NPE in JPEGImageReader.getSourceCSType in case of mime mixup

[huxi]

This is a followup to my comment in issue #22.

The exception is thrown with the following stack trace:

com.twelvemonkeys.imageio.plugins.jpeg.JPEGImageReader.getSourceCSType(Unknown Source)
com.twelvemonkeys.imageio.plugins.jpeg.JPEGImageReader.read(Unknown Source)
javax.imageio.ImageReader.read(ImageReader.java:939)

The code causing the issue looks like this:

ImageInputStream iis = ImageIO.createImageInputStream(new BufferedInputStream(inputStream));
Iterator readers = ImageIO.getImageReadersByMIMEType(mimeType);
ImageReader imageReader = (ImageReader) readers.next();
imageReader.setInput(iis, false);
BufferedImage bufferedImage = imageReader.read(0);

The mimeType is resolved by us based on the file extension so this is essentially our fault.

I thought I'd file this issue anyway so that other people hitting a similar problem can find it.

Is there a better way to create a BufferedImage from an InputStream, i.e. one that would resolve the actual mime type instead of using an inferred one to resolve the proper ImageReader?

[haraldk]

Hi Jörn,

Thanks for reporting!

I think the recommended way of obtaining an ImageReader is as follows:

ImageInputStream input = ImageIO.createInputStream(...);
Iterator<ImageReader> readers = ImageIO.getImageReaders(input);

if (readers.hasNext()) {
    ImageReader reader = readers.next();
    reader.setInput(input);

    // etc...
}

This way, each plugin provider (ImageReaderSpi instance) will be passed the actual payload (input in the example above), to see if this is a format it understands, and if so, it can create an ImageReader instance to read it.

I prefer this way, because file extensions can be wrong and there are often multiple extensions for each format. MIME types from browsers can also not be fully trusted. Inspecting the actual payload is just safer and more reliable (although there can be false positives here as well).

Best regards,

Harald K

[huxi]

Thanks a lot!