Add support for Substrait Struct literals and type #10622
Conversation
|
@jonahgao second extracted part :)
|
Adds support for converting from DataFusion Struct ScalarValues into Substrait Struct Literals and back. All structs are assumed to be unnamed - ie fields are renamed into "c0", "c1", etc
Blizzara
force-pushed
the
avo/substrait-literal-structs
branch
from
4f0dfef to
bd964f8
Compare
There was a problem hiding this comment.
Thank you @Blizzara -- this looks good to me. I have one more test I suggested test, but I think we could do it as a follow on PR too.
Please let me know if you would prefer to modify this PR or make a follow on.
Thanks again for the contribution!
| @@ -2125,6 +2141,15 @@ mod test { | |||
| ), | |||
| )))?; | |||
|
|
|||
| let struct_field_1 = Field::new("c0", DataType::Boolean, true); | |||
| let struct_field_2 = Field::new("c1", DataType::Int32, true); | |||
| round_trip_literal( | |||
There was a problem hiding this comment.
I suggest one more test with null values too -- specifically something like
ScalarStructBuilder::new()
.with_scalar(struct_field_1, ScalarValue::Boolean(None))
.with_scalar(struct_field_2, ScalarValue::Int32(None))
.build()?,There was a problem hiding this comment.
Added (though incorporated into the existing test) in 84ccaa3!
Thank you for the review! I added the null case here :) |
|
Not sure what's up with the checks here - cargo check runs fine for me locally and I don't see how the errors would be related to my changes? |
CI is attempting to merge the current PR into a problematic commit d2f6faf commit 90ed2f64c675f0a52114b8fcdfccef45fa258181 (grafted, HEAD, pull/10622/merge)
Author: Arttu <[email protected]>
Date: Thu May 23 13:59:56 2024 +0200
Merge 84ccaa3361f64dc4a247487b43a57367734b9548 into d2f6faf56784867beafaa2ee88df37ffc3d20720I am not sure if pushing a new commit to the current PR will fix it. git commit --allow-empty -m "retry ci" |
|
Thanks @jonahgao , that seems to work indeed. This is ready to merge by me :) |
I think this was a logical conflict that @phillipleblanc fixed earlier today in #10637 🙏 |
* Add support for (un-named) Substrait Struct literal Adds support for converting from DataFusion Struct ScalarValues into Substrait Struct Literals and back. All structs are assumed to be unnamed - ie fields are renamed into "c0", "c1", etc * add converting from Substrait Struct type * cargo fmt --all * Unit test for NULL inside Struct * retry ci
Which issue does this PR close?
Closes #.
Extracted part of #10531 - not necessary part for it but somewhat related
Rationale for this change
What changes are included in this PR?
Adds support for converting from un-named DataFusion Struct ScalarValues into Substrait Struct Literals and back, as well as converting from Substrait Struct type into DF Struct type (other direction already existed).
Substrait doesn't contain names in its Struct fields. For e.g. initial schema, the names are provided separately. This PR only properly supports un-named structs - which in DataFusion get default field names of form "c0", "c1", ... Other structs can also be converted but they'll be renamed automatically into the default names.
The VirtualTable PR will then provide better support for named structs.
Are these changes tested?
Adds round-trip unit tests
Are there any user-facing changes?
More things are now supported, but I don't think Substrait support status is covered by documentation currently?