Closed ZENOTME closed 1 week ago
Sorry for late reply, let's discuss here:https://github.com/apache/iceberg-rust/issues/398
Can we at least split the code that makes Predicate
serializable out of this PR into it's own PR? I think that is pretty uncontroversial and very useful on its own.
Can we at least split the code that makes
Predicate
serializable out of this PR into it's own PR? I think that is pretty uncontroversial and very useful on its own.
I agree with @sdd about this and suggest to split it into three prs:
We can still keep this pr open and waiting for 1 and 2 to get merged.
Can we at least split the code that makes
Predicate
serializable out of this PR into it's own PR? I think that is pretty uncontroversial and very useful on its own.I agree with @sdd about this and suggest to split it into three prs:
- Ser/de of datum
- Ser/de of expression
- Refine arrow builder
We can still keep this pr open and waiting for 1 and 2 to get merged.
I separate a PR for 1 and 2 #406 because I find that 2 is dependent on 1. Let's complete it first.
After #406 is complete, let's move on this.
cc @sdd Do you have other concerns?
No more concerns - looks good, thanks! In fact I've been already using this in some prototypes in anticipation of it getting merged, it works well 🙌🏼
This PR is a draft for #398. If it looks good, I will fill out the test later. The basic here is to move the info needed for the reader to FileScanTask. In this way, we can avoid the inconsistency between the reader and FileScanTask and it's more friendly for the user to use the reader.
To store the predicate in FileScanTask, this PR also makes the predicate serializable and deserializable.