-
Notifications
You must be signed in to change notification settings - Fork 412
Read partitioned tables with source field missing from schema #2367
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
| source_field = schema.find_field(field.source_id) | ||
| result_type = field.transform.result_type(source_field.field_type) | ||
| nested_fields.append(NestedField(field.field_id, field.name, result_type, required=source_field.required)) | ||
| else: |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Wanted to get some opinions for just allowing this for VoidTransforms fields, as as of now we can drop columns without dropping the partition first
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Unfortunally, dropping is not an option. The V1 tables do not have field-IDs encoded in the struct, and is purely positional based. See the spec for details. Dropping a field, would change the position, potentially resulting in data integrity issues.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@Fokko I was referring more to that if the field's transform is not VoidTransform then we should potentially fail here. Java does not allow dropping a schema column if there is an non-void partition referencing that field, but here it is not guaranteed that if there is no source-field for a partition then the partition will be void.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@Fokko gentle pin as this might gone off radar :)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think this is fine, if I understand the concern correctly. The worry is that we're allowing unknown type for non void transforms when the source field is missing.
On the Java side, when we're reading partition specs from the metadata, we are using the allow missing fields equal to true, which skips the validation if the source is missing(here).
This validation is only hit when we're constructing new partition specs. When we're reading, it's fine.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Sorry, I think I misunderstood the question, but nothing is being dropped here, so that all looks good.
a7b118b to
d93868c
Compare
d93868c to
65b14f9
Compare
Fokko
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks @gabeiglio for fixing this, and thanks @geruh for the review 🙌
Rationale for this change
Following with the Java solution implementation on how to read partition specs when a source field was dropped.
Are these changes tested?
Yes, added one integration tests, and one unit test
Are there any user-facing changes?
No