[
https://issues.apache.org/jira/browse/FLINK-29090?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17601826#comment-17601826
]
dalongliu edited comment on FLINK-29090 at 9/8/22 1:48 PM:
-----------------------------------------------------------
I didn't reproduce the exception through the following sql case in master
branch:
```java
@Test
publicvoidtestMap() throws Exception
{ super.tableEnv() .executeSql( String.format( "create table parquet_source
(\n" + "f1 int,\n" + " f_map map<varchar(20), int>\n" + ") with (\n" + "
'connector' = 'filesystem',\n" + " 'path' = '%s',\n" + " 'format' =
'parquet'\n" + ");", super.resultPath())); super.tableEnv().executeSql("insert
into parquet_source select 1, map['k1', 1]").await(); List<Row> results =
CollectionUtil.iteratorToList( super.tableEnv() .executeSql("select f1,
f_map['k1'] from parquet_source") .collect());
assertThat(results.size()).isEqualTo(1); }
```
In the codegen code, the ColumnarMapData will be convert to BinaryMapData
before get it, the codegen related code as following:
!image-2022-09-08-21-47-47-325.png!
Can you give me more context about reproducing the bug?
was (Author: lsy):
I didn't reproduce the exception through the following sql case:
```java
@Test
publicvoidtestMap() throws Exception {
super.tableEnv()
.executeSql(
String.format(
"create table parquet_source (\n"
+ "f1 int,\n"
+ " f_map map<varchar(20), int>\n"
+ ") with (\n"
+ " 'connector' = 'filesystem',\n"
+ " 'path' = '%s',\n"
+ " 'format' = 'parquet'\n"
+ ");",
super.resultPath()));
super.tableEnv().executeSql("insert into parquet_source select 1, map['k1',
1]").await();
List<Row> results =
CollectionUtil.iteratorToList(
super.tableEnv()
.executeSql("select f1, f_map['k1'] from parquet_source")
.collect());
assertThat(results.size()).isEqualTo(1);
}
```
> Fix the code gen for ColumnarMapData and ColumnarArrayData
> ----------------------------------------------------------
>
> Key: FLINK-29090
> URL: https://issues.apache.org/jira/browse/FLINK-29090
> Project: Flink
> Issue Type: Bug
> Components: Table SQL / Runtime
> Affects Versions: 1.16.0
> Reporter: Danny Chen
> Priority: Major
> Fix For: 1.16.0
>
> Attachments: image-2022-08-24-10-15-11-824.png,
> image-2022-09-08-21-47-47-325.png
>
>
> !image-2022-08-24-10-15-11-824.png|width=589,height=284!
> Currently, the code generation for {{MapData}} assumes that it is the
> {{{}GenericMapData{}}}, but the new introduced {{ColumnarMapData}} and
> {{ColumnarArrayData}} can not be casted to {{{}GenericMapData{}}}.
> {{ColumnarMapData}} and {{ColumnarArrayData}} are introduced in
> FLINK-24614
> [https://github.com/apache/flink/commit/5c731a37e1a8f71f9c9e813f6c741a1e203fa1a3]
> How to reproduce:
> {code:sql}
> create table parquet_source (
> f_map map<varchar(20), int>
> ) with (
> 'connector' = 'filesystem',
> 'format' = 'parquet'
> );
> select f_map['k1'] from table parquet_source;
> {code}
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
