Open andygrove opened 2 weeks ago
@andygrove
Thank you for the issue description.
I would like to work on this issue. If no one else is working on it. Can I take it up ?
Definitely yes. Thanks @prashantksharma
I suppose for this issue it is okay to violate the rule to not modify a field in a protobuf message because we haven't published anything yet.
I suppose for this issue it is okay to violate the rule to not modify a field in a protobuf message because we haven't published anything yet.
Yes, absolutely.
@andygrove , cc: @viirya
Summary:
expr.proto
and QueryPlanSerde.scala
. The changes have been pushed to a separate branch on my fork. The comparison can be seen here: make test-rust
gives following error. Minor Query:
Before opening the PR, I would like to confirm: if I should go ahead and resolve the error for planner.rs
or a separate issue needs to be created for this ?
On the Rust side you will need a match
statement to convert the protobuf i32 to the Rust enum (0 -> legacy, 1 -> try, 2 -> ansi). Perhaps take a look at how we handle one of the other enums defined in the proto and follow the same pattern?
@prashantksharma also, feel free to create a draft PR as it can be easier for maintainers to make suggestions on the PR
@andygrove
Thank you so much for your feedback:
On the Rust side you will need a match statement to convert the protobuf i32 to the Rust enum (0 -> legacy, 1 -> try, 2 -> ansi). Perhaps take a look at how we handle one of the other enums defined in the proto and follow the same pattern?
I will make necessary changes based on my understanding of the above comment by you.
And, Thank you for your suggesting creation of draft PR, after making necessary updates, I will do that.
@andygrove cc: @viirya
I have opended a draft PR.
I have tested the changes using
make test-rust
make test-jvm
Details on PR message:https://github.com/apache/datafusion-comet/pull/415
What is the problem the feature request solves?
When I added
eval_mode
to theCast
message inexpr.proto
I defined it as a string but it should really be an enum.Describe the potential solution
No response
Additional context
No response