feat(duckdb): transpile BigQuery IN UNNEST(array) to ARRAY_CONTAINS#7892
Open
sirockin wants to merge 2 commits into
Open
feat(duckdb): transpile BigQuery IN UNNEST(array) to ARRAY_CONTAINS#7892sirockin wants to merge 2 commits into
sirockin wants to merge 2 commits into
Conversation
…TAINS BigQuery uses `value IN UNNEST(array_expr)` to test array membership. DuckDB has no `UNNEST` in this position but provides `ARRAY_CONTAINS`. The transpilation wraps the result in `COALESCE(..., FALSE)` to handle NULL arrays safely — `ARRAY_CONTAINS(NULL, x)` returns NULL in DuckDB, but BigQuery's `IN UNNEST(NULL)` returns FALSE. Adds: - `IN UNNEST(arr)` → `COALESCE(ARRAY_CONTAINS(arr, val), FALSE)` in DuckDB generator via `in_sql` override - Transpilation test for BigQuery → DuckDB
geooo109
reviewed
Jul 17, 2026
Replace COALESCE(ARRAY_CONTAINS(arr, x), FALSE) with a CASE expression that preserves BigQuery's three-valued NULL semantics: - NULL IN UNNEST([1, 2]) → NULL (not FALSE) - 3 IN UNNEST([1, NULL]) → NULL (not FALSE) - 1 IN UNNEST(NULL) → FALSE - 1 IN UNNEST([]) → FALSE Use expression API instead of f-string interpolation per review feedback. Add NOT IN UNNEST test case. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Collaborator
|
@sirockin is this still a draft? What's pending? |
geooo109
marked this pull request as ready for review
July 21, 2026 11:02
geooo109
reviewed
Jul 21, 2026
|
|
||
| def in_sql(self, expression: exp.In) -> str: | ||
| unnest = expression.args.get("unnest") | ||
| if unnest: |
Collaborator
There was a problem hiding this comment.
This case only exists for bigquery (IN/NOT IN and an UNNEST) ? Can you check if it can occur for other dialects? @sirockin
edit: I asked this ^ question because we need to isolate the execution path whenever the AST originates from BigQuery, in case other dialects use similar syntax with different semantics
Comment on lines
+3140
to
+3183
| # BigQuery's `x IN UNNEST(arr)` → DuckDB CASE expression that preserves | ||
| # NULL semantics. The default `IN (SELECT UNNEST(...))` creates a correlated | ||
| # subquery that DuckDB rejects inside non-inner joins. | ||
| # | ||
| # NULL semantics (matching BigQuery): | ||
| # NULL IN UNNEST([1, 2]) → NULL (not FALSE) | ||
| # 3 IN UNNEST([1, NULL]) → NULL (not FALSE) | ||
| # 3 IN UNNEST([1, 2]) → FALSE | ||
| # 2 IN UNNEST([1, 2]) → TRUE | ||
| # 1 IN UNNEST(NULL) → FALSE | ||
| # 1 IN UNNEST([]) → FALSE | ||
| # | ||
| # Generated SQL: | ||
| # CASE | ||
| # WHEN arr IS NULL OR ARRAY_LENGTH(arr) = 0 THEN FALSE | ||
| # WHEN ARRAY_CONTAINS(arr, x) THEN TRUE | ||
| # WHEN x IS NULL OR ARRAY_LENGTH(arr) <> LIST_COUNT(arr) THEN NULL | ||
| # ELSE FALSE | ||
| # END | ||
| arr = unnest.expressions[0] | ||
| value = expression.this | ||
|
|
||
| arr_is_null = exp.Is(this=arr.copy(), expression=exp.null()) | ||
| arr_is_empty = exp.EQ( | ||
| this=exp.ArraySize(this=arr.copy()), | ||
| expression=exp.Literal.number(0), | ||
| ) | ||
| contains = exp.ArrayContains(this=arr.copy(), expression=value.copy()) | ||
| value_is_null = exp.Is(this=value.copy(), expression=exp.null()) | ||
| arr_has_nulls = exp.NEQ( | ||
| this=exp.ArraySize(this=arr.copy()), | ||
| expression=exp.func("LIST_COUNT", arr.copy()), | ||
| ) | ||
|
|
||
| case_expr = ( | ||
| exp.Case() | ||
| .when(exp.or_(arr_is_null, arr_is_empty, copy=False), exp.false(), copy=False) | ||
| .when(contains, exp.true(), copy=False) | ||
| .when(exp.or_(value_is_null, arr_has_nulls, copy=False), exp.null(), copy=False) | ||
| .else_(exp.false(), copy=False) | ||
| ) | ||
|
|
||
| return self.sql(case_expr) | ||
| return super().in_sql(expression) |
Collaborator
There was a problem hiding this comment.
In duckdb generator, we have a set of template variables to build these complex expressions.
These templates are in the form of *_TEMPLATE: exp.Expr = exp.maybe_parse(...). Let's leverage this in order to build this. Thus, creating a new one.
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
3 participants
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.
Summary
BigQuery uses
value IN UNNEST(array_expr)to test array membership. DuckDB has noUNNESTin this position but providesARRAY_CONTAINS.Changes
IN UNNEST(arr)→COALESCE(ARRAY_CONTAINS(arr, val), FALSE)in DuckDB generator viain_sqloverrideCOALESCE(..., FALSE)for null safety —ARRAY_CONTAINS(NULL, x)returns NULL in DuckDB, but BigQuery'sIN UNNEST(NULL)returns FALSEMotivation
We use sqlglot to transpile BigQuery SQL models to DuckDB for local testing via sqlmesh.
IN UNNEST(...)is used in several of our production models for array membership checks but was not previously transpiled to DuckDB.