[
https://issues.apache.org/jira/browse/FLINK-41537?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Martijn Visser updated FLINK-41537:
-----------------------------------
Release Note: Column names in the Cassandra SQL source are case-sensitive,
as in the JDBC connector. A Flink column MyCol reads the Cassandra column
"MyCol", not mycol. Declare the Flink column in lower case to read a column
that was created without quotes.
> CassandraSource can fail on tables with a composite or case-sensitive
> partition key
> -----------------------------------------------------------------------------------
>
> Key: FLINK-41537
> URL: https://issues.apache.org/jira/browse/FLINK-41537
> Project: Flink
> Issue Type: Bug
> Components: Connectors / Cassandra
> Affects Versions: cassandra-3.1.0, cassandra-3.2.0
> Reporter: Martijn Visser
> Priority: Major
> Fix For: cassandra-4.0.0, cassandra-3.3.0
>
>
> The source builds {{token(...)}} over the partition key columns. Since
> FLINK-38042 the columns are collected into a {{HashSet}}, so a composite key
> can come out in the wrong order, and case-sensitive names have never been
> quoted. Both make the read fail:
> * {{PRIMARY KEY ((z_key, a_key))}}: {{InvalidQueryException: The token
> function arguments must be in the partition key order: z_key, a_key}}
> * {{"MyKey" int PRIMARY KEY}}: {{InvalidQueryException: Undefined column
> name mykey in table flink.pk_quoted}}
> Through SQL the select list is built from unquoted field names as well, so
> a case-sensitive column fails there too ({{Undefined column name mykey}} on a
> Flink 2.3 cluster).
> Reproduced on main (3146c5b) against Cassandra 4.1.12. The PR for
> FLINK-41444 fixes both on driver 4, but the fix needs its own commit for v3.3.
--
This message was sent by Atlassian Jira
(v10.3.26#10030026)