[
https://issues.apache.org/jira/browse/SPARK-34365?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Erik Krogen updated SPARK-34365:
--------------------------------
Description:
When reading an Avro dataset (using the dataset's schema or by overriding it
with 'avroSchema') or writing an Avro dataset with a provided schema by
'avroSchema', currently the matching of Catalyst-to-Avro fields is done by
field name.
This behavior is somewhat recent; prior to SPARK-27762 (fixed in 3.0.0), at
least on the write path, we would match the schemas by positionally
("structural" comparison). While I agree that this is much more sensible for
default behavior, I propose that we make this behavior configurable using an
{{option}} for the Avro datasource. Even at the time that SPARK-27762 was
handled, there was [interest in making this behavior
configurable|https://github.com/apache/spark/pull/24635#issuecomment-494205251],
but it appears it went unaddressed.
There is precedence for configurability of this behavior as seen in
SPARK-32864, which added this support for ORC. Besides this precedence, the
behavior of Hive is to perform matching positionally
([ref|https://cwiki.apache.org/confluence/display/Hive/AvroSerDe#AvroSerDe-WritingtablestoAvrofiles]),
so this is behavior that Hadoop/Hive ecosystem users are familiar with:
{quote}
Hive is very forgiving about types: it will attempt to store whatever value
matches the provided column in the equivalent column position in the new table.
No matching is done on column names, for instance.
{quote}
was:
When reading an Avro dataset (using the dataset's schema or by overriding it
with 'avroSchema') or writing an Avro dataset with a provided schema by
'avroSchema', currently the matching of Catalyst-to-Avro fields is done by
field name.
This behavior is somewhat recent; prior to SPARK-27762 (fixed in 3.0.0), at
least on the write path, we would match the schemas by positionally
("structural" comparison). While I agree that this is much more sensible for
default behavior, I propose that we make this behavior configurable using an
{{option}} for the Avro datasource.
There is precedence for configurability of this behavior as seen in
SPARK-32864, which added this support for ORC. Besides this precedence, the
behavior of Hive is to perform matching positionally
([ref|https://cwiki.apache.org/confluence/display/Hive/AvroSerDe#AvroSerDe-WritingtablestoAvrofiles]),
so this is behavior that Hadoop/Hive ecosystem users are familiar with:
{quote}
Hive is very forgiving about types: it will attempt to store whatever value
matches the provided column in the equivalent column position in the new table.
No matching is done on column names, for instance.
{quote}
> Support configurable Avro schema field matching for positional or by-name
> -------------------------------------------------------------------------
>
> Key: SPARK-34365
> URL: https://issues.apache.org/jira/browse/SPARK-34365
> Project: Spark
> Issue Type: Improvement
> Components: SQL
> Affects Versions: 3.0.1
> Reporter: Erik Krogen
> Priority: Major
>
> When reading an Avro dataset (using the dataset's schema or by overriding it
> with 'avroSchema') or writing an Avro dataset with a provided schema by
> 'avroSchema', currently the matching of Catalyst-to-Avro fields is done by
> field name.
> This behavior is somewhat recent; prior to SPARK-27762 (fixed in 3.0.0), at
> least on the write path, we would match the schemas by positionally
> ("structural" comparison). While I agree that this is much more sensible for
> default behavior, I propose that we make this behavior configurable using an
> {{option}} for the Avro datasource. Even at the time that SPARK-27762 was
> handled, there was [interest in making this behavior
> configurable|https://github.com/apache/spark/pull/24635#issuecomment-494205251],
> but it appears it went unaddressed.
> There is precedence for configurability of this behavior as seen in
> SPARK-32864, which added this support for ORC. Besides this precedence, the
> behavior of Hive is to perform matching positionally
> ([ref|https://cwiki.apache.org/confluence/display/Hive/AvroSerDe#AvroSerDe-WritingtablestoAvrofiles]),
> so this is behavior that Hadoop/Hive ecosystem users are familiar with:
> {quote}
> Hive is very forgiving about types: it will attempt to store whatever value
> matches the provided column in the equivalent column position in the new
> table. No matching is done on column names, for instance.
> {quote}
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
