Files
accounted/lib/cash-accounts/service.ts
T
MattssonandClaude Fable 5.1 d670fe6663 feat(invoices): named payee accounts and per-invoice choice of bank account (#2233)
* fix(enable-banking): read BBAN from AccountIdentification.other and store it on the account

Enable Banking has no top-level `bban` key on AccountIdentification: a
Swedish BBAN (clearing + account number) arrives as `other.identification`
with `other.scheme_name = 'BBAN'`, or in `all_account_ids`. The client typed
`bban?: string` and read `.bban`, so the value was always undefined: no
connected account ever carried its clearing + account number, and domestic
counterparty accounts on transactions were dropped.

Type the identifiers per the OpenAPI spec, add extractBban() and
pickAccountIdentifier(), read counterparty identifiers through the scheme
list (IBAN, then BBAN/BGNR/PGNR, then anything), and store `bban` on
StoredAccount from the OAuth callback. The external_id dedup scope stays
IBAN-then-uid and is untouched.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UaZTY21HVN57hJoPXKSLjV

* feat(invoices): named payee accounts on cash_accounts with a default per currency

A company had exactly one set of payment instructions per invoice currency
(company_settings.invoice_payment_accounts), picked by currency alone. A
second SEK bank account, or a second bankgiro number, had nowhere to live.

cash_accounts is already the per-company bank-account entity. Migration
20260903150000 adds the payee fields (bankgiro, plusgiro, clearing +
account number, BBAN, BIC, Swish, foreign routing) plus invoice_payee, a
small invoice_payee_defaults table (one default account per currency; one
account may be the default for several currencies, a SEK account with an
IBAN is the usual EUR payee), and a SECURITY DEFINER mirror that rewrites
the legacy map and the SEK bank columns from the default accounts. Every
existing reader (PDF, email, reminders, v1, MCP) keeps working; the three
writers that only touched legacy columns (PUT /api/settings, v1 settings,
MCP update_company_settings) now write through to the default account, so
what an agent sets is what the PDF prints. Peppol PaymentMeans is built
from the resolver instead of the raw legacy column. bg_pg is dropped
(never read or written; NULL on every prod and staging row).

Backfill lands only on existing cash accounts (primary, IBAN match, or the
only enabled account in the currency). Entries with no target stay in the
map as the resolver fallback and get an attach action in settings.

New: POST /api/cash-accounts (manual bank account on the next free 19xx),
PATCH /api/cash-accounts/[id] payee fields (owner/admin), GET/PUT
/api/cash-accounts/payee-defaults. Settings page rewritten as an account
list with per-currency defaults. Behandlingshistorik and the full archive
cover the new table and columns.

Verified on staging: migration applied (11 defaults landed), mirror
trigger observed rewriting company_settings from a payee edit.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UaZTY21HVN57hJoPXKSLjV

* feat(invoices): choose which bank account an invoice is paid to, frozen at issue

Migration 20260903160000 adds invoices.payment_cash_account_id (FK to
cash_accounts, SET NULL) and invoices.payment_details, the payee fields
frozen when the account is chosen and refreshed at issue.

Resolver: resolveInvoicePaymentAccount / companyWithInvoicePaymentAccount /
assertInvoicePaymentAccountForRender take an optional override, and
hasRequiredInvoicePaymentAccount reads it from the invoice row, so every
surface (PDF, Swish QR, email, reminders, payment confirmation, Peppol,
recurring, staged MCP send) prints the frozen payee when one exists and the
company default per currency otherwise. Invoices that never chose an
account behave exactly as before.

Issue paths (mark-sent, send, v1 send, v1 mark-sent, Peppol send,
recurring, MCP send and mark-sent) refresh the snapshot from the account as
it is at issue; a chosen account that is disabled, un-flagged or unusable
for the currency blocks with INVOICE_SEND_PAYMENT_ACCOUNT_INVALID.

Writers: dashboard POST/PATCH, v1 create/update and MCP create_invoice
accept payment_cash_account_id and validate it against the company's payee
accounts (INVOICE_PAYEE_ACCOUNT_INVALID). Credit notes inherit the
original's payee; copies carry the choice; preview-pdf renders the chosen
account. The editor shows "Betalas till" under the currency when the
company has two or more usable payee accounts for that currency.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UaZTY21HVN57hJoPXKSLjV

* feat(invoices): book manual payments on the invoice's chosen bank account

Manual mark-paid (dashboard, v1, MCP gnubok_mark_invoice_as_paid) and the
booking dialog's proposed lines debited 1930 regardless of which bank
account the invoice asked to be paid to. They now resolve the chosen
payee account's ledger account (resolveInvoiceSettlementAccount) and fall
back to 1930 only when no account was chosen or the row is gone.

