fix: exclude internal VLE helper vars from RETURN * expansion (#2540)

[waterWang]

Description

Fixes #2540

MATCH p = ()-[*1]->(n:End {id: 2})
RETURN *

fails with ERROR: return row and column definition list do not match while the equivalent explicit RETURN p, n succeeds.

Root cause

When a variable-length edge pattern has no explicit start/end variable, build_VLE_relation in cypher_gram.y assigns internal helper names via create_unique_name(AGE_DEFAULT_PREFIX"vle_function_{start,end}_var") (e.g. _age_default_vle_function_start_var_1). Those columns land in the parse namespace and are picked up by RETURN * expansion (ExpandAllTables -> expand_pnsi_attrs). The SRF is then invoked with one extra column, producing the row/column mismatch error.

Fix

In expand_pnsi_attrs, skip columns whose name begins with AGE_DEFAULT_PREFIX "vle_function_", the same way the existing hidden-var (AGE_DEFAULT_VARNAME_PREFIX) and hidden-alias (AGE_DEFAULT_ALIAS_PREFIX) filters work.

Verification

• make installcheck: 41/43 pass (only pre-existing age_load/file-path and age_upgrade env failures, unrelated to this change)
• Manual repro now returns the expected (n, p) columns matching the AGE convention that RETURN * orders entities before the path variable

[Copilot]

Pull request overview

This PR fixes a RETURN * expansion bug in Apache AGE Cypher parsing where internal variable-length edge (VLE) helper variables leaked into the parse namespace expansion, causing ERROR: return row and column definition list do not match for patterns like MATCH p = ()-[*1]->(n:End {id: 2}) RETURN *.

Changes:

• Extend expand_pnsi_attrs()’s hidden-column filtering to also exclude internal VLE helper columns whose names begin with AGE_DEFAULT_PREFIX "vle_function_".
• Add explanatory in-code documentation clarifying why these VLE helper columns must not be included in RETURN *.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[jrgemignani]

@waterWang Two things to point out, please address the code change, which is minor. Also, as this is fixing an issue, please provide regression tests showing that it resolves that issue.