Spark SQL 1.3 support - esDF function incorrectly maps column names

[difin]

Hi,

Environment:
Spark 1.3.0
elasticsearch 1.4.4
elasticsearch-hadoop 2.1.0.Beta4

When using esDF function to load data from Elasticsearch into Spark, columns get assigned to incorrect column names in the resulted DataFrame. The columns names that get assigned, they do exist in the index/type, but the mapping between the columns and column names is totally incorrect.

This bug doesn't happen when using the general Spark SQL 1.3 "load" function, columns are mapped to correct column names, but when using the load function instead of esDF and then doing JOIN with DataFrame loaded from Oracle, an exception happens (see issue #449: #449).

[jmahonin]

I ran into this same issue myself using 2.1.0-Beta4. It seems to be resolved in the master branch, possibly by this commit:
279e893

It's been some time, and a number of changes since the Beta4 release. A Beta5 (or full) 2.1.0 release would be great!

[costin]

@FDmitriy @jmahonin Indeed, it has. Can you please try out the snapshot/dev build and report back?

Thanks,

[costin]

Duplicate of #441