Bank-transaction matching keeps debiting the account the money landed on
and does not filter by the chosen account; between equal-confidence
candidates it prefers the invoice that asked to be paid to the landing
account. Scores are untouched, so nothing new auto-matches.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UaZTY21HVN57hJoPXKSLjV

* chore(invoices): keep the payload-size and phantom-column ceilings after the payee work

Shorten the new gnubok_create_invoice argument description (tools/list
payload was 29 bytes over the 60 kB budget), inline the cash-account payee
UPDATE/INSERT payloads and the settings select strings as literals so the
phantom-column scanner can read their columns, and reuse ACCOUNT_NUMBER_RE
instead of a hand-rolled copy.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UaZTY21HVN57hJoPXKSLjV

* fix(invoices): harden the payee model after review (admin-only payee columns, separate payee IBAN, company-scoped FK)

Review findings from CodeRabbit, Superagent, the Swedish accounting review
and three skeptic passes, resolved in one batch:

Schema (both migrations are unshipped and edited in place):
- cash_accounts.payee_iban: the printed IBAN is its own column. iban stays
  the bank identity written by every sync and used to re-pair on reconnect,
  so a sync can no longer rewrite an invoice instruction or resurrect a
  cleared IBAN. The backfill copies each currency entry verbatim onto the
  target account (IBAN match first, then primary), so every invoice keeps
  printing exactly what it printed before; the bank IBAN is never pushed
  onto invoices that did not carry one.
- Payee columns are owner/admin-only at the database (BEFORE trigger,
  service role exempt): cash_accounts is member-writable for bank sync, and
  the SECURITY DEFINER mirror would otherwise have let a member rewrite
  where customers pay.
- Revoking an account as payee or disabling it drops its defaults; deleting
  a default drops that currency from the map and clears the legacy SEK
  columns (an admin saying "nothing to print" must not keep printing a
  closed account). The mirror leaves the legacy SEK columns alone when the
  map has no SEK entry, so legacy-only companies are never wiped by a
  mirror run for another currency.
- Audit and mirror triggers fire on the same column set; anon and
  authenticated can no longer execute the trigger-only definer functions.
- invoices.payment_cash_account_id is a composite same-company FK with
  SET NULL scoped to the account column.

Code:
- Only 19xx bank accounts can be payee: PATCH, the defaults PUT (which now
  also requires enabled, payee-flagged and usable for the currency),
  resolveInvoicePayeeChoice, and the mark-paid settlement resolver (which
  also refuses disabled rows and logs every fallback to 1930).
- createManualBankAccount excludes every ledger slot any row already holds
  (findFreeLedgerAccount treats a manual holder as free; this path inserts).
- The legacy settings writers (PUT /api/settings, v1, MCP) write through to
  the account BEFORE updating company_settings and fail the request on
  error; the account is written before it is adopted as default so the
  mirror never sees an empty payee.
- snapshotInvoicePayee: dry runs no longer persist; a failed snapshot write
  blocks issue (INVOICE_PAYEE_SNAPSHOT_FAILED). v1 mark-sent/mark-paid
  projections carry the payee columns; v1 create validates the payee
  before the dry-run return and echoes it in the preview.
- pickAccountIdentifier: supplementary IBAN wins over a primary BBAN, and
  non-account schemes (card PANs) are never persisted.
- Editor shows the payee select for a single usable account with no
  default; the booking dialog waits for cash accounts before proposing
  lines; a failed default write no longer hides a created account.
- Behandlingshistorik names the account on created/deleted defaults.
- Regenerated skills/accounted-api; MCP argument description trimmed under
  the tools/list payload ceiling.

