisible") or layer and (layer.is_off() or layer.is_frozen()):
+            continue
+        if entity["type"] == "DIMENSION" and not entity.get("rendered_text"):
+            continue
+        text = " / ".join(entity.get("rendered_text", [])) or entity.get("plain_text", "")
+        text = re.sub(r"(?<!\\)\\U\+[0-9A-Fa-f]{4}(?![0-9A-Fa-f])",
+                      lambda match: (decode_dxf_unicode(match.group())
+                                     if not 0xD800 <= int(match.group()[3:], 16) <= 0xDFFF else match.group()), text)
+        item = {"type": entity["type"], "handle": entity["handle"], "layer": entity["layer"],
+                "space": entity["space"], "source_space": entity["source_space"], "text": text[:300]}
+        if entity["type"] == "DIMENSION":
+            item["text_override"] = entity["attributes"].get("text", "")
+        size = len(json.dumps(item, ensure_ascii=False))
+        if len(previews) >= 20 or preview_chars + size > 6000:
+            break
+        previews.append(item)
+        preview_chars += size
+    result.update(ok=any(len(layout) for layout in layouts), pages=0, entity_count=len(entities),
+                  entity_counts=dict(Counter(e["type"] for e in entities)),
+                  entity_evidence=str(evidence_path),
+                  annotation_preview=previews,
+                  drawing_units_code=doc.header.get("$INSUNITS", 0),
+                  layout_count=len(layouts), source_space="layouts_and_block_definitions")
+    if not result["ok"]:
+        result["warnings"].append("所有模型空间和图纸布局均无图元，没有可读取的图纸内容。")
+    result["warnings"].extend([
+        "DIMENSION 原始属性及文字只是候选，尺寸值、标注样式、文字覆盖和对应几何必须核对；文件单位代码不保证标注单位。",
+        "原始数据包含布局和块定义；文字摘要排除块定义、隐藏图元及关闭或冻结图层。块坐标未经插入变换，不能当作实际位置；显示尺寸文字取自原有标注图形，不重新计算。摘要最多 20 条且截断长文字。",
+        "预览受字体、外部引用、代理对象和布局视口支持影响，可能有缺失，不可仅凭预览认定图纸完整；缺失、冲突和不清晰数据应集中询问用户。"])
+    if preview_limit:
+        try:
+            from ezdxf.addons.drawing.matplotlib import qsave
+            from PIL import Image
+        except ImportError:
+            result["warnings"].append("缺少 matplotlib，DXF 预览不可用；请安装 matplotlib 或使用 CAD 打开。")
+        else:
+            nonempty_layouts = [(index, layout) for index, layout in enumerate(layouts, 1) if len(layout)]
+            for index, layout in nonempty_layouts[:preview_limit]:
+                image_path = output_dir / ("dxf-preview-%04d.png" % index)
+                if image_path.exists():
+                    raise FileExistsError("预览图片已存在，请选择空目录")
+                try:
+                    qsave(layout, image_path, bg="#FFFFFF", dpi=120, size_inches=(12, 9))
+                    if not image_path.is_file() or not image_path.stat().st_size:
+                        raise RuntimeError("预览图片为空")
+                    with Image.open(image_path) as image:
+                        if all(low == high == 255 for low, high in image.convert("RGB").getextrema()):
+                            raise RuntimeError("预览仅有白色背景，没有可见图形，不能视为图纸已显示")
+                    result["images"].append({"page": index, "path": str(image_path), "layout": layout.name})
+                except Exception as exc:
+                    result["warnings"].append("布局 %s 预览失败: %s" % (layout.name, exc))
+            result["pages"] = len(result["images"])
+            if len(nonempty_layouts) > preview_limit:
+                result["warnings"].append("仅尝试预览前 %d 个非空空间，共有 %d 个；可调整 --preview-limit。" % (preview_limit, len(nonempty_layouts)))
+    return result
+
+
+def run(command):
+    return subprocess.run(command, capture_output=True, timeout=120, check=True)
+
+
+def prepare(source, output_dir, dwg2dxf=None, preview_limit=3, page=None):
+    source = Path(source).resolve(strict=True)
+    result = {"ok": False, "input": str(source), "pages": 0,
+              "extracted_text": [], "images": [], "ocr": "not_run",
+              "warnings": ["图中文字仅供候选，必须结合页图确认数值、尺寸对象和单位；缺失或模糊时向用户询问。禁止从像素比例推测尺寸。"]}
+    suffix = source.suffix.lower()
+    if page is not None and suffix != ".pdf":
+        raise ValueError("--page 仅用于 PDF")
+    if suffix == ".dxf":
+        return read_dxf(source, result, output_dir, preview_limit)
+    if suffix == ".dwg":
+        local_converter = Path(sys.executable).with_name("dwg2dxf.exe")
+        converter = dwg2dxf or shutil.which("dwg2dxf") or (local_converter if local_converter.is_file() else None)
+        if not converter:
+            raise RuntimeError("未找到 LibreDWG 的 dwg2dxf；请提供 --dwg2dxf 可执行文件路径，或先转换为 DXF。")
+        output_dir = Path(output_dir).resolve()
+        output_dir.mkdir(parents=True, exist_ok=True)
+        converted = output_dir / "converted.dxf"
+        if converted.exists():
+            raise FileExistsError("转换结果已存在，请选择空目录")
+        with tempfile.TemporaryDirectory(prefix="dwg-input-", dir=output_dir) as tmp:
+            copied = Path(tmp) / "input.dwg"
+            shutil.copyfile(source, copied)
+            run([str(converter), "-o", str(converted), str(copied)])
+        if not converted.is_file() or converted.stat().st_size == 0:
+            raise RuntimeError("DWG 转换器未生成有效输出")
+        result["converted_dxf"] = str(converted)
+        result["warnings"].append("DWG 已经转换为 DXF；转换可能丢失信息，需对照原图核验。")
+        return read_dxf(converted, result, output_dir, preview_limit)
+    if suffix in {".png", ".jpg", ".jpeg", ".webp", ".bmp", ".tif", ".tiff"}:
+        from PIL import Image
+        with Image.open(source) as image:
+            if getattr(image, "n_frames", 1) != 1:
+                raise ValueError("多页图片暂不支持，请先拆分为单页图片")
+            if image.format not in {"PNG", "JPEG", "WEBP"}:
+                out = Path(output_dir).resolve()
+                out.mkdir(parents=True, exist_ok=True)
+                preview = out / "image.png"
+                with preview.open("xb") as target:
+                    image.convert("RGB").save(target, "PNG")
+            else:
+                image.load()
+                preview = source
+        result.update(ok=True, pages=1, images=[{"page": 1, "path": str(preview)}])
+        result["warnings"].append("图片没有嵌入文字提取结果；本工具没有运行 OCR，需用视觉工具查看原图。")
+        return result
+    if suffix != ".pdf":
+        raise ValueError("仅支持 PDF、DXF 和常见图片格式；DWG 需要先转换")
+    programs = {name: shutil.which(name) for name in ("pdfinfo", "pdftotext", "pdftoppm")}
+    missing = [name for name, path in programs.items() if not path]
+    if missing:
+        raise RuntimeError("缺少 PDF 工具: " + ", ".join(missing))
+    info = run([programs["pdfinfo"], str(source)]).stdout.decode("utf-8", "replace")
+    match = re.search(r"^Pages:\s*(\d+)", info, re.MULTILINE)
+    if not match or int(match.group(1)) < 1:
+        raise RuntimeError("无法读取 PDF 页数")
+    count = int(match.group(1))
+    output_dir = Path(output_dir).resolve()
+    output_dir.mkdir(parents=True, exist_ok=True)
+    if page is not None and not 1 <= page <= count:
+        raise ValueError("请求的页码超出范围，图纸共 %d 页" % count)
+    pages = [page] if page is not None else range(1, min(count, 3) + 1)
+    existing = [output_dir / ("page-%04d.png" % p) for p in pages]
+    if any(path.exists() for path in existing):
+        raise FileExistsError("输出图片已存在，请选择空目录")
+    result["pages"] = count
+    if count > len(pages):
+        result["warnings"].append("还有未读页；使用 --page N 查看其余页再判断图纸是否完整。")
+    for page in pages:
+        prefix = output_dir / ("page-%04d" % page)
+        image = prefix.with_suffix(".png")
+        run([programs["pdftoppm"], "-f", str(page), "-l", str(page), "-singlefile", "-cropbox", "-r", "150", "-png", str(source), str(prefix)])
+        if not image.is_file() or image.stat().st_size == 0:
+            raise RuntimeError("PDF 页面渲染失败: " + str(page))
+        text = run([programs["pdftotext"], "-f", str(page), "-l", str(page), "-layout", "-enc", "UTF-8", str(source), "-"]).stdout.decode("utf-8", "replace").strip("\f\n\r ")
+        result["images"].append({"page": page, "path": str(image)})
+        item = {"page": page, "source": "pdf_embedded_text", "text": text[:2400]}
+        if len(text) > 2400:
+            text_file = prefix.with_suffix(".txt")
+            with text_file.open("x", encoding="utf-8") as full:
+                full.write(text)
+            item["full_text_path"] = str(text_file)
+        result["extracted_text"].append(item)
+        if not text:
+            result["warnings"].append("第 %d 页没有可提取的嵌入文字；可能是扫描图，需查看页图。" % page)
+    result["warnings"].append("未运行 OCR；嵌入文字可能缺字或顺序错误，不能单独确认尺寸。")
+    result["ok"] = True
+    return result
+
+
+def main():
+    if hasattr(sys.stdout, "reconfigure"):
+        sys.stdout.reconfigure(encoding="utf-8")
+    parser = argparse.ArgumentParser(description=__doc__)
+    parser.add_argument("input")
+    parser.add_argument("--page", type=int, help="只读取指定 PDF 页（从 1 开始）；默认返回前 3 页")
+    parser.add_argument("--output-dir", required=True, help="PDF 页面图片输出目录；已有同名页面不会覆盖")
+    parser.add_argument("--dwg2dxf", help="LibreDWG dwg2dxf 可执行文件路径（不填则查找 PATH）")
+    parser.add_argument("--preview-limit", type=int, default=3, help="DXF/DWG 预览最多的模型空间和布局数（0–20，默认 3）")
+    args = parser.parse_args()
+    try:
+        if not 0 <= args.preview_limit <= 20:
+            raise ValueError("--preview-limit 必须在 0 到 20 之间")
+        result = prepare(args.input, args.output_dir, args.dwg2dxf, args.preview_limit, args.page)
+    except (OSError, ValueError, RuntimeError, SyntaxError, ImportError, subprocess.SubprocessError) as exc:
+        result = {"ok": False, "input": args.input, "error": str(exc), "ocr": "not_run"}
+    print(json.dumps(result, ensure_ascii=False))
+    return 0 if result["ok"] else 1
+
+
+if __name__ == "__main__":
+    sys.exit(main())