[BUG] Schema probe on a single column returns all columns

[LepsyMikolaj3301]

Hello,
following a recent fix of schema caching (Spark/pyspark) using the WHERE 1=0 clause in PR #4118, the RowDescription returns all columns, rather than only one, queried column.

Example (Python 3.12.7 using psycopg==3.3.3) querying lesmiserables dataset:

1. full select

>>> SELECT * FROM Character WHERE 1=0

Description returns an array:

[<Column 'name', type: varchar (oid: 1043)>, <Column '@rid', type: varchar (oid: 1043)>, <Column '@type', type: varchar (oid: 1043)>, <Column '@cat', type: "char" (oid: 18)>]

2. 'name' Column only

>>> SELECT name FROM Character WHERE 1=0

Description returns an array:

[<Column 'name', type: varchar (oid: 1043)>, <Column '@rid', type: varchar (oid: 1043)>, <Column '@type', type: varchar (oid: 1043)>, <Column '@cat', type: "char" (oid: 18)>]

Expected: [<Column 'name', type: varchar (oid: 1043)>]

[lvca]

Confirmed, thanks for the report.

When a query returns no rows (like the WHERE 1=0 probe used by Spark/PySpark and JDBC drivers), ArcadeDB has no data to describe, so it falls back to reading the type definition to build the column metadata. That fallback looked only at the type, never at the list of columns you actually selected, so it always returned every property plus the system columns.

Fixed in main: the column metadata is now built from the projection of the parsed query, so:

• SELECT name FROM Character WHERE 1=0 -> name only
• SELECT name AS hero FROM ... -> hero
• SELECT * FROM ... and SELECT FROM ... -> all columns, as before

A regression test covering the empty and the populated case has been added.

---

❤️ Thanks for helping make ArcadeDB better! ArcadeDB is free and open source, sustained by its community. If it's useful to you or your company, please consider becoming a sponsor to keep development going.