Declined: clearing legacy columns via a forward migration (the mirror now
does it on delete); Swedish review's "show the debit account in the
mark-paid UI" (the booking dialog already proposes and lets the user edit
the debit line); manual ledger collision (UNIQUE exists, and the create
path now rejects it with a clear error); Peppol aligning to the PDF value
for companies whose legacy column had drifted from the map (the PDF is the
customer-facing document; both now agree).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UaZTY21HVN57hJoPXKSLjV

* fix(invoices): read NEW.invoice_payee only on the cash_accounts branch of the mirror trigger

trg_mirror_invoice_payee_defaults fires for both tables; plpgsql resolves
record fields per expression, so the combined condition failed with
"record new has no field invoice_payee" whenever a default row changed,
which took down every pg-real case on the payee tables. The revoke/disable
check now sits inside its own TG_TABLE_NAME branch. The MCP settings
executor test mocks the payee write-through like the settings route test
already does.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UaZTY21HVN57hJoPXKSLjV

* fix(invoices): keep member disables from revoking payee defaults, gate payee on 1920-1999, fit the MCP payload

Cycle 3 of /resolve-pr on #2233.

Superagent P1: the SECURITY DEFINER mirror trigger deleted an admin's
invoice_payee_defaults rows whenever cash_accounts.enabled flipped to
false, and enabled is member-writable (the bank picker's "Synkas ej"), so
a member could undo an admin's payee decision. The trigger now drops
defaults only on the admin-only invoice_payee true -> false revoke; the
mirror trigger's WHEN no longer lists enabled. Disabled accounts stay out
of the pick lists and the send gate already refuses an invoice that chose
one. Applied to staging as the same function + trigger definition and
probed inside a rolled-back block: disable keeps the default and the
mirrored bankgiro, revoke clears both.

pg-real: the admin-guard test ran three expectations inside one
withUserContext transaction; the first raise aborted it and the next
statement failed with "current transaction is aborted". One transaction
per expectation now, and the member case also flips enabled to prove the
column stays member-level.

Swedish review: payee eligibility was /^19\d\d$/, which admits 1910 Kassa
and the 1911-1919 tills. A customer pays to a giro or bank account, so
isBankCashAccount, CreateCashAccountSchema.ledger_account and the PATCH
route now require BAS 1920-1999; tests cover 1910 and 1919.

Unit tests (3/4): the tools/list payload guard read 60 025, then 60 014
tokens after main merged #2166 and #2163 alongside this branch. The
ceiling is not bumped and no read on this surface is a demotion
candidate, so gnubok_create_invoice drops payment_cash_account_id;
agent-created invoices print the per-currency default and v1 REST plus
the editor keep the field. Recorded in DECISIONS.md.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UaZTY21HVN57hJoPXKSLjV

* chore(migrations): move invoices_payment_cash_account to 20260903183000 after colliding with main's KPI migration

origin/main merged 20260903160000_kpi_monthly_include_reversed_originals
while this branch held the same version; identical versions abort the
Supabase apply. Staging's schema_migrations row was moved to the new
version with the file.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UaZTY21HVN57hJoPXKSLjV

* fix(invoices): gate invoice_payee on BAS 1920-1999 at the database, and unblock the typecheck ratchet

Cycle 4 of /resolve-pr on #2233, on Emil's go.

Swedish review: the 1920-1999 payee rule lived only in the routes. The
cash_accounts_payee_admin_only trigger now also refuses invoice_payee on
any other ledger (INVOICE_PAYEE_ACCOUNT_INVALID, 23514), whoever writes
it, and the backfill only targets giro/bank rows, so a company whose
single enabled cash_accounts row is a Stripe clearing account keeps its
legacy bankgiro in company_settings instead of landing it on 1686. pg
test covers insert and update on 1686 and 1910; the function was applied
to staging and probed.

Typecheck ratchet: main is red from two merges that landed with failing
Checks, and every branch that syncs it inherits the errors.
  - #2242 added POST(req) calls to the fiscal-periods route test without
    the route params argument withRouteContext handlers take (25 errors
    in the file, baseline 23). All 25 calls now pass
    createMockRouteParams({}).
  - #2247 made SyncResult.requestedFromDate and historyNarrowed required;
    the 13 mockedSync results in the enable-banking accounts-route test
    lacked them. They now carry a fixed date and historyNarrowed: false.
Both files' tests pass unchanged in behaviour.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

* chore(migrations): move invoices_payment_cash_account to 20260903193000 after colliding with main's party_promotion

origin/main merged 20260903183000_party_promotion while this branch held
the same version. Staging's schema_migrations row must follow (pending:
the Supabase MCP was disconnected at the time of this commit).

