Seeing a colleague's draft turned out to be half a feature: the point of
sharing it is to finish it while they are away. So writing is allowed
now -- but never by two people at once.
A draft is a single JSONB field. Whoever saves writes the whole state,
not the changed field, so two open wizards overwrite each other
completely and the second person sees nothing wrong: their own state is
right there on screen. That is why writing stayed with the owner until
now, and a lock is what makes giving that up safe.
The lock lives in the row (locked_by, locked_at) and is enforced by the
update and delete policies, not by the application. It expires, and that
is the important half: releasing happens when the wizard closes, and a
closed laptop never closes a wizard. Without expiry one crashed tab
would take a draft away for good -- worse than the problem being solved.
The wizard refreshes its lock while open so a long form does not lose it
mid-way.
Delete had to widen too, which reads like more than was asked for: the
wizard deletes the draft once the person is hired. Without it the hire
would go through and the draft would sit there forever. The card still
only offers delete on your own drafts.
Four of five mutations against the lock go red. The fifth -- dropping
`!open` from the refresh guard -- does not, because freigeben() already
nulls the ref the interval checks. The condition stays as the readable
statement of intent, now with a comment saying so.
Not run against a live database here; the CI migration job is the first
real execution.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The picker in the bell now governs both lists, so note_subscriptions is
renamed to colleague_subscriptions -- a name that only mentions notes would
mislead the next reader.
Reading and writing a draft now reach differently far. hire_drafts_owner
(for all) is split into four policies: select lets in your own drafts and
those of the people you added, while insert/update/delete stay with the
owner. A draft is unfinished work with no lock and no history; two people
writing into the same row would overwrite each other silently.
That split forces a change in the actions: a policy does not reject a write,
it lets it hit no rows. saveHireDraft and deleteHireDraft now read the row
count instead of reporting success over a row that never changed.
The card shows a foreign draft with its author and without Fortsetzen or
Loeschen -- offering a button that reliably ends in a database error is a
promise without cover.
check-schema-types.mjs learns `alter table ... rename to`; without it the
drift check reports one rename as two errors.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Yesterday's version had it the other way — everyone visible, untick to
hide. The decision from the business side is the opposite: you see your
own notes, and you tick the colleagues you also want. So note_mutes
becomes note_subscriptions and the predicate flips from `not exists` to
`exists`.
The existing rows are not carried over. The meaning inverts rather than
the sign: converting faithfully ("everyone except the muted") would write
almost the whole roster into the new table and reproduce exactly the state
the change is meant to end. Anyone opening the setting tomorrow would
think it had not taken effect. The table is a day old; what is lost is a
few ticks from trying it out.
What this costs is worth saying plainly: the silent case that could not
happen under exceptions can happen now. Do not tick a colleague and you
will not see her follow-ups — not while she is on holiday either. That is
the flip side of the decision, and it is written down in the migration
rather than discovered later.
Each note now says who wrote it. Own notes read "von mir" rather than
repeating your own name, which would sit on every second line and tell
nobody anything. The flag is computed on the server: the user id is
already there, and threading it through four components for one word is a
poor trade. The counter on the button follows the same turn — "+2" for
what you added, nothing when you added nothing.
Verified: 21 tests, five mutation-checked (restoring `not exists`,
dropping the own-notes clause, inverting the default, hiding the author,
and printing your own name instead of "von mir" each turn them red). 489
tests, typecheck, lint, schema drift and build clean. The migration is
reviewed but not run — no reachable database here.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The bell is a shared pile: every active HR person sees every open note,
regardless of who wrote it. That was agreed and it stays the default —
this narrows it, it never widens it. You can now untick colleagues whose
notes you do not want to see.
What gets stored is the *exceptions*, not the selection. The difference
shows the day someone new joins HR: had the selection been stored, she
would be invisible to everyone until each person ticked her, and nobody
would notice her follow-ups piling up. This way she is visible from day
one and hiding her is a deliberate act. Same reasoning that made notes a
shared inbox in the first place — the silent gap is worse than a row too
many.
Own notes always come through: `note_mutes` rejects a self-reference, and
the predicate says so again rather than depending on a check constraint
staying put. Notes with no author come through too — hiding one because
nobody knows who wrote it is exactly the loss this list exists to prevent.
The rule lives in lib/notes.ts as one SQL expression because two places
need it: the bell in the header and the "Anstehend" card on the dashboard.
Two copies drift, and then the card counts something the bell does not
show.
No SQL function and no audit row, unlike anything that touches employee
data — this is a personal display preference, and an audit trail recording
every tick would make finding real changes harder. Same pattern as saved
reports and hire drafts, and the owner policy on note_mutes means a row
for someone else cannot be written even with invented values.
The checkbox flips immediately and flips back if saving fails; the list
gets clicked through several at a time and a round trip per tick feels
like hesitation.
Verified: 19 tests, five mutation-checked (or→and, dropping the own-notes
clause, inverting `not exists`, inverting the default, and losing the
email fallback each turn them red). Typecheck, lint, schema drift, 477
tests and the build are clean. Not seen in a browser: login goes through
the company account and the database is unreachable — the migration is
reviewed but has not been run.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The app got slower as pages grew, and the reason was not the queries. It
was their number.
A transaction is pinned to one connection, and a connection runs queries
one after another. Every Promise.all in a withUser block looked like
concurrency and was a queue. Measured against the real database: the
round trip is ~36 ms, ten trivial `select 1` over one connection take
343 ms, over ten connections 39 ms. Nothing here is slow — the whole
dashboard payload is under 200 kB, and every table is around a thousand
rows.
More connections is the wrong answer: the RLS session context is per
transaction, so parallel reads mean parallel transactions, and those
multiply the connections the database will grant. Fewer round trips
instead. Postgres will return each sub-select as its own JSON column of
one result.
Per page view, counting the transaction frame:
shell (paid by every page) 10 → 4
overview 14 → 5
employee file 14 → 7
employee list 8 → 6
The overview plus its shell went from 24 round trips to 9 — about 860 ms
of pure waiting down to about 320 ms.
The one trap is documented where it bites: inside json_agg, Postgres
formats values itself and the driver's parsers (lib/db/pool.ts) never
see them. Dates, numerics and uuids come out identical; timestamptz does
not — "+00:00" where the driver gives "…Z". Timestamps are compared as
strings in lib/history.ts to decide what happened later, and those two
forms sort against each other wrongly. Every timestamptz in a bundled
query therefore goes through zeitstempel(), which was checked
character-for-character against the driver.
Four loaders moved out of their pages into lib/ so the number of round
trips can be measured without building a React tree, and so the new path
could be held against the old one field by field: same rows, same order,
same strings, for the overview and for four employee files chosen to
differ (with history, a chief, a planned entry, one with dependents).
withUser now counts the queries in each transaction and says so in
development past a threshold. Without that, this grows back: each new
tile brings its own query, and nobody notices until everybody does.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
