* feat(bokslut): kontantmetoden year-end cut-off for fordringar and skulder
Under kontantmetoden nothing reaches 1510/2440 during the year, but BFL
5 kap 2 § still requires fordringar och skulder to be booked at
rakenskapsarets utgang. That conversion did not exist: the AR/AP tie-outs
were permanently unreconciled by construction for all cash companies, and
the balance sheet omitted every open invoice.
Adds lib/core/bookkeeping/kontantmetod-cutoff.ts:
Fordringar: Debit 1510 / Credit 30xx / Credit 2618|2628|2638
Skulder: Debit 4-6xxx / Debit 2648 / Credit 2440
Moms goes to the VILANDE accounts, never 2611/2641. Under bokslutsmetoden
moms is reported at payment, and the vilande accounts are deliberately
absent from ACCOUNT_RUTA / ACCOUNT_TO_BOX, so parking it there keeps it out
of the momsdeklaration until the invoice is actually paid. Booking it to
2641 would claim the deduction a period early.
Two aggregate verifikat, each reversed on day 1 of the next period, and no
invoices.journal_entry_id link: the payment flows route on that link, so
per-invoice linking would send every new-year payment down the accrual
clearing path against a receivable the vandning already removed. Leaving it
unset means a new-year payment still books the normal kontantmetoden cash
entry at the real payment date.
Outstanding is computed from payment DATES, not remaining_amount: an
invoice settled in January was still a fordran on 31 December, and reading
remaining_amount would shrink the cut-off every day the bokslut is delayed.
Surfaced as a bokslut wizard reminder (warning, not a blocker: promoting it
would newly block every cash company mid-bokslut, which is a separate call).
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(bokslut): address compliance review on the kontantmetoden cut-off
Three findings from the Swedish compliance review, all real:
1. BFL 5 kap 6-7 § traceability. The aggregate verifikat collected invoice
references but never wrote them, so an examiner could not trace the
1510/2440 posting back to the affarshandelser behind it. Invoice numbers
now go into the entry `notes` via buildCutoffNote(), truncated past 50 so
the note stays a pointer to the reskontra rather than a copy of it.
2. Non-atomic posting. The cut-off and its vandning were two sequential
creates with no rollback: if the reversal threw, 1510/2440 stayed
permanently inflated and every new-year payment would double-book, which
is exactly what the module docstring warns about.
postKontantmetodCutoff now asserts the target period exists, is open, and
contains the reversal date BEFORE posting anything, so the common failures
refuse without writing. If a reversal still fails after its cut-off
committed, the cut-off is stornoed through reverseEntry() (BFL 5 kap 5 §:
never edit or delete a posted entry) and the original error is rethrown.
3. Silent vat_treatment default. Missing vat_treatment fell back to 25 %,
which would route a 12/6/undantagen invoice to the wrong vilande account
AND the wrong revenue account. Such rows are now collected into
CutoffCollection.unknownVatTreatment, excluded from the cut-off, refused
by the posting step, and surfaced as their own wizard reminder.
Adds 11 cases for postKontantmetodCutoff, which had none: every refusal path
asserts nothing was posted, and the storno-compensation path is covered in
both the happy and the storno-also-failed direction.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(bokslut): never split a reverse charge across the cut-off
Second compliance round. Verified the three data-dependent findings against
production before changing anything; two needed no change, one is hardened:
- Credit notes are NOT silently dropped: all 22 credit notes on prod carry
document_type='invoice', so they are inside the collected set exactly as
the comment claims. The filter only excludes proforma and delivery_note.
- Vilande account numbers verified against the BAS 2026 chart in
lib/bookkeeping/bas-data: 2618/2628/2638 utgaende, 2648 ingaende. The
suggested 2617/2627/2637 do not exist.
- Reverse charge: all 123 RC supplier invoices on prod carry vat_amount = 0,
so no RC moms could reach 2648 today. That was an implicit data invariant,
not an enforced one. CutoffPayable now carries reverseCharge and forces the
cut-off moms to 0 for those rows, so a stray amount can never post a
one-sided reverse charge into the single vilande bucket. The self-assessed
output/input pair stays with the payment entry, after the vandning.
Also names the reskontra as the underlag in a truncated aggregate note, so
the verifikat points at its specification rather than implying the listed
subset is the whole of it.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
* fix(bokslut): surface stray moms on momsfri invoices instead of absorbing it
Third compliance round, one legitimate new finding: moms on a treatment that
cannot carry Swedish output moms (export, omvand betalningsskyldighet,
undantagen) was folded into the revenue line with only a log.warn. That
balances the verifikat while silently swallowing a real invoicing error,
which is the netting the swedish-vat reference prohibits, and it was
inconsistent with how the same module already treats a missing
vat_treatment.
Those rows now travel the same path as a missing treatment: collected into
CutoffCollection.strayVatOnZeroRate, excluded from the cut-off, refused by
the posting step, and surfaced as their own wizard reminder.
buildCutoffLines keeps its balancing fallback for the case where such a row
reaches it directly: it is now a last resort rather than the normal path,
and it must still never invent a moms account nor unbalance the verifikat.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
---------
Co-authored-by: Jakob Wennberg <311770904+jakobwennberg-oss@users.noreply.github.com>
Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
Jakob Wennberg
2dff83e2f3