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>

---------

Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
2026-09-03 21:06:45 +02:00

1574 lines
62 KiB
TypeScript
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
import { chunk as chunkIds } from '@/lib/utils'
import type { SupabaseClient } from '@supabase/supabase-js'
import type { CashAccount, CashAccountSource, MappingResult } from '@/types'
import { createLogger } from '@/lib/logger'
import { syncMappedAccounts } from '@/lib/import/account-sync'
import { getBASReference } from '@/lib/bookkeeping/bas-reference'
import { fetchAllRows } from '@/lib/supabase/fetch-all'
const log = createLogger('cash-accounts')
/**
* Suggested BAS account per currency. Single source — the enable-banking
* callback and the AccountPickerDialog both key off these.
*/
export const CURRENCY_LEDGER_DEFAULTS: Record<string, string> = {
SEK: '1930',
EUR: '1932',
USD: '1933',
GBP: '1934',
}
export function defaultLedgerForCurrency(currency: string): string {
return CURRENCY_LEDGER_DEFAULTS[currency.toUpperCase()] ?? '1930'
}
/**
* Canonical read/write surface for cash_accounts.
*
* Replaces ad-hoc reads of bank_connections.accounts_data for routing decisions.
* UI panels that just display balances may still read accounts_data until the
* follow-up migration drops that column.
*
* All methods accept an authenticated SupabaseClient and rely on RLS for tenancy
* isolation. Defense-in-depth filter by company_id is applied regardless.
*/
export interface ListCashAccountsOptions {
enabledOnly?: boolean
}
export interface UpsertFromPsd2Input {
bank_connection_id: string
external_uid: string
currency: string
ledger_account: string
iban?: string | null
/** Raw BBAN from the ASPSP (Swedish: clearing + account number). */
bban?: string | null
name?: string | null
balance?: number | null
available_balance?: number | null
balance_updated_at?: string | null
enabled?: boolean
/**
* Existing cash_accounts row this PSD2 account was matched to by IBAN
* (see resolvePsd2LedgerAccount). The row is promoted in place: it keeps its
* id, its ledger_account and its linked transactions, and is re-pointed at
* this connection + external_uid. Without this the reconnect path would try
* to INSERT a second row on the same ledger and trip the
* (company_id, ledger_account) UNIQUE constraint.
*/
reuse_cash_account_id?: string | null
}
/**
* Normalize an IBAN for comparison: ASPSPs format the same account both as
* "SE45 5000 0000 0583 9825 7466" and "SE4550000000058398257466", and a plain
* string compare would read those as two different accounts. Mirrors the
* normalization the sync path already applies when deriving external_ids.
*/
export function normalizeIban(iban: string | null | undefined): string | null {
if (!iban) return null
const normalized = iban.replace(/\s+/g, '').toUpperCase()
return normalized || null
}
export async function listForCompany(
supabase: SupabaseClient,
companyId: string,
opts: ListCashAccountsOptions = {},
): Promise<CashAccount[]> {
let q = supabase
.from('cash_accounts')
.select('*')
.eq('company_id', companyId)
.order('is_primary', { ascending: false })
.order('ledger_account', { ascending: true })
if (opts.enabledOnly) q = q.eq('enabled', true)
const { data, error } = await q
if (error) {
log.error('listForCompany failed', { companyId, error: error.message })
return []
}
return (data ?? []) as CashAccount[]
}
/**
* Primary cash account for a company. Filters by currency when provided. Falls
* back to the global primary (`is_primary = true`) when no currency-specific
* match exists.
*
* Used by skattekonto-booking's __PRIMARY_SEK__ sentinel and by transfer-pairing
* to identify the company's default settlement account.
*/
export async function getPrimary(
supabase: SupabaseClient,
companyId: string,
currency?: string,
): Promise<CashAccount | null> {
let q = supabase
.from('cash_accounts')
.select('*')
.eq('company_id', companyId)
.eq('is_primary', true)
.limit(1)
if (currency) q = q.eq('currency', currency.toUpperCase())
const { data, error } = await q.maybeSingle()
if (error) {
log.warn('getPrimary failed', { companyId, currency, error: error.message })
}
if (data) return data as CashAccount
if (currency) {
// Fall back to any-currency primary so a company without a SEK account still
// resolves the sentinel: rare but possible (manual cash-on-hand only).
const { data: anyPrimary } = await supabase
.from('cash_accounts')
.select('*')
