PartitionSpec Schema error

[Zhanxiao-Ma]

Apache Iceberg version

1.5.2

Query engine

None

Please describe the bug 🐞

The table schema: a, b, c, with the schemaId 0. Column a is selected as the partition column to obtain PartitionSpec1. Then, the partition information on column a is deleted to obtain PartitionSpec2, followed by the deletion of column a, resulting in schemaId1. When retrieving PartitionSpec1 again, it is found that its schemaId is 1, whereas theoretically, the schemaId should remain 0.

Willingness to contribute

• I can contribute a fix for this bug independently
• I would be willing to contribute a fix for this bug with guidance from the Iceberg community
• I cannot contribute a fix for this bug at this time

[amogh-jahagirdar]

@Zhanxiao-Ma I'm not sure I completely followed the issue. Do you have a reproducible example via reference implementation or sequence of SQL for an engine? You may be hitting an issue that was fixed in #11868

When retrieving PartitionSpec1 again, it is found that its schemaId is 1, whereas theoretically, the schemaId should remain 0.

Partition specs and schemas evolve independently so long as if the partition spec is compatible with the schema (implementations check to make sure the current spec is compatible with the current spec). Other than time travel queries to a specific snapshot, the current table schema is always used on read.

That PR I referenced addressed an issue where files written with historical partition specs would fail to be read after dropping a column that was referenced in those historical specs.

[Zhanxiao-Ma]

@Zhanxiao-Ma I'm not sure I completely followed the issue. Do you have a reproducible example via reference implementation or sequence of SQL for an engine? You may be hitting an issue that was fixed in #11868我不确定我是否完全关注了这个问题。您是否有一个通过引擎的参考实现或 SQL 序列的可重现示例？您可能遇到了在 #11868

When retrieving PartitionSpec1 again, it is found that its schemaId is 1, whereas theoretically, the schemaId should remain 0.再次检索 PartitionSpec1 时，发现其 schemaId 为 1，而理论上，schemaId 应保持 0。

Partition specs and schemas evolve independently so long as if the partition spec is compatible with the schema (implementations check to make sure the current spec is compatible with the current spec). Other than time travel queries to a specific snapshot, the current table schema is always used on read.只要分区规范与模式兼容，分区规范和模式就会独立发展（实现检查以确保当前规范与当前规范兼容）。除了对特定快照的时间旅行查询外，读取时始终使用当前表架构。

That PR I referenced addressed an issue where files written with historical partition specs would fail to be read after dropping a column that was referenced in those historical specs.我引用的那个 PR 解决了一个问题，即在删除这些历史规范中引用的列后，使用历史分区规范编写的文件将无法读取。

Yes, that's the question. Thank you for your answer.