pg-inventory-synthesizer

name	pg-inventory-synthesizer
description	Synthesizes the outputs of pg-bnf-source-mapper, pg-write-trace-skill, and pg-read-trace-skill into a master PG_SEMANTIC_INVENTORY.json. It maps statements to regression tests (pg_regress) to provide a complete coverage plan.

Deprecated: use pg/semantic/SKILL.md and the staged pipeline under pg/semantic/ instead.

PostgreSQL Inventory Synthesizer

You are the project manager and architect of the PostgreSQL Semantic Fidelity Pipeline. Your task is to synthesize all audit results into a master inventory.

This is the final Step (Step 4) of building the PostgreSQL Semantic Inventory.

Your Task

1. Collect the JSON outputs from:
   • pg-bnf-source-mapper (Syntax to Source)
   • pg-write-trace-skill (Side-effects)
   • pg-read-trace-skill (Dependencies)
2. Map each statement to its relevant pg_regress test files (search for SQL statements in src/test/regress/sql/).
3. Compile a master PG_SEMANTIC_INVENTORY.json that represents the "Ground Truth" for full semantic coverage.
4. Update the master PROGRESS.json for the pg/catalog implementation.

Strategy

• Group scenarios by logical statement type (e.g., CREATE TABLE, ALTER TABLE).
• For each statement, list all identified "Semantic Scenarios" (basic, inheritance, partitioned, etc.).
• Ensure every BNF statement from the parser is represented in this inventory.

Expected Master JSON Structure

{
  "statement": "CREATE TABLE",
  "bnf_slug": "create-table-stmt",
  "pg_source": "src/backend/commands/tablecmds.c:DefineRelation",
  "semantic_batches": [
    {
      "scenario": "basic_heap_table",
      "affects": ["pg_class", "pg_attribute", "pg_type", "pg_namespace"],
      "reads": ["pg_namespace", "pg_authid"],
      "regress_files": ["create_table.sql"]
    },
    {
      "scenario": "inherited_table",
      "affects": ["pg_class", "pg_attribute", "pg_inherits", "pg_depend"],
      "reads": ["pg_class", "pg_attribute"],
      "regress_files": ["inherit.sql"]
    }
  ]
}

Guidelines

• This is the final blueprint for "Semantic Completeness".
• Ensure that the regress_files field contains enough tests to trigger all identified affects and reads.
• If a statement has zero regression tests (unlikely), mark it for "Manual Semantic Review".