feat(gql): implement standalone INSERT statement

[qishipengqsp]

Summary

Implements GQL INSERT (standalone, no MATCH context, system-allocated vertex/edge ids) end-to-end across parser → planner → executor, plus a sqllogictest suite and a test-fixture procedure.

• Parser: replaces the empty LinearDataModifyingStatement stub with a real DataModifyingStatement::Insert enum; INSERT paths reuse the existing ElementPattern AST so labels, properties, and variables share code with MATCH.
• Planner: full bind → plan → optimizer-passthrough chain for BoundInsertStatement (resolved vertices with property literals, edges referenced by endpoint index). Binder reports focused errors for unknown labels/properties, type mismatches, undefined endpoints, duplicate variables, and edge-endpoint type mismatches.
• Storage: adds AtomicU64 vid/eid allocators on MemoryGraph. Counters are observed past externally supplied ids inside create_vertex/create_edge, WAL replay, and checkpoint restore so allocator output never collides with persisted state.
• Execution: InsertBuilder follows the existing catalog_modify generator pattern — open a Serializable txn, allocate ids, insert vertices then edges, commit.
• Tests: new minigu-test/gql/dml/insert.slt with 7 happy paths, 8 error paths, and 4 MATCH-based verifications. Adds the create_insert_test_graph fixture procedure (PERSON/COMPANY/KNOWS/WORKS_AT, primitive types only — GQL has no vector literal syntax, so create_test_graph_data with its required vector embeddings can't be used for INSERT tests).

Supported syntax

INSERT (a:Person {name: 'Alice', age: 30})
INSERT (a:Person {name: 'A'}), (b:Company {name: 'B'})
INSERT (a:Person {name: 'A'})-[:KNOWS {since: 2020}]->(b:Person {name: 'B'})

All vertex variables used by edges in the same INSERT must be defined inline; MATCH-context bindings, anonymous vertices, undirected edges, quantifiers, and SET/REMOVE/DELETE are explicitly out of scope for this PR.

Side effect on existing snapshots

8 insta snapshots in ddl/, dml/, misc/ are updated. INSERT statements in those test programs previously failed at parse time (Parser(Unexpected)); they now parse cleanly and fall through to bind-time errors like CurrentGraphNotSpecified because those programs don't issue SESSION SET GRAPH. New snapshots reflect the new (more accurate) error path.

Test plan

• cargo test --workspace — all green (one pre-existing flaky test in minigu-storage is unrelated)
• cargo test -p gql-parser --features serde insert — 4 new parser snapshot tests pass
• cargo test -p minigu-test --test sqllogictest — new dml/insert.slt passes (all 7 statement ok, all 8 statement error, all 4 MATCH verifications)
• Verified the parser-only failures previously masked by parse-fail now reach the binder with the correct diagnostic

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR adds end-to-end support for standalone GQL INSERT statements (no MATCH context) across the parser → binder/planner → executor pipeline, and introduces storage-side ID allocation so inserts can use system-assigned vertex/edge IDs without colliding with recovered state.

Changes:

• Parser: implements a real DataModifyingStatement::Insert AST + parsing for INSERT <path>(, <path>)*, reusing ElementPattern for nodes/edges.
• Planner/Executor: binds INSERT into a physical plan node and executes it by allocating IDs, inserting vertices first then edges in a serializable transaction.
• Storage/Tests: adds monotone AtomicU64 ID allocators to MemoryGraph (observed during WAL replay/checkpoint restore) and adds a sqllogictest suite + fixture procedure for INSERT.

Reviewed changes

Copilot reviewed 34 out of 35 changed files in this pull request and generated 5 comments.

