Release 2.3.0; documentation style pass

Bump the extension to 2.3 (all plruby changes are in the shared
library, so plruby--2.2--2.3.sql is a no-op version move) and date the
changelog. Documentation cleanup: em and en dashes replaced with
standard punctuation, arrows spelled out, decorative adjectives
removed, and version-specific wording where "modern Ruby" was vague.

Verified: all three suites pass with zero failures on PostgreSQL 11
through 18, and CREATE EXTENSION VERSION '2.0' followed by ALTER
EXTENSION UPDATE chains to 2.3.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

Author: jdatcmd

CHANGELOG.md

@@ -5,13 +5,18 @@ All notable changes to PL/Ruby are documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and the project aims to follow [Semantic Versioning](https://semver.org/).
 
-## [Unreleased]
+## [2.3.0] - 2026-07-05
+
+Inline class definitions, the hstore transform, and transform coverage in
+every conversion context, including triggers. All plruby changes are in the
+shared library; `ALTER EXTENSION plruby UPDATE` completes the upgrade after
+installing the new binary.
 
 ### Added
 
 - **Inline `class`/`module` definitions in function bodies.** Bodies now
   compile to a top-level lambda (with a delegator method), where `class` is
-  legal — it was a SyntaxError anywhere inside the previous `def`-based
+  legal; it was a SyntaxError anywhere inside the previous `def`-based
   compilation. Definitions register globally for the session, exactly like
   `plruby_modules` code, and work in functions, triggers, and `DO` blocks.
 - **Transforms reach nested contexts.** A `TRANSFORM FOR TYPE` declaration
@@ -22,50 +27,50 @@ and the project aims to follow [Semantic Versioning](https://semver.org/).
   `return_next` rows carry nested composite values.)
 - **Transforms in triggers.** A trigger function that declares
   `TRANSFORM FOR TYPE` receives transformed columns in `$_TD['new']`/
-  `$_TD['old']` and may put native Ruby data back via `'MODIFY'` — going
+  `$_TD['old']` and may put native Ruby data back via `'MODIFY'`, going
   beyond PL/Perl, whose transforms skip trigger data. Without the clause,
   triggers behave exactly as before.
-- **`hstore_plruby`** — a companion extension (in `hstore_plruby/`) providing
+- **`hstore_plruby`**: a companion extension (in `hstore_plruby/`) providing
   `TRANSFORM FOR TYPE hstore`: opted-in functions receive hstore arguments as
   a Ruby `Hash` of String keys to String-or-`nil` values and may return a
-  `Hash` into an hstore result (keys/values stringified, `nil` ↔ `NULL`).
+  `Hash` into an hstore result (keys/values stringified, `nil` maps to `NULL` in both directions).
   Implemented against hstore's public SQL functions, so it needs no hstore
-  headers and works with the packaged hstore on PostgreSQL 11–18.
+  headers and works with the packaged hstore on PostgreSQL 11-18.
 
-## [2.2.0] — 2026-07-05
+## [2.2.0] - 2026-07-05
 
-Native jsonb, modern Ruby, and CI. The `jsonb_plruby` companion extension
+Native jsonb, Ruby 3.3/3.4 support, and CI. The `jsonb_plruby` companion extension
 exchanges jsonb with native Ruby data; RubyGems is enabled in the embedded
 interpreter (restoring csv/bigdecimal/base64 on Ruby 3.4 and making
 installed gems requirable); every push is now verified by GitHub Actions on
-PostgreSQL 12–18 and Ruby 3.2/3.3/3.4 (PG 11 verified out-of-band). All
+PostgreSQL 12-18 and Ruby 3.2/3.3/3.4 (PG 11 verified out-of-band). All
 plruby changes are in the shared library; `ALTER EXTENSION plruby UPDATE`
 completes the upgrade after installing the new binary.
 
