@ai-matrx/records-ui 0.77.0 → 0.79.0

package/CHANGELOG.md

@@ -1,5 +1,59 @@
 # Changelog — @ai-matrx/records-ui
 
+## 0.79.0
+
+**A joined column says what joins it.** (lane STORE-TAILS-2, VERIFIER-11 finding F5.)
+
+Greenline Landscaping Crew's Jobs board, "Job label" joining Job Number and Service Type,
+read `GL-2034mow_edge` on all 35 rows — beside a cell reading `Mow & Edge`. Two wrongs in
+one string. The machine's word is the store's half and is fixed there (`custom.rule_eval`'s
+concat arm now resolves a choice to its label and a relation to its display words through
+`custom.field_words`, while every comparison keeps the stored value). The missing separator
+is this package's: nothing on the field panel ever asked for one, so there was no way to
+send one.
+
+* **"Joined by"** — a new box on the worked-out-column builder, offered only for *Join them
+  together*, defaulting to ` — `. It is sent as `expr.separator`, which the store has
+  accepted since the migration above. Emptying it sends no key at all, which is exactly what
+  every formula written before today already holds, so nothing saved changes its answer.
+* A sum, a product or a difference is not offered the box and never sends the key — a
+  control that set something the evaluator ignores would be a setting that silently does
+  nothing.
+* **The panel reads an existing worked-out column back** — its operation, its columns and
+  its separator — instead of starting at "concat" with nothing picked whatever the column
+  actually held.
+* **And it says out loud what it still cannot do.** Opening an existing worked-out column
+  draws no formula controls (this panel never has), and now it says so in a sentence instead
+  of standing silent over the one thing that column is. Re-aiming an existing formula needs
+  `expr` on `FieldPatch` in `@ai-matrx/records`; the store door already takes it.
+
+Consumer action: none.
+
+## 0.78.0
+
+**The button that could not get rid of a table now can, and says what it does.** (lane
+FIX-11A, VERIFIER-11 finding F3.)
+
+`TableSettings` called `useRecordMutation().remove` with the TABLE's id — one unbounded
+statement — so on a 1,430-record table it died twelve seconds in on `57014` and told the
+person nothing had changed and to "try a smaller page". There is no page.
+
+* The panel now drives `useTableArchive` from `@ai-matrx/records` 0.53.0: it previews
+  first (`chunk: 0`) so the confirm names the real number of records, then archives in
+  passes and shows "N of M records put away so far" while it works.
+* **It never claims nothing happened.** A refused pass or a stopped run says what went
+  wrong, how many records are already archived and how many are left, and offers **Carry on
+  archiving**, which picks up exactly where it stopped. There is a **Stop** while it runs.
+* When the store is busy the loop takes smaller passes, and the screen SAYS so rather than
+  going quietly four times slower.
+* **The control's word matches what it does.** It said "Delete this table" over a sentence
+  explaining the store keeps everything for a year; the store's own verb is archive
+  (`table_archive`, `record_restore`, `read_records_archived`), so that is the word on the
+  button and the sentence says plainly it can be brought back.
+
+Consumer action: none — the panel's props are unchanged; `onDeleted` still fires, now when
+the Table itself has been archived.
+
 ## 0.77.0
 
 **A publish button that would hand out a dead link is not drawn.** (lane STORE-OFF, from