.eq('company_id', companyId)
.eq('is_primary', true)
.maybeSingle()
if (anyPrimary) return anyPrimary as CashAccount
}
return null
}
/**
* One in-memory picture of the company's cash_accounts rows and the status of
* the bank connections holding them. Every #1643 helper below derives from it,
* so "orphaned", "live" and "same physical account" mean the same thing in the
* transfer detector, the match/link flows and the commit guards.
*/
interface CashAccountTopology {
rows: CashAccount[]
/** bank_connection_id -> bank_connections.status */
statuses: Map<string, string>
/** Ledger accounts that must never be PROPOSED or accepted as a counter leg. */
orphaned: Set<string>
/** A row on an ACTIVE connection: the live claim on that physical account. */
isLive: (row: CashAccount) => boolean
/**
* A row no connection holds a claim on any more: demoted to manual
* (bank_connection_id null) or still pointing at a REVOKED connection.
* Distinct from "not live": an expired/error/pending connection still
* holds the row and can come back through re-auth.
*/
isReleased: (row: CashAccount) => boolean
}
function currencyKey(currency: string | null | undefined): string {
return String(currency ?? '').toUpperCase()
}
/**
* Load the topology, or null when the row lookup fails. A failed connection
* lookup degrades to "no connection is known to be active": nothing is
* flagged orphaned (the conservative pre-fix behavior) and no row ranks as
* live.
*/
async function loadCashAccountTopology(
supabase: SupabaseClient,
companyId: string,
): Promise<CashAccountTopology | null> {
const { data, error } = await supabase
.from('cash_accounts')
.select('*')
.eq('company_id', companyId)
if (error) {
log.warn('cash_accounts topology lookup failed', { companyId, error: error.message })
return null
}
const rows = (data ?? []) as CashAccount[]
const connectionIds = [
...new Set(rows.map((r) => r.bank_connection_id).filter((id): id is string => id !== null)),
]
const statuses = await getConnectionStatuses(supabase, companyId, connectionIds)
// Two enabled rows on ONE active connection sharing (IBAN, currency) are
// deliberately BOTH live: no liveness signal has held up on prod (see
// DECISIONS.md, #1643), so nothing ranks them here.
const isLive = (r: CashAccount): boolean =>
r.enabled && r.bank_connection_id !== null && statuses.get(r.bank_connection_id) === 'active'
const isReleased = (r: CashAccount): boolean =>
r.bank_connection_id === null || statuses.get(r.bank_connection_id) === 'revoked'
const orphaned = new Set<string>()
// Stale IBAN twins of a live row: the live row IS that physical account
// now, so a demoted-to-manual, disabled, revoked-held, or expired/error-
// connection row carrying the same IBAN in the same currency is a leftover
// of a broken reconnect. A row held by a REVOKED connection is NOT orphaned
// on its own: the disconnect and supersede paths demote such rows to
// manual holders (bank_connection_id null, #916), and a row revoked bank-
// side or before that demotion existed is the same thing with the stale
// FK kept, i.e. a real account the user still tracks (commonly the
// company's only 1930). The key is (IBAN, currency), never the IBAN alone: multi-
// currency accounts (Revolut, Wise) copy one IBAN onto every currency
// pocket, and a manual or deselected GBP pocket beside a live SEK pocket is
// a distinct account the user still tracks, not an orphan.
const liveByAccount = new Map<string, CashAccount>()
for (const row of rows) {
const key = physicalAccountKey(row)
if (key && isLive(row)) liveByAccount.set(key, row)
}
for (const row of rows) {
if (isLive(row)) continue
const key = physicalAccountKey(row)
if (!key) continue
const live = liveByAccount.get(key)
if (live && live.ledger_account !== row.ledger_account) orphaned.add(row.ledger_account)
}
return { rows, statuses, orphaned, isLive, isReleased }
}
/**
* Identity of the physical bank account a row represents: normalized IBAN
* plus currency, or null for rows without an IBAN (manual, CSV, kassa).
*/
function physicalAccountKey(row: Pick<CashAccount, 'iban' | 'currency'>): string | null {
const iban = normalizeIban(row.iban)
return iban ? `${iban}|${currencyKey(row.currency)}` : null
}
/**
* Ledger accounts of the OTHER rows that represent the same physical account
* as `own` (same normalized IBAN, same currency), whatever their liveness.
* The row on `settlementAccount` is never a twin of itself.
*/
function twinLedgersOf(
topology: CashAccountTopology,
own: CashAccount | null,
settlementAccount: string,
): Set<string> {
const twins = new Set<string>()
const ownKey = own ? physicalAccountKey(own) : null
if (!own || !ownKey) return twins
for (const row of topology.rows) {
if (row.id === own.id || row.ledger_account === settlementAccount) continue
if (physicalAccountKey(row) !== ownKey) continue
twins.add(row.ledger_account)
}
return twins
}
/**
* Find the cash account an own-account TRANSFER may pair with, by IBAN.
*
* Replaces the old findByIban for this purpose (issue #1643): a broken reconnect
* can leave several rows carrying the same IBAN (the live account plus orphans
* held by a revoked connection, or demoted to manual), and proposing an orphan
* as the transfer's counter-account books real money onto a junk balance-sheet
* ledger. This finder therefore:
* - tolerates multiple rows on one IBAN (the old single-row lookup errored),
* - drops disabled rows and every row in the orphaned set (the same
* definition the commit guards use, so a proposal is never rejected later),
* - treats the transaction's OWN IBAN as "not a transfer": when the bank
* stamps the account's own IBAN as counterparty (interest, fees) every
* row on that IBAN in the same currency is the same physical account,
* whichever of them happens to be live. Only a pocket in ANOTHER currency
* on that IBAN (a multi-currency account exchanging between pockets) can
* still pair.
* When more than one candidate survives (two active twins of one account, or
* several currency pockets with nothing to pick between them) the finder
* returns null rather than guessing by ledger number: no proposal beats a
* wrong one, and that is also what the single-row lookup did before.
*/
export async function findPairableCashAccountByIban(
supabase: SupabaseClient,
companyId: string,
iban: string,
opts: { excludeCashAccountId?: string | null } = {},
): Promise<CashAccount | null> {
const wanted = normalizeIban(iban)
if (!wanted) return null
const topology = await loadCashAccountTopology(supabase, companyId)
if (!topology) return null
const onIban = topology.rows.filter((row) => normalizeIban(row.iban) === wanted)
if (onIban.length === 0) return null
const ownId = opts.excludeCashAccountId ?? null
const own = ownId ? (onIban.find((row) => row.id === ownId) ?? null) : null
let rows = onIban.filter(
(row) => row.enabled && row.id !== ownId && !topology.orphaned.has(row.ledger_account),
)
if (own) {
const ownCurrency = currencyKey(own.currency)
rows = rows.filter((row) => currencyKey(row.currency) !== ownCurrency)
}
if (rows.length === 0) return null
if (rows.length > 1) {
log.warn('several pairable cash accounts share the counterparty IBAN: not pairing', {
companyId,
ledgers: rows.map((row) => row.ledger_account),
})
return null
}
return rows[0]
}
export interface SiblingCashAccount {
id: string
ledger_account: string
currency: string | null
/** Held by an ACTIVE bank connection and enabled. */
live: boolean
/**
* No connection holds the row any more (bank_connection_id null or the
* connection is revoked). False for an expired/error/pending connection,
* which can still be renewed onto this row.
*/
released: boolean
}
export interface CashAccountSiblings {
own: SiblingCashAccount
siblings: SiblingCashAccount[]
}
/**
* The transaction's own cash_accounts row plus the OTHER rows that carry the
* same (normalized) IBAN in the same currency, i.e. the same physical bank
* account on a different ledger. Multi-currency accounts (Revolut, Wise) copy
* one IBAN onto every currency pocket, so an IBAN match alone would present a
* EUR pocket as a sibling of the SEK pocket; the currency key keeps those
* apart.
*
* A broken reconnect strands transactions on an orphaned row (e.g. 1931) while
* the live claim on the same underlying account sits on another row (e.g.
* 1940). Matching and linking against "the transaction's own ledger" then
* permanently misses vouchers booked on the live ledger; this helper names the
* sibling rows such flows may additionally consider, and lets manualLink
* re-point a stranded row at the live sibling (issue #1643).
*
* Returns null when the row cannot be found or on any lookup failure, and an
* empty sibling list when the row has no IBAN (manual/CSV accounts):
* broadening is an enhancement, never a requirement.
*/
export async function describeCashAccountSiblings(
supabase: SupabaseClient,
companyId: string,
cashAccountId: string,
): Promise<CashAccountSiblings | null> {
