[
https://issues.apache.org/jira/browse/SPARK-61074?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Hyukjin Kwon reassigned SPARK-61074:
------------------------------------
Assignee: Hyukjin Kwon
> Write and read the collations of string fields in the metadata of the JSON of
> a StructType, and order the keys of jsonValue, in the Python client as
> PySpark does
> -----------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Key: SPARK-61074
> URL: https://issues.apache.org/jira/browse/SPARK-61074
> Project: Spark
> Issue Type: Bug
> Components: Connect
> Affects Versions: connect-rust-4.3.0
> Reporter: Hyukjin Kwon
> Assignee: Hyukjin Kwon
> Priority: Major
> Labels: pull-request-available
> Fix For: connect-rust-4.3.0
>
>
> In PySpark Connect, {{StructType.json()}} and {{jsonValue()}} write the
> collation of a string field, also of one in an array or a map, in the
> {{__COLLATIONS}} of the metadata of the field, as Spark does, with the type
> of the field as "string":
> {{{"metadata":{"__COLLATIONS":{"s":"spark.UTF8_LCASE"}},"name":"s",...,"type":"string"}}}.
> {{StructField.fromJson}}, {{StructType.fromJson}}, and the {{fromJson}} of
> {{ArrayType}} and {{MapType}} (with their {{fieldPath}} and
> {{collationsMap}}) read them back. They raise {{COLLATION_INVALID_PROVIDER}}
> for an unknown provider, and {{INVALID_JSON_DATA_TYPE_FOR_COLLATIONS}} for a
> collation of a field that is not a string. The dicts of {{jsonValue()}} have
> their keys in a fixed order, such as name, type, nullable and metadata for a
> field.
> The Rust-backed Python client differs:
> * It writes the type of the field as "string collate UTF8_LCASE", without the
> metadata.
> * {{ArrayType.fromJson}} and {{MapType.fromJson}} ignore their
> {{collationsMap}}.
> * {{StructType.fromJson}} and {{StructField.fromJson}} raise
> {{INVALID_DATATYPE_FORMAT}} for a field of a user-defined type.
> * {{jsonValue()}} sorts its keys.
> Write and read them as PySpark does.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
