Support dictionary literals in Substrait producer#22041
Draft
alexanderbianchi wants to merge 1 commit intoapache:branch-53from
Draft
Support dictionary literals in Substrait producer#22041alexanderbianchi wants to merge 1 commit intoapache:branch-53from
alexanderbianchi wants to merge 1 commit intoapache:branch-53from
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Rationale for this change
The Substrait producer currently fails when a logical plan contains a
ScalarValue::Dictionary, for example a string literal that was coerced to match a dictionary-encoded string column. Dictionary encoding is a physical representation detail; the Substrait literal should carry the logical inner value.A bit more context on why this takes the approach of serializing the inner value:
There are two unrelated "dictionary" concepts that are easy to conflate here:
This PR is about the second one. The logical value is still just a string.
The failing path we hit was:
where the table schema exposes
metric_nameas:DataFusion type coercion makes both sides of the equality compatible, so the predicate becomes conceptually:
That scalar is not a map/object value. It is DataFusion representing a string scalar that has been coerced to match a dictionary-encoded string column.
The Substrait producer then failed with:
In Substrait, there is no useful distinction between a string scalar and a "dictionary-encoded string scalar" here. Dictionary encoding is meaningful for arrays/columns, not for a single scalar literal. So the intended encoding is just the logical literal value:
The column/scan can still be dictionary encoded when the plan is consumed against a table schema where
metric_nameisDictionary(Int32, Utf8). At that point DataFusion can again apply its normal coercion/execution behavior for comparing the dictionary column to the string literal.So the key point is: this PR is not trying to encode dictionary array layout into Substrait literals. It is preserving the logical scalar value while avoiding a producer failure caused by DataFusion's internal
ScalarValue::Dictionaryrepresentation after type coercion.What changes are included in this PR?
This unwraps
ScalarValue::Dictionary(_, value)in the Substrait literal producer and serializes the inner scalar value. It also adds a regression test showing a dictionary-encoded UTF8 scalar is emitted as a string literal and round trips as the logical UTF8 scalar.Are these changes tested?
Yes: