Hi all,

Will Berkeley has been working on some proposed updates updates to the Kudu
Spark connector (see the code in review here
<https://gerrit.cloudera.org/#/c/3871>).  The high level overview of the
patch is

* Explicit support for inserting, dropping, upserting, and deleting rows in
DataFrames into Kudu tables (via methods on KuduContext).

* Possibly dropping the implementation of CreatableRelationProvider.  I
originally suggested this because the Spark connector can't create a table
without specifying many table properties that aren't part of the
interface.  Since then, I've been convinced by Chris George that this is a
bad idea in terms of usability, and that we should just limit the
implementation to only support the "append" save mode, and error if the
table doesn't exist.

If you use the Kudu Spark connector, please take a look at the patch and
weigh in!  We're trying to get this in for the upcoming 0.10 release.

- Dan

Reply via email to