gengliangwang edited a comment on issue #23836: [SPARK-26915][SQL] DataFrameWriter.save() should write without schema validation URL: https://github.com/apache/spark/pull/23836#issuecomment-465428121 @rdblue Sorry if I cause some misunderstanding. Let's focus on the API `DataFrameWriter.save` here: 1. The expressions `AppendData` and `OverwriteByExpression` always get table schema and validate the output schema with table schema. This is a serious behavior change. Before we have a final agreement about file source V2 behavior(mostly likely we will keep the V1 behavior in this API), I would suggest to revert them for now. 2. Keeping ORC V2 working helps us to find problems, while disabling it might lead to problematic design which we won't notice until we enable the file source V2 again. The ORC V2 is implemented as per the API we have been working for a long time: `table -> write builder -> write`. The only problem is how we deal with save modes. This PR is not proposing to keep `SupportsSaveMode` or to say this is the final solution. I am suggesting that we should keep original behavior in this API, unless there is a clear solution that eventually the following partial development will work. The above solution of @cloud-fan is good directions to go. We can discuss the solution in tomorrow's meetup. For the code changes in `DataFrameWriter.save`, I think we should remove the new expressions for now.
---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: [email protected] With regards, Apache Git Services --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