 ### Added
 
-- **`jsonb_plruby`** — a companion extension (in `jsonb_plruby/`) providing
+- **`jsonb_plruby`**: a companion extension (in `jsonb_plruby/`) providing
   `TRANSFORM FOR TYPE jsonb`: opted-in functions receive jsonb arguments as
   native Ruby data (`Hash`/`Array`/`String`/`Integer`/`Float`/booleans/`nil`)
   and return Ruby data into jsonb directly. Integers beyond `Float` precision
   and `BigDecimal` values serialize exactly.
-- **Structured `pg_raise`** — `pg_raise(level, message, detail:, hint:,
+- **Structured `pg_raise`**: `pg_raise(level, message, detail:, hint:,
   sqlstate:)` maps the keywords onto the corresponding `ereport` fields (like
   PL/pgSQL's `RAISE ... USING`); a PL/Ruby caller that rescues the resulting
   error reads them back via `PLRuby::Error#detail` / `#hint` / `#sqlstate`.
-- **`PLRuby::Error#detail` / `#hint`** — a caught database error now carries
+- **`PLRuby::Error#detail` / `#hint`**: a caught database error now carries
   its `DETAIL` and `HINT` alongside the SQLSTATE.
-- **Streaming `spi_query_prepared`** — executes a prepared plan through a
+- **Streaming `spi_query_prepared`**: executes a prepared plan through a
   cursor (block form and `PLRuby::Cursor` handle form) instead of
   materializing; the plan stays reusable after the cursor closes. It was
   previously an alias of `spi_exec_prepared`.
-- **Regression suite 35 → 37 tests** (`replace`: mid-session
+- **Regression suite 35 to 37 tests** (`replace`: mid-session
   `CREATE OR REPLACE` recompilation; extended `misc`/`sqlstate`/`prepare`),
   plus the `jsonb_plruby` suite.
-- **CI (GitHub Actions)** — every push and pull request builds both
-  extensions and runs both suites on PostgreSQL 12–18 (system Ruby 3.2) and
+- **CI (GitHub Actions)**: every push and pull request builds both
+  extensions and runs both suites on PostgreSQL 12-18 (system Ruby 3.2) and
   on Ruby 3.3/3.4 (PostgreSQL 18).
-- **Ruby 3.3 and 3.4 support** — alternate expected files cover their
+- **Ruby 3.3 and 3.4 support**: alternate expected files cover their
   changed `NoMethodError`/`Hash#inspect` output; validation errors are
   reported as a single line (Ruby 3.4's parser quotes the generated wrapper
   source over many lines); the cookbook's token recipe encodes with
@@ -78,48 +83,48 @@ completes the upgrade after installing the new binary.
   arrived as the text-form String.
 - **RubyGems is enabled in the embedded interpreter** (was `--disable-gems`):
   Ruby 3.4 ships `csv`, `bigdecimal`, and `base64` as bundled gems, which
-  plain `require` cannot see without it — and installed gems become
+  plain `require` cannot see without it, and installed gems become
   requirable as a side benefit. `did_you_mean`/`error_highlight` are disabled
   so error messages stay deterministic. Verified on Ruby 3.2, 3.3, and 3.4
   (CI covers all three).
 
-## [2.1.0] — 2026-07-05
+## [2.1.0] - 2026-07-05
 
 Feature and hardening release: broader trigger and argument-mode coverage,
-richer error objects, streaming SPI, a tested cookbook, and a regression
+error objects carrying SQLSTATE, streaming SPI, a tested cookbook, and a regression
 suite grown to 35 tests verified on PostgreSQL 11 through 18. All changes
 are in the shared library; `ALTER EXTENSION plruby UPDATE` completes the
 upgrade after installing the new binary.
 
 ### Added
 
-- **Expanded regression suite** — 20 → 28 tests, verified on **PostgreSQL 11,
+- **Expanded regression suite**: 20 to 28 tests, verified on **PostgreSQL 11,
   12, 13, 14, 16, and 18** with Ruby 3.2. New coverage: `bytea` conversion,
   special numeric/float values (`NaN`, `±Infinity`, integer overflow), nested
   and array-bearing composites, SPI DML status codes and NULL columns,
   DELETE-trigger `$_TD['old']` handling, the error/`ensure` model, prepared-plan
   reuse, standard-library `require`, `$stdout`/`$stderr` redirection, quoting
   helpers, and `plruby.start_proc`.
 
