Avro, mail # user - Re: How to deserialize avro file with union/many schemas? - 2014-07-23, 22:43
 Search Hadoop and all its subprojects:

Switch to Threaded View
Copy link to this message
Re: How to deserialize avro file with union/many schemas?

To see a union schema, do the following:
System.out.println (ReflectData.AllowNull.get().getSchema(YourClass.class));

And then do the following:
System.out.println (ReflectData.get().getSchema(YourClass.class));

Diff the two outputs.
First one generates a UNION of each and every field with a null.

Hope that helps.
Date: Wednesday, July 23, 2014 at 3:09 PM
Subject: Re: How to deserialize avro file with union/many schemas?

Hi Mike,

I read through most of the doc on avro site, don't see anything about the "union schema", Mike, would you mind give me some example here how the union schma is defined? also what package/method can retrieve the master schema from avro file? is that "getschema()" should work? and how to read in each Avro datums whithout knowing their corresponding schema?....

very much appreciate your help!
On Tue, Jul 22, 2014 at 10:25 PM, Michael Pigott <[EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>> wrote:

It's just a regular Union :-) http://avro.apache.org/docs/1.7.6/spec.html#Unions


On Jul 23, 2014 1:22 AM, "Echo" <[EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>> wrote:
Thanks Mike, it sounds make sense, is there any doc I can read about union schema?

On Jul 22, 2014, at 2:32 PM, Michael Pigott <[EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>> wrote:

    Just to make sure I understand you correctly - do you have a file with multiple Avro datums in it, each one following a separate schema?  And are all of these schemas unioned together in a file-level "master schema?"  (As far as I know, Avro file readers and writers only support one schema per file, so this is the only way your question makes sense to me.)
    If that's the case, then you can get the file's "master schema" and determine what all of the different types are:

List<Schema> allTypes = masterSchema.getTypes(); // Assumes masterSchema is of Type.UNION

Then when you read each Avro datum in the file, you can check which of the schemas it conforms to, and write a new file with just that sub-schema and the one datum in it.

Does that make sense?
On Tue, Jul 22, 2014 at 3:22 PM, Lewis John Mcgibbney <[EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>> wrote:
For the purpose of others on this list, can ytou please provide an example of your schema?
On Tue, Jul 22, 2014 at 12:06 PM, Echo Li <[EMAIL PROTECTED]<mailto:[EMAIL PROTECTED]>> wrote:

I'm new here, hope I can get help from you guys. Basically I have an avro file with union/many schemas and mixed records. I will need to split it to many avro file, one schema per file. All the stuff I've been reading is about serializing and deserializing avro file with one schema, which is pretty straightforward, but in my case I have no clue, any ideas?


NEW: Monitor These Apps!
elasticsearch, apache solr, apache hbase, hadoop, redis, casssandra, amazon cloudwatch, mysql, memcached, apache kafka, apache zookeeper, apache storm, ubuntu, centOS, red hat, debian, puppet labs, java, senseiDB