Koert Kuipers 2012-07-18, 03:05
Record names should match between the reader and writer. It may be
that the implementation does not check this in some cases.
If you've changed the record name you can alias it in the reader's
schema to the name in the writer's schema.
On Tue, Jul 17, 2012 at 8:05 PM, Koert Kuipers <[EMAIL PROTECTED]> wrote:
> i am a little confused about the roles of record schema names when using
> avro in hadoop.
> when i read from avro files in hadoop with a reader schema that has a
> different name but otherwise is a sound projection, then everything seems to
> work fine. however if the writer and reader schema names have different
> names, then field aliases don't seem to work anymore. is this what should be
> expected? and what is the rationale for this behavior? at one point the
> schema name seems to be ignored (how else can i read using a schema with a
> different name), but at another point it seems to matter (for field
> aliases). according to the avro spec the record schemas do not match...