Josh Rosen

Even after fixing #19, the final clusterings still don't seem very intuitive:  I might try using cosine similarity instead of Euclidean distance to see if that gives us more...

There doesn't seem to be an incremental search option in Vrome. I'd like an option to make Vrome's search behave like Chrome's default search, so that the scroll position moves...

Actually, one trick that might be helpful is to define separate SBT tasks for deploying the dependency and user code assemblies; although this would be somewhat more manual than figuring...

It looks like this is duplicated by #27. We should definitely fix this.

@bdrillard, given that `ObjectType` now appears to be public in Spark, would you be interested in submitting a pull request to contribute your custom encoder to this project?

Per my comment at https://github.com/databricks/spark-avro/pull/89#issuecomment-263147247, this is a feature that we absolutely should add, so I'll gladly review any patches for this and will help to carry them across the...

@bdrillard, although it would be ugly, there's a possibility that we use an ugly hack to either construct `ObjectType` reflectively or to place a helper "shim" class into the `org.apache.spark.sql`...

I think you need to set this in the Hadoop configuration used by Spark, not in SparkConf itself. Try either ``` spark.sparkContext.hadoopConfiguration.set("avro.mapred.ignore.inputs.without.extension", "false") ``` or ``` val conf = new...

If we do this, it's maybe worth feature-flagging.

The case where another application has started writing to the file seems like it could manifest itself as corrupted / partially-readable files, so it sounds like you might want a...