PDFBox 锁定 JPEG 输入文件，直到应用程序退出

Question

我在Windows 7环境中使用PDFBox RC2，Java 1.8_66。我正在使用它从 200dpi 页面大小的图像文件集合（JPEG 和 PNG）创建 PDF。

事实证明，当将JPEG文件添加到PDF时，PDImageXObject.createFromFile（）例程无法关闭内部文件句柄，从而在应用程序的生命周期内锁定图像文件。将PNG文件添加到PDF时，没有问题。

下面是重现该问题的一些示例代码。使用进程资源管理器（从系统内部），查看 java.exe 进程的打开文件句柄并运行此代码。我的测试使用了大约 20 个全尺寸的 JPEG 文件。请注意，方法退出后，仍保留几个锁定的文件。

public Boolean CreateFromImages_Broken(String pdfFilename, String[] imageFilenames) {
    PDDocument doc = new PDDocument();        
    for (String imageFilename : imageFilenames) {
        try {
            PDPage page = new PDPage();
            doc.addPage(page);
            PDImageXObject pdImage = PDImageXObject.createFromFile(imageFilename, doc);
            // at this point, if the imageFilename is a jpeg, pdImage holds onto a handle for 
            // the given imageFilename and that file remains locked until the application is closed
            try (PDPageContentStream contentStream = new PDPageContentStream(doc, page)) {
                float scale = (float)72.0 / 200;
                page.setMediaBox(new PDRectangle((int)(pdImage.getWidth() * scale), (int)(pdImage.getHeight() * scale)));
                contentStream.drawImage(pdImage, 0, 0, pdImage.getWidth()*scale, pdImage.getHeight()*scale);
            }
        } catch (IOException ioe) {
            return false;
        }                          
    }
    try {
        doc.save(pdfFilename);
        doc.close();                  
    } catch (IOException ex) {
        return false;
    }
    return true;
} 

Answer 1

作为一种解决方法，我查看了 PNG 和 JPEG 处理的源代码，并且通过实现这一点取得了成功，这似乎适用于两种文件类型：

public Boolean CreateFromImages_FIXED(String pdfFilename, String[] imageFilenames) {
    PDDocument doc = new PDDocument();        
    for (String imageFilename : imageFilenames) {
        FileInputStream fis = null;
        try {
            PDPage page = new PDPage();
            doc.addPage(page);
            PDImageXObject pdImage = null;
            // work around JPEG issue by opening up our own stream, with which
            // we can close ourselves instead of PDFBOX leaking it. For PNG
            // images, the createFromFile seems to be OK
            if (imageFilename.toLowerCase().endsWith(".jpg")) {
                fis = new FileInputStream(new File(imageFilename));
                pdImage = JPEGFactory.createFromStream(doc, fis);
            } else {
                pdImage = PDImageXObject.createFromFile(imageFilename, doc);
            }
            try (PDPageContentStream contentStream = new PDPageContentStream(doc, page)) {
                float scale = (float)72.0 / 200;
                page.setMediaBox(new PDRectangle((int)(pdImage.getWidth() * scale), (int)(pdImage.getHeight() * scale)));
                contentStream.drawImage(pdImage, 0, 0, pdImage.getWidth()*scale, pdImage.getHeight()*scale);                    
                if (fis != null) {
                    fis.close();
                    fis = null;
                }                    
            }
        } catch (IOException ioe) {
            return false;
        }                          
    }
    try {
        doc.save(pdfFilename);
        doc.close();                  
    } catch (IOException ex) {
        return false;
    }
    return true;        
}