-- **Cursor streaming** — `spi_query(sql)` opens a portal and reads rows a batch
+- **Cursor streaming**: `spi_query(sql)` opens a portal and reads rows a batch
   at a time, so large results stream without materializing. Block form
-  (`spi_query(sql) { |row| … }`), handle form (`spi_fetchrow` /
+  (`spi_query(sql) { |row| ... }`), handle form (`spi_fetchrow` /
   `spi_cursor_close`), and `PLRuby::Cursor#each` (Enumerable).
-- **Broader string encoding** — text is tagged with the Ruby encoding matching
-  the database encoding (LATIN*, WIN*, EUC*, SJIS, KOI8, Big5, GBK, …), not just
+- **Broader string encoding**: text is tagged with the Ruby encoding matching
+  the database encoding (LATIN*, WIN*, EUC*, SJIS, KOI8, Big5, GBK, ...), not just
   UTF-8; unmapped encodings fall back to ASCII-8BIT.
-- **`PLRuby::Error#sqlstate`** — a caught PostgreSQL error now carries its
+- **`PLRuby::Error#sqlstate`**: a caught PostgreSQL error now carries its
   five-character `SQLSTATE` (e.g. `42P01`, `22012`) on the Ruby exception; it
   is `nil` on a PL/Ruby error not backed by a database error.
-- **`VARIADIC` arguments** — the variadic tail arrives as a single Ruby
+- **`VARIADIC` arguments**: the variadic tail arrives as a single Ruby
   `Array` argument (`VARIADIC "any"` remains unsupported).
-- **`INSTEAD OF` triggers** on views — `$_TD['when']` is `INSTEAD OF`, with
+- **`INSTEAD OF` triggers** on views: `$_TD['when']` is `INSTEAD OF`, with
   the same `nil`/`'SKIP'`/`'MODIFY'` return handling as BEFORE row triggers.
-- **Cookbook** — `doc/cookbook.md`, practical recipes built on Ruby's stdlib
+- **Cookbook**: `doc/cookbook.md`, recipes built on Ruby's stdlib
   (JSON reshaping, HMAC/PBKDF2, audit trigger, CSV, Zlib-to-bytea, BigDecimal,
   slugify, ERB, streaming scans, batch commits); every recipe in the "tested"
   section runs verbatim in the regression suite (`cookbook` test).
-- **Regression suite 28 → 35 tests**, adding `datetime` (date/time/timestamp/
+- **Regression suite 28 to 35 tests**, adding `datetime` (date/time/timestamp/
   interval), `jsonb`, `misc` (uuid/inet/enum/domain), `variadic`, `trigger2`
   (INSTEAD OF, `WHEN` clauses, deferred constraint triggers, composite-column
   `'MODIFY'`), `hostile` (mid-SRF errors, cursor misuse, ~1MB TOAST values,
@@ -135,7 +140,7 @@ upgrade after installing the new binary.
   statement-level, return value ignored.
 - **`spi_freeplan`** now invalidates the plan handle; reusing a freed plan
   raises a clear PL/Ruby error instead of silently failing.
-- **Trigger `'MODIFY'` with composite columns** — the modified-tuple path is
+- **Trigger `'MODIFY'` with composite columns**: the modified-tuple path is
   now datum-based, so assigning a `Hash` to a composite-typed field of
   `$_TD['new']` works (it previously failed with "malformed record literal").
 - **`spi_fetchrow` after `spi_cursor_close`** now returns `nil`; previously it
@@ -146,7 +151,7 @@ upgrade after installing the new binary.
 - A function body compiles to a method, so a `class`/`module` definition cannot
   appear inline in the body (use a `plruby_modules` module instead).
 
-## [2.0.0] — 2026-07-01
+## [2.0.0] - 2026-07-01
 
 The initial release of the modernized PL/Ruby: an MRI Ruby interpreter embedded
 in PostgreSQL, packaged as a `CREATE EXTENSION` and offering the same feature
@@ -155,34 +160,34 @@ library, non-thread-safe embedding).
 