Show a summary per file

File	Description
minigu/storage/src/tp/memory_graph.rs	Adds atomic ID allocators (next_vertex_id/next_edge_id) + observe helpers to keep allocator monotone across recovery and explicit IDs.
minigu/storage/src/tp/checkpoint.rs	Observes restored vertex/edge IDs during checkpoint restore to advance allocators past persisted maxima.
minigu/gql/planner/src/plan/mod.rs	Registers a new physical plan node variant for INSERT.
minigu/gql/planner/src/plan/data_modify.rs	Introduces Insert physical plan node and explain output.
minigu/gql/planner/src/optimizer/mod.rs	Passes through PhysicalInsert during physical plan creation.
minigu/gql/planner/src/logical_planner/procedure_spec.rs	Routes bound data-modifying statements into the data-modify planner path.
minigu/gql/planner/src/logical_planner/mod.rs	Adds the data logical planner module.
minigu/gql/planner/src/logical_planner/data.rs	Plans BoundDataModifyingStatement::Insert into PlanNode::PhysicalInsert.
minigu/gql/planner/src/bound/procedure_spec.rs	Adds BoundStatement::Data(Vec<...>) to carry bound data-modifying statements.
minigu/gql/planner/src/bound/mod.rs	Exposes new bound data-modify structures.
minigu/gql/planner/src/bound/data.rs	Defines BoundInsertStatement (vertices + edges referencing endpoints by vertex index).
minigu/gql/planner/src/binder/procedure_spec.rs	Enables binding of Statement::Data into BoundStatement::Data.
minigu/gql/planner/src/binder/mod.rs	Adds the binder data module.
minigu/gql/planner/src/binder/error.rs	Adds INSERT-specific bind errors (unknown labels/properties, endpoint/type mismatch, etc.).
minigu/gql/planner/src/binder/data.rs	Implements binding for standalone INSERT, including schema/type checking and literal coercions.
minigu/gql/parser/src/parser/impls/data.rs	Implements parsing for linear data-modifying statements and INSERT paths; adds parser snapshot tests.
minigu/gql/parser/src/ast/data.rs	Replaces the stub linear DML AST with DataModifyingStatement::Insert + InsertStatement/InsertPath.
minigu/gql/parser/src/parser/impls/snapshots/gql_parser__parser__impls__data__tests__insert_single_vertex.snap	New parser snapshot for single-vertex INSERT.
minigu/gql/parser/src/parser/impls/snapshots/gql_parser__parser__impls__data__tests__insert_multiple_vertices.snap	New parser snapshot for multi-vertex INSERT.
minigu/gql/parser/src/parser/impls/snapshots/gql_parser__parser__impls__data__tests__insert_vertex_edge_vertex.snap	New parser snapshot for vertex-edge-vertex INSERT.
minigu/gql/parser/src/parser/impls/snapshots/gql_parser__parser__impls__data__tests__insert_reverse_edge.snap	New parser snapshot for reverse-direction edge INSERT.
minigu/gql/execution/src/executor/mod.rs	Registers the data_modify executor module.
minigu/gql/execution/src/executor/data_modify.rs	Adds INSERT executor that allocates IDs and writes vertices then edges in a serializable txn.
minigu/gql/execution/src/builder.rs	Wires PlanNode::PhysicalInsert into executor building.
minigu/core/src/procedures/mod.rs	Registers a new built-in procedure for INSERT testing.
minigu/core/src/procedures/create_insert_test_graph.rs	Adds create_insert_test_graph fixture procedure to set up a primitive-only schema for INSERT tests.
minigu-test/gql/dml/insert.slt	Adds sqllogictest coverage for INSERT happy/error paths + MATCH verification queries.
minigu-test/gql/misc/vector_index@parser.snap	Updates parser snapshot: INSERT now parses (previously failed at parse time).
minigu-test/gql/misc/vector_index@e2e.snap	Updates e2e snapshot: INSERT now reaches bind-time errors when no graph is set.
minigu-test/gql/dml/dml_dql@e2e.snap	Updates e2e snapshot: INSERT now reaches bind-time errors (no current graph) rather than parse errors.
minigu-test/gql/ddl/ddl_truncate@parser.snap	Updates parser snapshot: INSERT now parses in these programs.
minigu-test/gql/ddl/ddl_truncate@e2e.snap	Updates e2e snapshot: INSERT now fails at bind-time (no current graph) rather than parse-time.
minigu-test/gql/ddl/ddl_drop@parser.snap	Updates parser snapshot: INSERT now parses in these programs.
minigu-test/gql/ddl/ddl_drop@e2e.snap	Updates e2e snapshot: INSERT now fails at bind-time (no current graph) rather than parse-time.

---

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