const topology = await loadCashAccountTopology(supabase, companyId)
if (!topology) return null
const ownRow = topology.rows.find((row) => row.id === cashAccountId)
if (!ownRow) return null
return describeSiblingsFromTopology(topology, ownRow)
}
function describeSiblingsFromTopology(
topology: CashAccountTopology,
ownRow: CashAccount,
): CashAccountSiblings {
const toSibling = (row: CashAccount): SiblingCashAccount => ({
id: row.id,
ledger_account: row.ledger_account,
currency: row.currency ?? null,
live: topology.isLive(row),
released: topology.isReleased(row),
})
const own = toSibling(ownRow)
const wanted = normalizeIban(ownRow.iban)
if (!wanted) return { own, siblings: [] }
const ownCurrency = currencyKey(ownRow.currency)
const seenLedgers = new Set<string>()
const siblings: SiblingCashAccount[] = []
for (const row of topology.rows) {
if (row.id === ownRow.id) continue
if (row.ledger_account === ownRow.ledger_account) continue
// A row the user deselected is never a destination (round 5): an
// automatic move must not land on a row the transactions page hides.
// A voucher booked only there is then refused as a cross-account link.
if (!row.enabled) continue
if (normalizeIban(row.iban) !== wanted) continue
if (currencyKey(row.currency) !== ownCurrency) continue
// UNIQUE (company_id, ledger_account) makes this a no-op in practice;
// kept so a duplicate row can never yield two siblings on one ledger.
if (seenLedgers.has(row.ledger_account)) continue
seenLedgers.add(row.ledger_account)
siblings.push(toSibling(row))
}
return { own, siblings }
}
/**
* Whether a link against a voucher booked on `sibling` should MOVE the
* transaction's cash_account_id there (manualLink) and, equivalently, whether
* that sibling's vouchers should be offered to the row at all
* (unmatched-entries). The decision is about the destination: move onto a
* live sibling always; onto a dead one only when the own row's holder is
* definitively gone (released) and no sibling is live. An own row on an
* expired/error/pending connection is still the syncing account, so its
* transactions never leave it for a dead twin (issue #1643, round 3).
*/
export function shouldRepointToSibling(
described: CashAccountSiblings,
sibling: SiblingCashAccount,
): boolean {
if (sibling.live) return true
if (!described.own.released) return false
return !described.siblings.some((row) => row.live)
}
/**
* The sibling rows of `cashAccountId` (see describeCashAccountSiblings), or []
* when the row has no IBAN or the lookup fails.
*/
export async function listSiblingCashAccounts(
supabase: SupabaseClient,
companyId: string,
cashAccountId: string,
): Promise<SiblingCashAccount[]> {
const described = await describeCashAccountSiblings(supabase, companyId, cashAccountId)
return described?.siblings ?? []
}
/**
* Ledger accounts of the sibling rows returned by listSiblingCashAccounts.
*/
export async function listSiblingLedgerAccounts(
supabase: SupabaseClient,
companyId: string,
cashAccountId: string,
): Promise<string[]> {
const siblings = await listSiblingCashAccounts(supabase, companyId, cashAccountId)
return siblings.map((row) => row.ledger_account)
}
/**
* Ledger accounts that must never be PROPOSED (or accepted) as the
* counter-account of a booking, because their cash_accounts row is orphaned
* (issue #1643): a row that is not live (demoted-to-manual, disabled, held
* by a REVOKED connection, or by an expired/error connection) whose (IBAN,
* currency) also belongs to a row on an ACTIVE connection. The active row IS
* that physical account now, so the stale twin is a leftover of a broken
* reconnect.
* A manual/CSV account without a live twin is NOT orphaned, and neither is a
* row held by a revoked connection without a live twin (a disconnected but
* real account, often the company's only 1930): transfers to such an account
* are legitimate, and so is another currency pocket of a multi-currency
* account that shares the live pocket's IBAN.
*/
export async function getOrphanedCounterLedgers(
supabase: SupabaseClient,
companyId: string,
): Promise<Set<string>> {
const topology = await loadCashAccountTopology(supabase, companyId)
return topology?.orphaned ?? new Set()
}
/**
* The first account in a mapping result that would book the COUNTER leg onto
* an orphaned cash-account ledger, or null when the result is clean. The
* settlement account itself is exempt: a transaction stranded on an orphaned
* row still books its own bank leg there (the only leg that belongs there).