 ### Added
 
-- **Functions** with native Ruby argument and return types — `Integer`,
+- **Functions** with native Ruby argument and return types: `Integer`,
   `Float`, `true`/`false`, `String`, `nil`, nested `Array` (multidimensional
   PostgreSQL arrays), and composite/record types as `Hash`.
-- **Set-returning functions** — `RETURNS SETOF` and `RETURNS TABLE(...)` with
+- **Set-returning functions**: `RETURNS SETOF` and `RETURNS TABLE(...)` with
   `return_next` (both explicit-value and no-argument forms, the latter reading
   the TABLE/OUT column locals from the running body's binding).
-- **Argument modes** — `IN`, `OUT`, `INOUT`, `TABLE`, and named parameters
+- **Argument modes**: `IN`, `OUT`, `INOUT`, `TABLE`, and named parameters
   aliased as local variables.
-- **Trigger functions** — `RETURNS trigger` with the `$_TD` hash and
+- **Trigger functions**: `RETURNS trigger` with the `$_TD` hash and
   `nil`/`'SKIP'`/`'MODIFY'` return semantics.
-- **Event trigger functions** — `RETURNS event_trigger`, with `$_TD['event']`
+- **Event trigger functions**: `RETURNS event_trigger`, with `$_TD['event']`
   and `$_TD['tag']`.
-- **Anonymous `DO` blocks** — `DO $$ ... $$ LANGUAGE plruby`.
-- **Database access (SPI)** — `spi_exec`, `spi_fetch_row`, `spi_processed`,
+- **Anonymous `DO` blocks**: `DO $$ ... $$ LANGUAGE plruby`.
+- **Database access (SPI)**: `spi_exec`, `spi_fetch_row`, `spi_processed`,
   `spi_status`, `spi_rewind`.
-- **Prepared statements** — `spi_prepare`, `spi_exec_prepared`,
+- **Prepared statements**: `spi_prepare`, `spi_exec_prepared`,
   `spi_query_prepared`, and `spi_freeplan`.
-- **Transaction control** in procedures — `spi_commit` and `spi_rollback`.
-- **Explicit subtransactions** — `subtransaction { ... }` and
+- **Transaction control** in procedures: `spi_commit` and `spi_rollback`.
+- **Explicit subtransactions**: `subtransaction { ... }` and
   `subtransaction(callable, ...)`, rolling back and re-raising on any exception.
-- **Quoting helpers** — `quote_literal`, `quote_nullable`, `quote_ident`.
+- **Quoting helpers**: `quote_literal`, `quote_nullable`, `quote_ident`.
 - **`elog(level, message)`** supporting `DEBUG`/`LOG`/`INFO`/`NOTICE`/`WARNING`/`ERROR`,
   and the narrower `pg_raise(level, message)`.
-- **`$_SHARED`** — a hash persisting across calls within a session; standard
+- **`$_SHARED`**: a hash persisting across calls within a session; standard
   output (`puts`, `print`) is forwarded to the server log.
-- **Session initialization** — module autoloading from a `plruby_modules` table
+- **Session initialization**: module autoloading from a `plruby_modules` table
   and a `plruby.start_proc` configuration setting.
-- **UTF-8 string handling** — text from the database is tagged with the
+- **UTF-8 string handling**: text from the database is tagged with the
   database encoding (UTF-8 when the database is UTF-8), so multibyte Ruby String
   operations (`length`, `reverse`, regexp, ...) behave correctly.
 - Packaging as a first-class extension (`CREATE EXTENSION plruby`) and a

Makefile

@@ -17,7 +17,7 @@ MODULE_big = plruby
 OBJS = plruby.o plruby_io.o plruby_spi.o
 
 EXTENSION = plruby
-DATA = plruby--2.0.sql plruby--2.1.sql plruby--2.2.sql plruby--2.0--2.1.sql plruby--2.1--2.2.sql
+DATA = plruby--2.0.sql plruby--2.1.sql plruby--2.2.sql plruby--2.3.sql plruby--2.0--2.1.sql plruby--2.1--2.2.sql plruby--2.2--2.3.sql
 
