如何使用Apache POI XWPF获取docx文档绘图中的形状

问题场景 #

持有包含文本与若干形状的docx文档，已通过Apache POI成功获取文档内的CTDrawing绘图对象，可识别到inline（内嵌）、anchor（锚定）两类绘图节点下存在形状数据，但无法正确解析提取Shape对象，现有解析代码如下：

try (FileInputStream fis = new FileInputStream(file)) {
   document = new XWPFDocument(fis);
   parseDocument(document);
}

核心解析逻辑代码：

public void parseDocument(XWPFDocument document) throws Exception {
   Iterator<XWPFParagraph> it = document.getParagraphs().iterator();
   while (it.hasNext()) {
      XWPFParagraph paragraph = it.next();
      List<XWPFRun> runs = paragraph.getRuns();
      Iterator<XWPFRun> it2 = runs.iterator();
      while (it2.hasNext()) {
         XWPFRun run = it2.next();
         List<CTDrawing> drawings = getAllDrawings(run);
         if (!drawings.isEmpty()) {
            System.out.println("Found " + drawings.size() + " drawings");
            Iterator<CTDrawing> it3 = drawings.iterator();
            while (it3.hasNext()) {
               CTDrawing drawing = it3.next();
               visitDrawing(drawing);
            }
         }
      }
   }
}

private void visitDrawing(CTDrawing drawing) {
   List<CTInline> inlines = drawing.getInlineList();
   Iterator<CTInline> it = inlines.iterator();
   while (it.hasNext()) {
      CTInline inline = it.next();
      CTGraphicalObject graphic = inline.getGraphic();
      if (graphic != null) {
         CTGraphicalObjectData graphicData = graphic.getGraphicData();
         if (graphicData != null) {
            XmlCursor c = graphicData.newCursor();
            c.selectPath("./*");
            while (c.toNextSelection()) {
               XmlObject o = c.getObject();
               System.out.println(o);
            }
         }
      }
   }
   
   List<CTAnchor> anchors = drawing.getAnchorList();
   Iterator<CTAnchor> it2 = anchors.iterator();
   while (it2.hasNext()) {
      CTAnchor anchor = it2.next();
      CTGraphicalObject graphic = anchor.getGraphic();
      if (graphic != null) {
         CTGraphicalObjectData graphicData = graphic.getGraphicData();
         if (graphicData != null) {
            XmlCursor c = graphicData.newCursor();
            c.selectPath("./*");
            while (c.toNextSelection()) {
               XmlObject o = c.getObject();
               System.out.println(o);
            }
         }
      }
   }      
}

private static List<CTDrawing> getAllDrawings(XWPFRun run) throws Exception {
   CTR ctR = run.getCTR();
   XmlCursor cursor = ctR.newCursor();
   cursor.selectPath("declare namespace w='http://schemas.openxmlformats.org/wordprocessingml/2006/main' .//*/w:drawing");
   List<CTDrawing> drawings = new ArrayList<>();
   while (cursor.hasNextSelection()) {
      cursor.toNextSelection();
      XmlObject obj = cursor.getObject();
      CTDrawing drawing = CTDrawing.Factory.parse(obj.newInputStream());
      drawings.add(drawing);
   }
   return drawings;
}

实现步骤 #

1. 替换依赖 #

首先确保项目引入的是完整版POI OOXML依赖，精简版poi-ooxml-lite裁剪了形状相关的XML绑定类，会出现类找不到的问题，版本和当前使用的POI保持一致即可，Maven配置示例：

<dependency>
    <groupId>org.apache.poi</groupId>
    <artifactId>poi-ooxml-full</artifactId>
    <version>5.2.5</version>
</dependency>

2. 修改绘图节点解析逻辑 #

Word中插入的普通自选图形、文本框、艺术字都存储在graphicData节点下，对应http://schemas.microsoft.com/office/word/2010/wordprocessingShape命名空间的CTShape节点，遍历子节点时直接做类型判断即可拿到形状对象，替换原有visitDrawing方法及新增解析逻辑如下：

// 统一处理内嵌、锚定绘图的图形数据
private void visitDrawing(CTDrawing drawing) {
    for (CTInline inline : drawing.getInlineList()) {
        extractShapesFromGraphic(inline.getGraphic());
    }
    for (CTAnchor anchor : drawing.getAnchorList()) {
        extractShapesFromGraphic(anchor.getGraphic());
    }
}

private void extractShapesFromGraphic(CTGraphicalObject graphic) {
    if (graphic == null) return;
    CTGraphicalObjectData graphicData = graphic.getGraphicData();
    if (graphicData == null) return;

    try (XmlCursor cursor = graphicData.newCursor()) {
        cursor.selectPath("./*");
        while (cursor.toNextSelection()) {
            XmlObject node = cursor.getObject();
            // 优先使用POI封装的高层Shape API，使用更简单
            if (node instanceof XWPFShape) {
                XWPFShape shape = (XWPFShape) node;
                // 读取基础属性
                System.out.printf("找到形状，名称：%s，类型：%s%n", 
                    shape.getShapeName(), shape.getShapeType());
                // 读取形状内的文本内容
                if (shape instanceof XWPFShapeBase) {
                    List<XWPFParagraph> paragraphs = ((XWPFShapeBase) shape).getTextParagraphs();
                    for (XWPFParagraph para : paragraphs) {
                        System.out.println("形状内文本：" + para.getText());
                    }
                }
            }
            // 兼容低版本POI无高层API的场景，直接操作底层XML对象
            else if (node instanceof com.microsoft.schemas.office.word.x2010.wordprocessingShape.CTShape) {
                var ctShape = (com.microsoft.schemas.office.word.x2010.wordprocessingShape.CTShape) node;
                // 读取基础属性
                var shapePr = ctShape.getCnVPr();
                System.out.printf("找到底层形状对象，ID：%d，名称：%s%n",
                    shapePr.getId(), shapePr.getName());
                // 读取形状内文本
                CTTextBody textBody = ctShape.getTxBody();
                if (textBody != null) {
                    for (CTParagraph para : textBody.getPList()) {
                        StringBuilder text = new StringBuilder();
                        for (CTRun run : para.getRList()) {
                            text.append(run.getT());
                        }
                        System.out.println("形状内文本：" + text);
                    }
                }
                // 还可以读取位置、大小、填充色、边框等所有属性，直接调用ctShape对应的get方法即可
            }
        }
    }
}

注意事项 #

• 如果需要解析组合形状，拿到CTShape对象后递归遍历其下的grpSp子节点即可，逻辑和上述解析流程一致。
• 高层XWPFShape API封装了常用属性读取方法，开发效率更高；如果需要读取特殊属性（比如形状的自定义XML属性、特效参数），直接操作底层CTShape对象即可拿到全量数据。
• 原有getAllDrawings方法不需要修改，已经可以正确拿到所有绘图节点。

内容的提问来源于stack exchange，提问作者Hervé Girod