Update from upstream #1
Merged
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…1721) ## Which issue does this PR close? PR fixes schema mismatch errors (similar to the example shown below) when using `IcebergCommitExec` with DataFusion. This occurs when `IcebergCommitExec` is not the top-level plan but is instead wrapped as the input to another plan node, for example when added by a custom optimization rule (cache invalidation step for example). >An internal error occurred. Internal error: PhysicalOptimizer rule 'OutputRequirements' failed. Schema mismatch. Expected original schema: Schema { fields: [Field { name: "count", data_type: UInt64, nullable: false, dict_id: 0, dict_is_ordered: false, metadata: {} }], metadata: {} }, got new schema: Schema { fields: [Field { name: "r_regionkey", data_type: Int32, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {"PARQUET:field_id": "1"} }, Field { name: "r_name", data_type: Utf8, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {"PARQUET:field_id": "2"} }, Field { name: "r_comment", data_type: Utf8, nullable: true, dict_id: 0, dict_is_ordered: false, metadata: {"PARQUET:field_id": "3"} }], metadata: {} }. This issue was likely caused by a bug in DataFusion's code. Please help us to resolve this by filing a bug report in our issue tracker: https://github.com/apache/datafusion/issues ## What changes are included in this PR? PR updates `compute_properties` logic to use target (output) schema instead of input schema. Below is example DataFusion `DataSinkExec` implementation demonstrating that properties must be created based on target schema, not input. https://github.com/apache/datafusion/blob/4eacb6046773b759dae0b3d801fe8cb1c6b65c0f/datafusion/datasource/src/sink.rs#L101C1-L117C6 ```rust impl DataSinkExec { /// Create a plan to write to `sink` pub fn new( input: Arc<dyn ExecutionPlan>, sink: Arc<dyn DataSink>, sort_order: Option<LexRequirement>, ) -> Self { let count_schema = make_count_schema(); let cache = Self::create_schema(&input, count_schema); Self { input, sink, count_schema: make_count_schema(), sort_order, cache, } } .... fn properties(&self) -> &PlanProperties { &self.cache } ``` ## Are these changes tested? Tested manually, expanded existing test to verify output schema, tested as part of [Spice Iceberg write automated tests](https://github.com/spiceai/spiceai/blob/trunk/crates/runtime/tests/iceberg/write/mod.rs)
## Which issue does this PR close? - Closes #1505 . ## What changes are included in this PR? - Adds `table_properties.rs` to hold and validate properties and set default values. Uses macros to simplify setting new properties. ## Are these changes tested? Yes
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
No description provided.