Reading a BlobstoreInputStream >= 1MB in size

问题

Reading more than 1mb of data from a BlobstoreInputStream will throw an IOException "Blob fetch size too large."

You can use the ChainedBlobstoreInputStream class in the answer below to solve this problem.

回答1:

I've created a simple wrapper class that solves this problem. You should be able to directly swap ChainedBlobstoreInputStream for BlobstoreInputStream. I haven't tested the mark(), markSupported() or reset() methods.

You may use this code however you wish.

package net.magicscroll.server.blobstore;

import com.google.appengine.api.blobstore.BlobInfo;
import com.google.appengine.api.blobstore.BlobInfoFactory;
import com.google.appengine.api.blobstore.BlobKey;
import com.google.appengine.api.blobstore.BlobstoreInputStream;

import java.io.IOException;
import java.io.InputStream;

/**
 * ChainedBlobstoreInputStream works exactly like BlobstoreInputStream but does
 * not throw an error if more than 1mb is read.
 * 
 * @author Richard Wallis
 */
public class ChainedBlobstoreInputStream extends InputStream {
    /**
     * The maximum number of bytes that can be read by a single request.
     */
    private static final int MAX_READSIZE = 1015800;

    /**
     * The BlobKey of the blobstore item.
     */
    private BlobKey blobKey;
    /**
     * The current byte position of the reader.
     */
    private long offset;

    /**
     * The Total Size of the blob.
     */
    private long totalSize;
    /**
     * The current Input Stream being read.
     */
    private BlobstoreInputStream currentStream;
    /**
     * The next point at which a new InputStream will need to be initialized.
     */
    private long nextReadBreak;
    /**
     * The currentStream at the time of the last mark.
     */
    private BlobstoreInputStream markedStream;

    /**
     * Creates a new ChainedBlobstoreInputStream. This stream should behave
     * exactly the same as a BlobstoreInputStream and it should be possible to
     * interchange them.
     * 
     * @param theBlobKey
     *            - The blobkey of the object to be read.
     * @throws IOException
     *             - Thrown if there is an error reading the current stream.
     */
    public ChainedBlobstoreInputStream(final BlobKey theBlobKey)
            throws IOException {
        this(theBlobKey, 0);
    }

    /**
     * Creates a new ChainedBlobstoreInputStream. This stream should behave
     * exactly the same as a BlobstoreInputStream and it should be possible to
     * interchange them.
     * 
     * @param theBlobKey
     *            - The blobkey of the object to be read.
     * @param newOffset
     *            - The offset in the blob from where to read.
     * @throws IOException
     *             - Thrown if there is an error reading the current stream.
     */
    public ChainedBlobstoreInputStream(final BlobKey theBlobKey,
            final long newOffset) throws IOException {
        this.offset = newOffset;
        this.blobKey = theBlobKey;
        final BlobInfo blobInfo = 
            new BlobInfoFactory().loadBlobInfo(this.blobKey);
        this.totalSize = blobInfo.getSize();
        this.currentStream = 
            new BlobstoreInputStream(this.blobKey, this.offset);
        this.nextReadBreak = 
            this.offset + ChainedBlobstoreInputStream.MAX_READSIZE;
    }

    /*
     * (non-Javadoc)
     * @see java.io.InputStream#read()
     */
    @Override
    public final int read() throws IOException {
        if (this.offset < this.totalSize) {
            if (this.offset == this.nextReadBreak) {
                this.currentStream.close();
                this.currentStream = 
                    new BlobstoreInputStream(this.blobKey, this.offset);
                this.nextReadBreak = this.offset 
                + ChainedBlobstoreInputStream.MAX_READSIZE;
            }
            this.offset += 1;
            return this.currentStream.read();
        } else {
            this.currentStream.close();
            return -1;
        }
    }

    /*
     * (non-Javadoc)
     * @see java.io.InputStream#close()
     */
    @Override
    public final void close() throws IOException {
        this.currentStream.close();
        super.close();
    }

    /*
     * (non-Javadoc)
     * @see java.io.InputStream#mark(int)
     */
    @Override
    public final void mark(final int readlimit) {
        this.currentStream.mark(readlimit);
        this.markedStream = this.currentStream;
    }

    /*
     * (non-Javadoc)
     * @see java.io.InputStream#markSupported()
     */
    @Override
    public final boolean markSupported() {
        return this.currentStream.markSupported();
    }

    /*
     * (non-Javadoc)
     * @see java.io.InputStream#reset()
     */
    @Override
    public final void reset() throws IOException {
        this.currentStream = this.markedStream;
        this.currentStream.reset();
    }

}

回答2:

while ((size = zis.read(buffer)) > 0)
  contents.write(buffer, 0, size);

But the whole concept of reading entire files into memory has knobs on it. Find a way to process it as you read it.

or I get an error.

What error?

来源：https://stackoverflow.com/questions/3410886/reading-a-blobstoreinputstream-1mb-in-size