Implement predicate functions: all(), any(), none(), single()

[gregfelice]

Summary

Implements the four openCypher predicate functions (issues #552, #553, #555, #556):

• all(x IN list WHERE predicate) — true if all elements match
• any(x IN list WHERE predicate) — true if at least one matches
• none(x IN list WHERE predicate) — true if no elements match
• single(x IN list WHERE predicate) — true if exactly one matches

These are among the most requested Cypher features for AGE and are critical for users migrating from Neo4j and Kuzu (recently archived).

Implementation

Approach: Builds on the existing list comprehension infrastructure (unnest-based subqueries with child parsestates for variable scoping).

SQL transformation strategy: Each predicate function is lowered to an aggregate subquery over unnest(list) that preserves Cypher's three-valued (TRUE/FALSE/NULL) semantics:

• all(), any(), none() → SELECT CASE WHEN bool_or(<first-branch>) THEN <result1> WHEN bool_or(pred IS NULL) THEN NULL ELSE <result2> END FROM unnest(list) AS x — two bool_or() aggregates compute (1) whether any element satisfies the decisive condition and (2) whether any element's predicate is NULL, combined via CASE to return the correct three-valued result. On an empty list both aggregates return NULL, so the ELSE branch yields the vacuous-truth defaults (all()/none() → true, any() → false).
• single() → SELECT count(*) FROM unnest(list) AS x WHERE pred IS TRUE) = 1 — exact count of truthy matches. (LIMIT 2 short-circuit optimization deferred; current form evaluates the full list.)
• All four are wrapped in a CASE WHEN <list-expr> IS NULL THEN NULL ELSE <subquery> END guard so a NULL list expression propagates NULL rather than collapsing to the empty-list defaults.

This aggregate-based shape (rather than EXISTS/NOT EXISTS) was chosen specifically to preserve correct NULL handling: WHERE in a SQL subquery drops rows where the predicate is NULL, which would incorrectly collapse any(x IN [NULL] WHERE x > 0) from NULL to false under an EXISTS shape. The aggregate form keeps NULL-producing rows visible and folds them into a proper three-valued result.

Files changed (12):

File	Change
cypher_nodes.h	New cypher_predicate_function node type with CPFK_ALL/ANY/NONE/SINGLE enum
ag_nodes.h / ag_nodes.c	Node registration
cypher_outfuncs.h / cypher_outfuncs.c	Serialization
cypher_kwlist.h	Three new keywords: ANY_P, NONE, SINGLE
cypher_gram.y	Grammar rules in expr_func_subexpr, build_predicate_function_node() helper with NULL-list guard, extract_iter_variable_name() shared helper (rejects qualified ColumnRefs), keywords added to safe_keywords
cypher_clause.c	transform_cypher_predicate_function() — builds query tree with bool_or() aggregates + CASE for three-valued semantics; make_bool_or_agg() helper
cypher_analyze.c	Expression walker for new node type
Makefile	Register new regression test
regress/sql/predicate_functions.sql	New test file
regress/expected/predicate_functions.out	Expected output

Backward compatibility:

• ANY_P, NONE, SINGLE added to safe_keywords so they work as property keys and label names (e.g., {any: 1, none: 2, single: 3})
• ALL was already a reserved keyword with safe_keywords entry
• No grammar conflicts (verified: zero new shift/reduce or reduce/reduce warnings)

Regression Tests

Test queries covering:

• Basic true/false for each function
• Empty list edge cases (vacuous truth for all()/none(), false for any()/single())
• NULL list input (all four return NULL via the guard)
• NULL predicate results within non-empty lists (three-valued semantics)
• Graph data integration (MATCH (u) WHERE all(x IN u.vals WHERE ...))
• Boolean combinations (any(...) AND all(...))
• Nested predicates (any(x IN ... WHERE all(y IN ... WHERE ...)))
• Keyword backward compatibility ({any: 1, none: 2, single: 3})
• Deterministic ordering (all graph queries use ORDER BY)

All regression tests pass.

