Add ltree_plruby: TRANSFORM FOR TYPE ltree

[jdatcmd]

Summary

New companion extension ltree_plruby providing TRANSFORM FOR TYPE ltree, closing the last identified PL/Python parity gap (mirrors the in-core ltree_plpython transform, ltree ↔ list).

An opted-in function receives an ltree argument as a Ruby Array of its label Strings — 'Top.Science.Astronomy' → ['Top', 'Science', 'Astronomy'] — and may return an Array of labels into an ltree result (elements stringified and joined with ., so an invalid label is rejected by ltree's own parser). The empty ltree maps to the empty Array; a NULL ltree is nil.

CREATE EXTENSION ltree_plruby CASCADE;

CREATE FUNCTION ancestors(path ltree) RETURNS SETOF ltree
TRANSFORM FOR TYPE ltree
LANGUAGE plruby AS $$
    (1..path.length).each { |n| return_next(path[0...n]) }
    nil
$$;

Design

Follows the hstore_plruby pattern exactly: it never touches ltree's internal representation. It converts through ltree's public ltree2text(ltree) / text2ltree(text) functions, resolved at run time via the pg_transform row that points at the transform function (schema/search_path independent) and cached in fn_extra. So it needs no ltree headers and works with the packaged ltree on PG 11–18.

Testing

New ltree_plruby regression suite: argument → Array, path operations (length/parent/reverse), build & round-trip, empty and NULL, non-Array rejection (clean error), SETOF/composite rows via return_next, trigger $_TD Array + 'MODIFY', and untransformed passthrough (still a String). Green on PG 12 and PG 18; wired into CI beside the jsonb/hstore transforms. INSTALL, README, and the language reference updated. Additive; no compatibility impact.

🤖 Generated with Claude Code