Add jsonb[] array type pushdown support#286
Draft
sfc-gh-dachristensen wants to merge 6 commits into
Draft
Conversation
Remove the JSONB[] exclusion in GetDuckDBTypeForPGType() that returned DUCKDB_TYPE_INVALID for arrays with JSONB elements, preventing pushdown. JSONB[] now maps to DUCKDB_TYPE_LIST like all other array types. The existing infrastructure already handles this correctly: - ArrayOutForPGDuck serializes jsonb elements via jsonb_out - pgduck_server's list_to_text handles JSON (VARCHAR) child elements - The jsonb() macro in DuckDB maps jsonb casts to JSON Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Signed-off-by: David Christensen <david.christensen@snowflake.com>
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Signed-off-by: David Christensen <david.christensen@snowflake.com>
DuckDB does not have a jsonb type, only json. When the query rewriter encounters JSONBARRAYOID in expression nodes (Const, CoerceViaIO, RelabelType, Param, ArrayExpr), rewrite to JSONARRAYOID so the deparsed query emits "json[]" instead of "jsonb[]". Also handle JSONBARRAYOID -> JSONARRAYOID in insert_select.c for null column construction, matching the existing JSONBOID -> JSONOID handling. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Signed-off-by: David Christensen <david.christensen@snowflake.com>
Revert the JSONBARRAYOID→JSONARRAYOID OID rewriting in rewrite_query.c which broke operator resolution (operators resolved for jsonb[] args became inconsistent after type OIDs were changed to json[], causing DuckDB to receive unparseable OPERATOR(pg_catalog.=) syntax). Instead, map jsonb/jsonb[] to json/json[] in deparse_type_name() which only affects the SQL text sent to DuckDB without disturbing the expression tree's operator and function OIDs. Also revert the test assertion for col::jsonb[] cast back to expecting no pushdown, since CoerceViaIO casts are not shippable. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Signed-off-by: David Christensen <david.christensen@snowflake.com>
Scalar jsonb is already handled by rewrite_query.c which wraps jsonb expressions in __lake__internal__nsp__.jsonb() function calls. Mapping JSONBOID to JSONOID in deparse_type_name caused DuckDB to receive ::json casts that broke operator resolution (json = json doesn't exist in DuckDB). Only JSONBARRAYOID needs remapping since there's no equivalent array rewrite mechanism. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Signed-off-by: David Christensen <david.christensen@snowflake.com>
| * DuckDB does not have a jsonb[] type. Map it to json[] which DuckDB | ||
| * understands. We do this at deparse time rather than rewriting OIDs | ||
| * in the expression tree, because changing OIDs would break operator | ||
| * and function lookup. |
Collaborator
There was a problem hiding this comment.
This isn't the real deparse though, we still parse the output of this (foreignscan case). Not sure I understand the effect of this.
Collaborator
Author
There was a problem hiding this comment.
I'm just giving Claude a chance to bang its head against this for a bit, so it could definitely be a bit offbase here. (I guess I'll make this a draft so you don't waste your time on it until things are happy.)
The full query pushdown path uses PostgreSQL's pg_get_querydef (ruleutils) for deparsing, which has no knowledge of our type mappings. When RewriteConst creates RelabelType nodes with JSONBARRAYOID, or when NULL constants retain JSONBARRAYOID type, ruleutils emits ::jsonb[] which DuckDB doesn't understand. Fix by remapping JSONBARRAYOID to JSONARRAYOID directly in the expression tree at the end of RewriteQueryTreeForPGDuckMutator. Only value/output nodes (Const, RelabelType, Param, ArrayExpr) are changed to avoid breaking operator resolution which depends on type OIDs. This runs after RewriteConst/RewriteParam so their output nodes are also caught. Combined with the deparse_type_name fix in deparse.c, both deparse paths now emit json[] instead of jsonb[]. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Signed-off-by: David Christensen <david.christensen@snowflake.com>
sfc-gh-mslot
reviewed
Apr 1, 2026
| else if (IsA(node, RelabelType) && ((RelabelType *) node)->resulttype == JSONBARRAYOID) | ||
| ((RelabelType *) node)->resulttype = JSONARRAYOID; | ||
| else if (IsA(node, Param) && ((Param *) node)->paramtype == JSONBARRAYOID) | ||
| ((Param *) node)->paramtype = JSONARRAYOID; |
Collaborator
There was a problem hiding this comment.
this looks messy. How do we map JSONB to JSON? Seems like we'd want to do the same thing here?
Iirc we create an alias in DuckDB, so do we actually need to do anything special for arrays w.r.t. deparsing/rewriting?
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
2 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.
Remove the JSONB[] exclusion in GetDuckDBTypeForPGType() that returned DUCKDB_TYPE_INVALID for arrays with JSONB elements, preventing pushdown. JSONB[] now maps to DUCKDB_TYPE_LIST like all other array types.
The existing infrastructure already handles this correctly: