Correcting a single voucher cleared the momsdeklaration's RC_BASIS_MISSING
error and the whole per-voucher worklist with it: the check tested mere
presence of ruta 20-24 basis, the stepper re-derived its landing step and
yanked the user to Granska mid-work, and the remounted checks card never
refetched gaps once the aggregate check stopped firing. The declaration
then claimed "klart" while the remaining vouchers still under-reported
rutor 20-24 (FK004).
- Make RC_BASIS_MISSING/RC_OUTPUT_MISSING proportional: compare reported
basis against the basis the per-rate output boxes imply (moms/sats),
with a 0.5% + 1 kr tolerance for per-voucher ore rounding.
- Fetch the rc-basis-gaps worklist once per period, ungated from the
aggregate check, so remaining rows survive remounts.
- Latch the automatic stepper landing once per period so a refetch after
a korrigering cannot navigate the user off Kontrollera.
- Resolve rc-basis-gaps against the rakenskapsar (fiscal_period_id) for
helarsmoms, matching the declaration totals; a calendar span hid gap
vouchers in the tail of an extended first year.
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
* fix(reports): drive report line queries from journal_entries, not the unfiltered lines side
Every report generator fetched journal_entry_lines with a
journal_entries!inner(...) embed and put the tenant filter on the
embedded side (.eq('journal_entries.company_id', ...)). PostgREST
compiles that to a correlated INNER JOIN LATERAL with a parameterized
LIMIT inside, which blocks join reordering: Postgres walked the ENTIRE
journal_entry_lines table (603k rows, all tenants) per report query.
Measured in production: 13.6 s vs 2.7 ms for the equivalent plain join,
against Supabase's 8 s statement_timeout; nightly cloud backups failed
for 5 of 11 companies on 2026-07-09 and a GL report 500'd.
Introduce lib/bookkeeping/entry-lines.ts with a shared two-step fetch:
1. fetch matching journal_entries (id + caller-selected columns)
filtered by company_id / fiscal_period_id / status / entry_date /
source_type, paginated via fetchAllRows;
2. fetch journal_entry_lines with .in('journal_entry_id', chunk) in
chunks of 100 ids (URL-length safety), paginated per chunk;
3. reattach the parent entry to each line under the embed's key shape
(line.journal_entries = {...}, aliasable) and sort lines by id
ascending to preserve the old .order('id') semantics.
Converted call sites (selected columns and filters preserved):
trial-balance (x2), general-ledger, journal-register, sie-export
(reuses its existing entry list via fetchLinesByEntryIds),
vat-declaration, dimension-pnl, opening-balances, monthly-breakdown,
periodisk-sammanstallning, rc-basis-gaps (sibling-line fetch now also
chunked), ar-reconciliation, supplier-reconciliation,
bank-reconciliation, asset-service (x2), bolagsskatt-calculator,
sarskild-loneskatt-calculator.
Tests: unit tests for the helper (chunk size, reattachment shape,
forced id/journal_entry_id columns, empty result, cross-chunk sort,
error propagation); existing report/reconciliation/bokslut test mocks
updated to the two-step query shape, preserving every assertion about
report output.
From the 2026-07-09 production log triage.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
* fix(reports): stop echoing raw error messages from the general-ledger route
The catch handler returned err.message to the client in
details.reason; internal error strings (SQL fragments, table names,
timeout messages) must not reach the browser. The error is already
logged server-side with the request id, so the client envelope keeps
only the REPORT_GENERATION_FAILED code.
From the 2026-07-09 production log triage.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
---------
Co-authored-by: Claude Fable 5 <noreply@anthropic.com>
Em dashes (—) and en dashes (–) had spread across comments, docs, tests,
and a few UI strings, reading as AI-generated boilerplate rather than
house style. Replaced each with punctuation matching its context: colon
for explanatory clauses, comma for asides, plain hyphen for numeric/legal
ranges (e.g. "21-23§"), "to"/"till" for date ranges, parentheses for
paired-dash asides. messages/en.json and messages/sv.json were fixed by
hand together to keep sv/en in sync.
Left untouched where the dash is the functional subject rather than
decorative punctuation: date-range-parser.ts's separator regex,
charset-repair.ts's CP1252 byte-mapping table (and its test), the SIE
encoding mojibake docs, generic-csv.ts's minus-sign normalizer, the
agent system-prompt files that already instruct against em dashes, and
a golden iXBRL test fixture compared byte-for-byte.
Also fixes two bugs surfaced along the way: an off-by-one in
ApiKeysPanel's scope-label split (a leftover from an earlier partial
pass), and a charset-repair test that had lost the literal en-dash it
exists to verify.
Regenerated the agent atom seed migration (skills:generate) since 27
SKILL.md files changed. Added a CLAUDE.md rule against em/en dashes,
with an explicit carve-out for the functional-dash cases above.
Co-authored-by: Claude Sonnet 5 <noreply@anthropic.com>
* fix(reports): stabilize fetchAllRows paging to stop doubled/dropped balances (#790, #791)
PostgREST `.range()` paging is only correct when the underlying query has a
stable TOTAL order. Several aggregating report queries (general ledger, trial
balance, grundbok, supplier/AR ledgers, etc.) paginated without `.order()`, so
on datasets larger than one 1000-row page Postgres could return rows in a
different order between requests — silently DUPLICATING or SKIPPING rows on a
page boundary and doubling or dropping financial totals.
- fetch-all.ts: document the ordering invariant and add an optional
`dedupeBy` defense-in-depth that drops cross-page duplicates and warns when
it fires (surfaces a missing `.order()` in logs instead of corrupting money).
- Add a stable `.order()` (line PK or account_number) to every paginated query
in lib/reports/ and the account-balances route; pass `dedupeBy` on the
money-aggregating line queries.
- Add fetch-all unit tests and update report test fixtures to carry row ids.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(api): declare the real { data, meta } envelope on v1 single/write/204 endpoints (#794)
The OpenAPI generator derives each endpoint's documented body purely from its
registered `response.success` Zod schema, and that schema is never validated at
runtime — so a route could advertise a shape its handler never sends. #802
fixed this for list endpoints; the same drift was latent on single-resource and
write endpoints, which declared the bare resource schema instead of the
`{ data, meta }` envelope the handlers actually return.
- registry.ts: extend `ResponseMetaSchema` with the optional `audit` block and
`partial_expansions` list that writes/expansions emit; add the `NoBodyResponse`
sentinel so 204 DELETE handlers document a bare 204 instead of a phantom 200.
- Wrap every single/write endpoint's `response.success` in `dataEnvelope(...)`
(or `NoBodyResponse` for 204s) across the v1 routes.
- Add a response-envelope contract test that fails CI if any JSON endpoint
forgets to wrap its schema, with binary downloads and 204s as the only
exemptions.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* fix(reports): extend paging dedupeBy to rc-basis-gaps and opening-balances
Address PR review: these two money-aggregating line queries already had the
stable `.order('id')` (so paging was correct) but didn't carry `id` in the
select, so they couldn't use the `dedupeBy` defense-in-depth that general-ledger
and trial-balance got. Select `id` and pass `dedupeBy: r => r.id` so the whole
report layer applies the ordering invariant consistently.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Jakob Wennberg