Description

BABEL-6037: Cross-session ANTLR parse tree caching for T-SQL stored procedures.

Stored procedures with thousands of lines (e.g., ~1300 lines) take excessive time on first execution in each new session due to redundant ANTLR parsing. The PLtsql function hash table is session-scoped, so every new session re-parses from scratch.

This PR introduces persistent ANTLR parse tree caching. Serialized parse trees are stored in new column in sys.babelfish_function_ext using PostgreSQL's nodeToString()/stringToNode() framework. On first execution in a new session, cached results are deserialized to skip ANTLR re-parsing, with version and modify-date validation to prevent serving stale data.

Issues Resolved

BABEL-6037

Changes

Serialization Infrastructure

• gen_pltsql_node_support.pl code generator (modeled after PG's gen_node_support.pl) produces pltsql_nodetags.h, pltsql_outfuncs_defs.c, and pltsql_readfuncs_defs.c from annotated header files (pltsql_serializable_1.h, pltsql_serializable_2.h which are copies of pltsql.h, pltsql-2.h)
• Extension-owned T_PLtsql_* NodeTag values offset from 1000 to avoid collision with engine NodeTag enum, with an ABI stability check that fails the build if node types are added without updating the last nodetag constants
• Wrapper files pltsql_outfuncs.c and pltsql_readfuncs.c expose pltsql_outNode() and pltsql_parseNodeString() dispatch functions
• pltsql_serialize_macros.h replicates engine-internal WRITE_*/READ_* macros
• pltsql_node_stubs.c provides custom read/write handlers for nodes requiring special serialization logic (flexible array members, runtime-only fields, string/int arrays)
  • Redefines necessary WRITE_*/READ_* macros locally (not exposed in any PG header) (from postgresql_modified_for_babelfish/src/backend/nodes/readfuncs.c and outfuncs.c)
  • Custom read/write implementations for PLtsql nodes marked custom_read_write:
    • PLtsql_expr — skips runtime-only fields (plan, func, expr_simple_*)
    • PLtsql_nsitem — handles FLEXIBLE_ARRAY_MEMBER for name[]
    • PLtsql_row — handles string/int arrays (fieldnames, varnos) with array_size(nfields)
    • PLtsql_recfield — skips runtime cache fields (rectupledescid, finfo)
• pl_handler.c registers outNode_hook and parseNodeString_hook in _PG_init() so the engine delegates to extension code for PLtsql node types
• Upon encountering serialization errors, the code gracefully falls back to full ANTLR recompile.

Catalog Changes

Five new columns in sys.babelfish_function_ext:

• antlr_parse_tree_text TEXT — serialized parse tree (nodeToString output)
• antlr_parse_tree_datums TEXT — serialized datum array
• antlr_parse_tree_modify_date DATETIME — timestamp for staleness detection
• antlr_parse_tree_bbf_version TEXT — Babelfish version at serialization time
• antlr_cache_enabled BOOL — per-function cache flag (default false)

Upgrade SQL in babelfishpg_tsql--5.5.0--5.6.0.sql adds these columns with allow_system_table_mods guard for CI compatibility.

Cache Lifecycle

• CREATE/ALTER PROCEDURE: Compiles the procedure, serializes the parse tree to babelfish_function_ext, and populates the in-session PLtsql hash table
• 1st EXEC in new session: Hash table miss → deserializes from babelfish_function_ext, validates bbf_version and modify_date, populates hash table
• 2nd+ EXEC in same session: Hash table hit (no deserialization needed)
• ALTER with GUC disabled: Sets parse tree columns to NULL to force reparsing
• DROP PROCEDURE: Removes the catalog row and per-function flag
• MVU (Major Version Upgrade): Version mismatch rejects cache entries and forces ANTLR parsing

GUC Configuration

• babelfishpg_tsql.enable_routine_parse_cache (session-level, PGC_USERSET, default false) — global toggle for enabling/disabling cache reads and writes
• sys.enable_routine_parse_cache(func_identifier TEXT, enable_flag BOOLEAN) — per-function granularity; accepts schema.func, schema.func(argtypes), or func (defaults to dbo); disabling NULLs out cache columns for immediate invalidation

Build System

• Makefile updated to compile serialization wrapper .o files with dependency rules for generated files from gen_pltsql_node_support.pl
• Generated files added to .gitignore

Node Allocation

• All PLtsql node palloc0() calls replaced with makeNode() to set proper NodeTag values required by the serialization framework

Performance Results

• Customer procedure (~1300 lines): 2031ms → 15ms first-execution time (99% reduction)
• Stress test (1000 connections): Average first-execution time ~4314ms → ~175ms

Test Scenarios Covered

[TBD]

• Use case based -

• Boundary conditions -

• Arbitrary inputs -

• Negative test cases -

• Minor version upgrade tests -

• Major version upgrade tests -

• Performance tests -

• Tooling impact -

• Client tests -

Check List

• Commits are signed per the DCO using --signoff

By submitting this pull request, I confirm that my contribution is under the terms of the Apache 2.0 and PostgreSQL licenses, and grant any person obtaining a copy of the contribution permission to relicense all or a portion of my contribution to the PostgreSQL License solely to contribute all or a portion of my contribution to the PostgreSQL open source project.

For more information on following Developer Certificate of Origin and signing off your commits, please check here.