[Copilot]

Pull request overview

Adds support for the openCypher predicate functions all(), any(), none(), and single() by introducing a new AST node and transforming it into unnest(...)-based SQL subqueries during parsing/analyzing, with a new regression test suite.

Changes:

• Add new cypher_predicate_function AST node (+ enum kind) and register/serialize it as an ExtensibleNode.
• Extend the Cypher grammar with all/any/none/single(variable IN list WHERE predicate) and add keywords to the lexer + safe_keywords.
• Implement query-tree transformation for predicate functions and add regression tests (predicate_functions).

Reviewed changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
src/include/parser/cypher_kwlist.h	Adds any/none/single as Cypher keywords.
src/backend/parser/cypher_gram.y	Adds grammar rules + helper to build predicate-function nodes and wraps them in SubLinks.
src/include/nodes/cypher_nodes.h	Introduces cypher_predicate_function node + kind enum.
src/include/nodes/ag_nodes.h	Registers new node tag for predicate functions.
src/backend/nodes/ag_nodes.c	Adds node name and ExtensibleNode methods entry for predicate-function node.
src/include/nodes/cypher_outfuncs.h	Declares serialization function for the new node.
src/backend/nodes/cypher_outfuncs.c	Implements serialization for cypher_predicate_function.
src/backend/parser/cypher_clause.c	Transforms predicate-function node into unnest-based subqueries (EXISTS / count).
src/backend/parser/cypher_analyze.c	Adds expression walker support for the new node type.
Makefile	Registers the new predicate_functions regression test.
regress/sql/predicate_functions.sql	Adds regression SQL coverage for predicate functions.
regress/expected/predicate_functions.out	Adds expected output for the new regression test.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[jrgemignani]

@gregfelice Please see the above comments by Copilot

[gregfelice]

Addressed all 4 Copilot suggestions:

1. NULL semantics — Rewrote from EXISTS_SUBLINK to EXPR_SUBLINK with aggregate-based CASE expressions (bool_or(pred IS TRUE/FALSE) + bool_or(pred IS NULL)) to preserve three-valued Cypher NULL semantics. any(x IN [NULL] WHERE x > 0) now correctly returns NULL instead of false.

2. Unqualified iterator check — Added list_length(cref->fields) != 1 validation in extract_iter_variable_name(). Qualified names like x.y now error with "qualified name not allowed as iterator variable".

3. NULL regression tests — Added 13 new test cases covering: NULL list input for all four functions, null elements in lists, literal null predicates, and mixed null/non-null elements.

4. single() performance — Applied IS TRUE filtering so NULL predicates aren't counted. The LIMIT 2 optimization breaks correlated variable references in graph property contexts (e.g., MATCH (u) WHERE single(x IN u.vals WHERE ...)), documented with TODO for future optimization pass.

Bonus: Added copy/read support for cypher_predicate_function ExtensibleNode to prevent "unexpected copyObject()" crashes when PostgreSQL's query rewriter copies expression trees.

All 32 regression tests pass (predicate_functions: ok).

[jrgemignani]

@gregfelice We'll see what Copilot thinks ;) Btw, in the future, can you put your comments in the reply to Copilot, please.

[Copilot]

Pull request overview

Copilot reviewed 16 out of 16 changed files in this pull request and generated 3 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Pull request overview

Copilot reviewed 16 out of 16 changed files in this pull request and generated 2 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[jrgemignani]

@gregfelice Please see Copilot above

[gregfelice]

@jrgemignani — all Copilot items addressed and threads resolved. Round 3 commit (507a2e5) refactors build_list_comprehension_node() to reuse extract_iter_variable_name(), so iterator-variable validation is consistent between list comprehensions and predicate functions. The PR description already describes the bool_or() aggregate approach correctly. Ready for re-review when you have a moment. Thanks!