claude@clouddev1 275c726ad4 feat: DSL→SQL teaching echo — Bucket B renderer (ADR-0038 Phase 2)
Expands the renderer to Bucket B — resolved-name single-statement
echoes plus the two category-2 multi-statement forms. Every catalogue
row round-trips per line through the advanced-mode walker (the §1
copy-paste contract; §6 category 2 holds the contract per line):

  add index [as N] on T (cols)           → CREATE INDEX <name> ON T (cols)
  drop index on T (cols) (positional)    → DROP INDEX <name>
  add 1:n relationship [as N] …          → ALTER TABLE C ADD CONSTRAINT
                                            <name> FOREIGN KEY (cc)
                                            REFERENCES P (pc) [ON …]
  drop relationship (endpoints or named) → ALTER TABLE C DROP CONSTRAINT
                                            <name>
  drop column T.c --cascade              → DROP INDEX <ix1> ⏎ … ⏎
                                            ALTER TABLE T DROP COLUMN c
  add relationship … --create-fk         → ALTER TABLE C ADD COLUMN cc <ty>
   (child column newly created)            ⏎ ALTER TABLE … ADD CONSTRAINT
   (already existed) collapses to a single-line FK echo

Refactors the echo payload from Option<String> to Option<Vec<String>>
across the 7 success events + arms + render path — one entry per
statement; the Bucket A single-line echoes wrap as Some(vec![s]). Plain
rendering repeats `Executing SQL:` per line; the de-emphasised
styled-runs polish (ADR-0038 §4) will refine it later.

Adds the two echo build paths the handoff §5 ⚠️ gotcha foreshadowed:

* collect_echo_lookups (pre-execution, runtime): resolves names the
  dropped thing or not-yet-created column would erase post-execution —
  drop index (positional), drop relationship (both endpoints and named
  selectors, the latter via a list_tables scan acceptable for teaching-
  playground schemas), and the --create-fk pre-state (whether the child
  column existed + the parent PK type to derive the new column type via
  Type::fk_target_type).
* build_schema_echo (post-execution, runtime): subsumes the Bucket A
  pure-Command schema cases and renders Bucket B from the description +
  the lookups.

The DropColumn arm gains build_drop_column_cascade_echo, which reads
DropColumnResult.dropped_indexes to emit the multi-line cascade echo;
non-cascade falls through to the pre-execution Bucket A echo unchanged.

Tests: 2013 passed / 0 failed / 1 ignored (pre-existing); clippy clean
(`--all-targets -D warnings`, nursery). Two end-to-end runtime tests
exercise the resolved-name and multi-statement flows against a real
worker (auto-named index, both drop-relationship selector forms, both
--create-fk branches). One app-level test pins the multi-line rendering
(one Executing SQL: per statement, in order, beneath [ok]).

Phase 3 (category-3 prose — shortid generation, type-conversion
transforms, `change column --dont-convert` caveat) and the §4
de-emphasised styled-runs rendering polish remain per ADR-0038 §8
phasing.
2026-05-28 07:54:05 +00:00
2026-05-07 11:17:58 +00:00
S
Description
No description provided
15 MiB
v0.1.0 Latest
2026-06-15 17:57:18 +01:00
Languages
Rust 97.4%
MDX 1.3%
JavaScript 0.8%
Astro 0.2%
CSS 0.1%
Other 0.1%