[FLINK-32674][doc] Add documentation for the new getTargetColumns in DynamicTableSink #23075
Conversation
…umns in DynamicTableSink
luoyuxia
left a comment
luoyuxia
left a comment
There was a problem hiding this comment.
@lincoln-lil Thanks for the pr. I left minor comments. PTAL.
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| For the connector developers who want to avoid overwriting non-target columns with null values when processing partial column updates, | ||
| this user specified target column list can be found from the {{< gh_link file="flink-table/flink-table-common/src/main/java/org/apache/flink/table/connector/sink/DynamicTableSink.java" name="DynamicTableSink$Context.getTargetColumns()" >}}. |
There was a problem hiding this comment.
nit
| this user specified target column list can be found from the {{< gh_link file="flink-table/flink-table-common/src/main/java/org/apache/flink/table/connector/sink/DynamicTableSink.java" name="DynamicTableSink$Context.getTargetColumns()" >}}. | |
| this user-specified target column list can be found from the {{< gh_link file="flink-table/flink-table-common/src/main/java/org/apache/flink/table/connector/sink/DynamicTableSink.java" name="DynamicTableSink$Context.getTargetColumns()" >}}. |
There was a problem hiding this comment.
if the above new version is ok, then this dash will not be added
| Given a table T(a INT, b INT, c INT), Flink supports INSERT INTO T(c, b) SELECT x, y FROM S. The expectation is | ||
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| For the connector developers who want to avoid overwriting non-target columns with null values when processing partial column updates, |
There was a problem hiding this comment.
It seems like we miss how the connector devs can avoid it? may be some like:
For the connector .... when processing partial column updates, they can find the user-specified target column list and only update the target columns.
?
There was a problem hiding this comment.
For connector developers who want to avoid overwriting non-target columns with null values when processing partial column updates, you can get the information about the target columns specified by the user's insert statement from 'getTargetColumns(link...)' and decide how to process the partial updates.
There was a problem hiding this comment.
The corresponding chinese version: 连接器开发人员在处理部分列更新时，如果希望避免用空值覆盖非目标列，可以从 "getTargetColumns(link...) "中获取用户插入语句指定的目标列信息，然后决定如何处理部分更新。
There was a problem hiding this comment.
Thanks, I like the tweaked description.
| Given a table T(a INT, b INT, c INT), Flink supports INSERT INTO T(c, b) SELECT x, y FROM S. The expectation is | ||
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| For the connector developers who want to avoid overwriting non-target columns with null values when processing partial column updates, |
There was a problem hiding this comment.
Also, I'm wondering whether it's more clear to break the line in here since these lines are more likely to say other things.
There was a problem hiding this comment.
current version: 
a new line seems to be better, will update it.
lincoln-lil
left a comment
lincoln-lil
left a comment
There was a problem hiding this comment.
@luoyuxia thanks for reviewing this! What do you think of the tweaked description?
| Given a table T(a INT, b INT, c INT), Flink supports INSERT INTO T(c, b) SELECT x, y FROM S. The expectation is | ||
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| For the connector developers who want to avoid overwriting non-target columns with null values when processing partial column updates, |
There was a problem hiding this comment.
For connector developers who want to avoid overwriting non-target columns with null values when processing partial column updates, you can get the information about the target columns specified by the user's insert statement from 'getTargetColumns(link...)' and decide how to process the partial updates.
| Given a table T(a INT, b INT, c INT), Flink supports INSERT INTO T(c, b) SELECT x, y FROM S. The expectation is | ||
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| For the connector developers who want to avoid overwriting non-target columns with null values when processing partial column updates, |
There was a problem hiding this comment.
current version:
a new line seems to be better, will update it.
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| For the connector developers who want to avoid overwriting non-target columns with null values when processing partial column updates, | ||
| this user specified target column list can be found from the {{< gh_link file="flink-table/flink-table-common/src/main/java/org/apache/flink/table/connector/sink/DynamicTableSink.java" name="DynamicTableSink$Context.getTargetColumns()" >}}. |
There was a problem hiding this comment.
if the above new version is ok, then this dash will not be added
| Given a table T(a INT, b INT, c INT), Flink supports INSERT INTO T(c, b) SELECT x, y FROM S. The expectation is | ||
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| that 'x' is written to column 'c' and 'y' is written to column 'b' and 'a' is set to NULL (assuming column 'a' is nullable). | ||
| For the connector developers who want to avoid overwriting non-target columns with null values when processing partial column updates, |
There was a problem hiding this comment.
The corresponding chinese version: 连接器开发人员在处理部分列更新时，如果希望避免用空值覆盖非目标列，可以从 "getTargetColumns(link...) "中获取用户插入语句指定的目标列信息，然后决定如何处理部分更新。
luoyuxia
left a comment
luoyuxia
left a comment
There was a problem hiding this comment.
@lincoln-lil Thanks for updating. LGTM.
3 participants
What is the purpose of the change
This is an addition to the documentation for the newly added interface in FLINK-31487. It may be beneficial for connector developers who want to avoid overwriting non-target columns with null values when processing partial column updates.
Brief change log
Update the current sql doc dev/table/sql/insert.md