Read data from MongoDB (gridfs) via Matlab and Java driver

Question

I am working with Matlab and the Java driver. After writing some big data to the DB I want to read them out later. The files are about 100MB big and because of that I am writing them into GridFS.

I read the data with the Java driver into Matlab this way:

dbgridfs = GridFS(db, 'data_gridfs');
file1 = dbgridfs.findOne('bigdata');

After that code I will get an object in my Matlab workspace: Name Size Bytes Class file1 1x1 com.mongodb.gridfs.GridFSDBFile

Now I have a problem with converting the (Java?)-object into a nativ Matlab variable.

I searched a lot on different sites, but I don't get it. At the moment I am writing the data on my harddisk and after that I read it into a native Matlab variable - but this is a really dirty way and don't ask me relating to the performance :(

Are there any existing solutions I missed with the Java driver and/or do you know some code which could help me? Thanks for your help.

regards matl

Answer 1

I have never responded to any forum question, but I have benefitted a lot especially from stackoverflow, so I thought I have to give something back at least once. Since the above issue was bugging me now for nearly two days and I finally stitched together a solution I thought this would be a good topic to give something back to the community.

I had the same issue of retrieving an image (.png) from a MongoDB/GridFS with the Java driver. First, as also described above, you need to retrieve the file Java object:

    import com.mongodb.*;
    import com.mongodb.gridfs.*;

    mongoClient = MongoClient('server_name',27017);
    db = mongoClient.getDB('database_name');
    imgData = GridFS(db,'image_data');

The above code part is getting the collection with the image data. Afterwards, the image data Java object can be retrieved:

    javaIObj = imgData.findOne('image_name.png');

Now ByteArrayOutputStream is needed where the data stream can be piped into instead of a file:

    import java.io.ByteArrayOutputStream;

    baos = ByteArrayOutputStream();
    javaIObj.writeTo(baos);

The output stream needs again to be piped into a ByteArrayInputStream which can be used to construct an ImageIO object:

    import java.io.ByteArrayInputStream;
    import javax.imageio.ImageIO;    

    bais = ByteArrayInputStream(baos.toByteArray());
    jbi = ImageIO.read(bais);

Now the actual Matlab image as matrix can be retrieved and displayed:

    nrows = jbi.getHeight; ncols = jbi.getWidth;
    data = jbi.getData.getPixels(0,0,ncols,nrows,[]);
    matImg = reshape(data,ncols,nrows)';

    imagesc(matImg);

I do not know if this is the perfect solution, but it worked for me.

Cheers!