 # Ruby compile/link flags, discovered via RbConfig.
 RUBY ?= ruby

README.md

@@ -15,7 +15,7 @@
 PL/Ruby is a procedural-language handler that lets you write database functions
 in **Ruby**, stored and executed inside PostgreSQL. You get the expressiveness
 of Ruby and its standard library with the full power of a native PostgreSQL
-function — plain functions, set-returning functions, triggers, event triggers,
+function: plain functions, set-returning functions, triggers, event triggers,
 and procedures with transaction control.
 
 ```sql
@@ -30,7 +30,7 @@ SELECT hello('world');   -- Hello, world!
 
 > [!NOTE]
 > PL/Ruby embeds an MRI Ruby interpreter in the backend. It targets **PostgreSQL
-> 11–18** and **Ruby 3.x**, installs as a first-class `CREATE EXTENSION`, and
+> 11-18** and **Ruby 3.x**, installs as a first-class `CREATE EXTENSION`, and
 > mirrors the feature set of PL/php with a large set of PL/Perl- and
 > PL/Tcl-inspired capabilities.
 
@@ -40,20 +40,20 @@ SELECT hello('world');   -- Hello, world!
 
 | | |
 |---|---|
-| 🧩 **Scalars, arrays, composites** | Arguments arrive as native Ruby values — `Integer`, `Float`, `true`/`false`, `String`, nested `Array`, and composite/record types as `Hash`. |
+| 🧩 **Scalars, arrays, composites** | Arguments arrive as native Ruby values: `Integer`, `Float`, `true`/`false`, `String`, nested `Array`, and composite/record types as `Hash`. |
 | 🔁 **Set-returning functions** | `RETURNS SETOF` / `RETURNS TABLE` with `return_next`. |
 | ⚡ **Triggers** | Row & statement triggers via `$_TD` (with `'SKIP'` / `'MODIFY'`). |
 | 📣 **Event triggers** | Back `CREATE EVENT TRIGGER` with `RETURNS event_trigger`. |
 | 🗄️ **Database access (SPI)** | `spi_exec`, `spi_fetch_row`, `spi_processed`, `spi_status`, `spi_rewind`. |
-| 🌊 **Cursor streaming** | `spi_query` (block or handle), `spi_fetchrow`, `spi_cursor_close`, `Cursor#each` — consume huge results without materializing them. |
+| 🌊 **Cursor streaming** | `spi_query` (block or handle), `spi_fetchrow`, `spi_cursor_close`, `Cursor#each`. Consume large results without materializing them. |
 | 📝 **Prepared statements** | `spi_prepare` / `spi_exec_prepared` / `spi_query_prepared` / `spi_freeplan`. |
 | 🔐 **Transaction control** | `spi_commit` / `spi_rollback` in procedures, plus `subtransaction` blocks. |
 | 🧰 **Utilities** | `quote_literal` / `quote_nullable` / `quote_ident`, `elog`, `$_SHARED`. |
 | 📦 **Session setup** | Anonymous `DO` blocks, `plruby_modules` autoloading, and a `plruby.start_proc` hook. |
 | 🔄 **Transforms** | `jsonb_plruby` and `hstore_plruby`: functions declared `TRANSFORM FOR TYPE` exchange native Ruby Hashes/Arrays with `jsonb` and `hstore`. |
 
 See the [**language reference**](doc/plruby.md) for the full API, the
-[**cookbook**](doc/cookbook.md) for practical tested recipes, and the
+[**cookbook**](doc/cookbook.md) for tested recipes, and the
 [PL/Perl](doc/plperl-comparison.md) and [PL/Tcl](doc/pltcl-comparison.md)
 comparisons for feature-by-feature detail.
 
@@ -96,7 +96,7 @@ $$;
 
 ## Requirements
 
-- **PostgreSQL 11 or newer** (tested on 11–18; 18 recommended), with the server
+- **PostgreSQL 11 or newer** (tested on 11-18; 18 recommended), with the server
   development files that provide `pg_config`.
 - **Ruby 3.x** built as a shared library (`ENABLE_SHARED=yes`) with development
   headers. On Debian/Ubuntu, install `ruby-dev`.
@@ -120,9 +120,9 @@ See [**INSTALL**](INSTALL) for details, and run the regression suite with
 ## Security
 
 > [!WARNING]
-> **PL/Ruby is an untrusted language.** On modern Ruby there is no sandbox
+> **PL/Ruby is an untrusted language.** Ruby 3.0 and later have no sandbox
 > (`$SAFE` and object tainting were removed in Ruby 3.0), so a PL/Ruby function
-> can do anything the PostgreSQL server's operating-system user can — read and
+> can do anything the PostgreSQL server's operating-system user can: read and
 > write files, open network connections, run shell commands, and so on.
 
 The language is created **without** the `TRUSTED` attribute, so only superusers
@@ -132,7 +132,7 @@ to roles you would trust with the server's OS account.
 ## Documentation
 
 - [Language reference](doc/plruby.md)
-- [Cookbook: practical tested recipes](doc/cookbook.md)
+- [Cookbook: tested recipes](doc/cookbook.md)
 - [Installation](INSTALL)
 - [Changelog](CHANGELOG.md)
 - [Feature comparison: PL/Ruby vs PL/php vs PL/Perl vs PL/Tcl](doc/comparison.md)
@@ -141,4 +141,4 @@ to roles you would trust with the server's OS account.
 
 ## License
 
-PL/Ruby is licensed under the **MIT License** — see [LICENSE](LICENSE).
+PL/Ruby is licensed under the **MIT License**; see [LICENSE](LICENSE).