[
https://issues.apache.org/jira/browse/SPARK-61294?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Serge Rielau updated SPARK-61294:
---------------------------------
Attachment: (was: object-type.md)
> OBJECT type support
> -------------------
>
> Key: SPARK-61294
> URL: https://issues.apache.org/jira/browse/SPARK-61294
> Project: Spark
> Issue Type: Task
> Components: SQL
> Affects Versions: 5.0.0
> Reporter: Serge Rielau
> Priority: Major
>
> {{STRUCT}} fields are matched by position. Assignment, comparison, and set
> operations ignore field names, so a reordered query or table can write values
> into the wrong fields without an error. {{OBJECT}} makes field names part of
> the type: assignment, comparison, and set operations require the same names
> in the same order.
> Two further gaps motivate the type:
> * VARIANT objects have named fields but no declared order or schema.
> {{to_variant_object}} sorts fields, and {{schema_of_variant}} returns
> {{OBJECT<...>}} only as a string that cannot be stored. {{OBJECT}} gives a
> typed, storable home to a VARIANT object with a stated signature.
> * Snowflake's structured {{OBJECT}} is the common precedent for users moving
> semi-structured data. Matching its functions ({{{}object_delete{}}},
> {{{}object_pick{}}}, {{{}ADD FIELDS{}}}, {{{}RENAME FIELDS{}}}) lowers
> migration cost. The one deliberate rename is {{{}OBJECT_INSERT{}}}, which
> becomes {{{}object_append{}}}.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
