1) Concept framing (short)

• What problem it solves
  • Lets you define CRUD screens/forms (fields, labels, types), validations, and permissions in tables, so behavior can be adjusted after deployment without code changes.
  • Supports “FoxPro/4GL ergonomics”: tables → forms/behavior (and later reports) with metadata as the source of truth.
• What changes post-deploy (and who can change it)
  • Admin/config editors can change:
    • Which fields appear on a form, their order, labels, help text
    • Validation rules (required, regex, min/max, length limits)
    • Role-based permissions (view/edit/create/delete)
    • Overrides per environment (dev/test/prod) and/or tenant
  • Changes take effect immediately (with caching + hot-reload strategy below).
• What must remain code (safety boundaries)
  • DB access and parameterization (no dynamic SQL from config)
  • Allow-lists for:
    • Validation operators (required, regex, min, max, min_len, max_len)
    • Field types (text, int, decimal, date, bool)
  • Any “execute action” / “call method” behavior must be explicitly mapped in code (config may only select from an allow-list).

2) Canonical data model (portable)

PostgreSQL DDL

-- Schema versioning (simple + effective)
create table if not exists tdd_schema_version (
  id            bigserial primary key,
  version       int not null unique,
  applied_at    timestamptz not null default now()
);

-- Entities (one per screen/form)
create table if not exists tdd_entity (
  id            bigserial primary key,
  entity_key    text not null unique,   -- e.g. "customer"
  display_name  text not null,
  table_name    text not null,          -- physical table target, allow-listed by code if needed
  created_at    timestamptz not null default now(),
  updated_at    timestamptz not null default now()
);

-- Fields on a form (metadata-driven form layout)
create table if not exists tdd_field (
  id            bigserial primary key,
  entity_id     bigint not null references tdd_entity(id) on delete cascade,
  field_key     text not null,          -- e.g. "email"
  column_name   text not null,          -- e.g. "email"
  field_type    text not null,          -- allow-list: text,int,decimal,date,bool
  label         text not null,
  help_text     text null,
  is_required   boolean not null default false,
  sort_order    int not null default 0,

  -- override dimensions
  env           text null,              -- e.g. "dev","prod"
  tenant_id     text null,              -- e.g. "acme"

  -- conflict resolution
  priority      int not null default 0, -- higher wins within same specificity

  created_at    timestamptz not null default now(),
  updated_at    timestamptz not null default now(),

  unique (entity_id, field_key, env, tenant_id)
);

create index if not exists ix_tdd_field_entity on tdd_field(entity_id);
create index if not exists ix_tdd_field_lookup on tdd_field(entity_id, field_key);

-- Validations are separate rows (more flexible than columns)
create table if not exists tdd_validation (
  id            bigserial primary key,
  entity_id     bigint not null references tdd_entity(id) on delete cascade,
  field_key     text not null,
  rule_op       text not null,          -- allow-list: required,regex,min,max,min_len,max_len
  rule_arg      text null,              -- e.g. regex pattern, numeric threshold
  message       text not null,

  env           text null,
  tenant_id     text null,
  priority      int not null default 0,

  created_at    timestamptz not null default now(),
  updated_at    timestamptz not null default now()
);

create index if not exists ix_tdd_val_entity_field on tdd_validation(entity_id, field_key);

-- Permissions (role → allowed operations)
create table if not exists tdd_permission (
  id            bigserial primary key,
  entity_id     bigint not null references tdd_entity(id) on delete cascade,
  role_key      text not null,          -- e.g. "clerk","admin"
  can_view      boolean not null default true,
  can_create    boolean not null default false,
  can_edit      boolean not null default false,
  can_delete    boolean not null default false,

  env           text null,
  tenant_id     text null,
  priority      int not null default 0,

  created_at    timestamptz not null default now(),
  updated_at    timestamptz not null default now(),

  unique (entity_id, role_key, env, tenant_id)
);

create index if not exists ix_tdd_perm_entity_role on tdd_permission(entity_id, role_key);

SQLite DDL

• Same as PostgreSQL, except:
  • Use integer primary key autoincrement instead of bigserial
  • Use datetime('now') instead of now()
  • No timestamptz type (use text or integer)

3) Runtime algorithm

1. Load config
   • Cache key: (entity_key, tenant_id, env, role_key)
   • Strategy:
     • Start with lazy load on first access.
     • Cache with a TTL (e.g., 30–120s) or with a config revision.
     • Optional hot-reload: poll a tdd_schema_version or a dedicated tdd_config_revision table.
2. Conflict resolution (overrides)
   • Treat rows as layered:
     1. (tenant_id, env) most specific
     2. (tenant_id, null)
     3. (null, env)
     4. (null, null) base default
   • Within the same layer, higher priority wins.
   • For fields: choose the single “winning” row per (field_key).
   • For validations: you may keep multiple rules per field; you still resolve by specificity + priority.
3. Validate config
   • Schema validation: table constraints + allowed values (types, ops) enforced in code.
   • Semantic validation (on load):
     • Field types are in allow-list.
     • Column names are in allow-list (or verified against actual DB schema if you want).
     • Regex compiles (where applicable).
     • Numeric thresholds parse.
   • If invalid: fail “closed” (deny edit) or fall back to base config depending on your risk tolerance.
4. Failure modes
   • Missing entity/config: return “entity not configured” (safe error), no dynamic behavior.
   • Invalid overrides: ignore override layer and fall back to less-specific rows; log + alert.
   • Partial migrations: use tdd_schema_version gate; if app expects version N but DB has N-1, run in safe mode.
5. Migration strategy
   • Increment tdd_schema_version.
   • Backfill:
     • Add new columns as nullable with defaults.
     • Populate base rows first (tenant/env null), then overrides.
   • Optional: store a config_schema_version column per row if you want per-row evolution.

4) Cross-language implementation (repeat for each language)