Fix single-node labeled pattern expressions not filtering by label (#2443)

[gregfelice]

Problem

A single-node labeled pattern used as a boolean expression — e.g. WHERE (a:Person) or WHERE EXISTS((a:Person)) — is accepted but does not test the bound vertex's label. It evaluates as trivially true, so the predicate matches every row.

Repro (on master):

SELECT * FROM cypher('g', $$ CREATE (:Person {name:'Alice'}), (:Animal {name:'Rex'}) $$) AS (r agtype);

-- returns BOTH Alice and Rex; should return only Alice
SELECT * FROM cypher('g', $$ MATCH (a) WHERE (a:Person) RETURN a.name $$) AS (n agtype);

Root cause

A single-node pattern expression desugars to an EXISTS sub-pattern. make_path_join_quals() returns early for vertex-only patterns (list_length(entities) < 3), emitting no quals. With no edge to carry a correlation, the sub-pattern references nothing from the enclosing query, so the planner produces an uncorrelated one-time InitPlan that is true whenever any vertex of that label exists — independent of the outer row.

Relationship patterns are unaffected: the edge-driven join (start_id = a.id) correlates them to the outer variable, which is why WHERE (a)-[:R]->(b) works correctly.

EXPLAIN before the fix — uncorrelated InitPlan with a One-Time Filter:

InitPlan 1
  ->  Result
One-Time Filter: ((InitPlan 1).col1)::agtype

After the fix — correlated SubPlan referencing the outer a_1:

SubPlan 1
  ->  Result
        One-Time Filter: ((_extract_label_id(a_1.id))::integer = 3)

Fix

In make_path_join_quals(), for a vertex-only pattern whose vertex carries a non-default label and whose variable is declared in an ancestor parse state (a correlated reference), emit an explicit label-id filter. make_qual() builds a name-based id reference that resolves to the outer variable, so the filter both correlates the sub-pattern to that variable and enforces the label.

Freshly scanned, non-correlated vertices (no ancestor binding) are untouched, so MATCH (a:Person) and "does any X exist" EXISTS checks behave exactly as before.

Testing

Added regression coverage to pattern_expression against a graph containing a non-Person vertex:

• WHERE (a:Person) → only the :Person vertices
• WHERE NOT (a:Person) → only the non-Person vertex
• WHERE EXISTS((a:Company)) → only the :Company vertex

All 41 regression tests pass (make installcheck).

Note

RETURN (a:Label) in a projection on an unlabeled bound variable still errors with "multiple labels for variable" — that is a separate, pre-existing guard, orthogonal to this fix, and is intentionally left unchanged.

Fixes #2443

[gregfelice]

Ready for review.

• Fix verified in a clean PG18 build: WHERE (a:Person) / EXISTS((a:Person)) now filter correctly (correlated SubPlan on _extract_label_id(a.id) = <label_id>) instead of producing a trivially-true uncorrelated one-time InitPlan.
• Full regression suite green (41/41), regression.diffs empty, including the new pattern_expression cases.
• Scope kept tight: only correlated bound vertices with a non-default label are affected; plain MATCH (a:Person) and "does any X exist" EXISTS checks are unchanged.
• The pre-existing single-node-pattern NOTE in pattern_expression.sql was updated to reflect the corrected behavior (and the orthogonal, still-present "multiple labels" projection limitation).

Fixes #2443.