*/
export function findOrphanedCounterLedger(
accounts: Array<string | null | undefined>,
settlementAccount: string,
orphanedLedgers: ReadonlySet<string>,
): string | null {
for (const account of accounts) {
if (!account || account === settlementAccount) continue
if (!/^19\d{2}$/.test(account)) continue
if (orphanedLedgers.has(account)) return account
}
return null
}
export interface CounterLegGuardResult {
mappingResult: MappingResult
/** The 19xx ledger the result must not book its counter leg on, or null. */
refusedLedger: string | null
}
/**
* Commit-time guard shared by every categorize path (issue #1643 problem 4).
* Runs after applySettlementAccount, on the legs that are NOT the settlement
* account:
* 1. A 19xx leg that is a twin of the settlement row (same IBAN, same
* currency) is that same physical account's bank leg learned on another
* ledger (a counterparty template learned while the account sat on 1931,
* replayed after the reconnect moved it to 1940). It is rewritten to the
* settlement account rather than refused: the business account is fine,
* only the bank side is stale.
* 2. If that leaves the result booking the settlement account against
* itself (a "transfer" between two ledgers of one physical account, e.g.
* interest whose counterparty IBAN is the account's own), the twin is
* refused: no revenue or expense would reach the P&L.
* 3. Any remaining 19xx counter leg in the orphaned set is refused.
* The cash_accounts lookup only runs when a non-settlement 19xx leg is
* present, so ordinary bookings pay nothing.
*/
export async function guardCounterLegs(
supabase: SupabaseClient,
companyId: string,
mappingResult: MappingResult,
settlementAccount: string,
settlementCashAccountId: string | null | undefined,
): Promise<CounterLegGuardResult> {
const isCounterCashLeg = (a: string | null | undefined): a is string =>
!!a && a !== settlementAccount && /^19\d{2}$/.test(a)
const legs = [
mappingResult.debit_account,
mappingResult.credit_account,
...mappingResult.vat_lines.map((l) => l.account_number),
].filter(isCounterCashLeg)
if (legs.length === 0) return { mappingResult, refusedLedger: null }
const topology = await loadCashAccountTopology(supabase, companyId)
if (!topology) return { mappingResult, refusedLedger: null }
const own = settlementCashAccountId
? (topology.rows.find((row) => row.id === settlementCashAccountId) ?? null)
: null
const twins = twinLedgersOf(topology, own, settlementAccount)
let result = mappingResult
const rewrittenTwin = legs.find((leg) => twins.has(leg)) ?? null
if (rewrittenTwin) {
const rewrite = (a: string): string => (twins.has(a) ? settlementAccount : a)
result = {
...mappingResult,
debit_account: rewrite(mappingResult.debit_account),
credit_account: rewrite(mappingResult.credit_account),
vat_lines: mappingResult.vat_lines.map((l) => ({
...l,
account_number: rewrite(l.account_number),
})),
}
if (result.debit_account === settlementAccount && result.credit_account === settlementAccount) {
return { mappingResult, refusedLedger: rewrittenTwin }
}
}
const remaining = [
result.debit_account,
result.credit_account,
...result.vat_lines.map((l) => l.account_number),
].filter(isCounterCashLeg)
const orphaned = findOrphanedCounterLedger(remaining, settlementAccount, topology.orphaned)
return { mappingResult: result, refusedLedger: orphaned }
}
export interface CounterLegContext {
/** Ledger of the transaction's own cash_accounts row, or null when unknown. */
settlementLedger: string | null
/** Other ledgers of the same physical account (same IBAN, same currency). */
twins: ReadonlySet<string>
}
export interface CounterLegTopology {
/** Ledgers that must never be PROPOSED or accepted as a counter leg. */
orphaned: ReadonlySet<string>
contextFor: (cashAccountId: string | null | undefined) => CounterLegContext
}
/**
* One topology load for a batch of transactions (the suggest-categories
* route), exposing the same twin and orphan rules guardCounterLegs applies at
* commit: a learned 19xx leg that is a twin of the transaction's own row is
* that account's stale BANK leg (rewrite it to the settlement ledger, never
* withhold), and only a true counter-position orphan disqualifies a
* suggestion. Returns null when the lookup fails (nothing is withheld).
*/
export async function loadCounterLegTopology(
supabase: SupabaseClient,
companyId: string,
): Promise<CounterLegTopology | null> {
const topology = await loadCashAccountTopology(supabase, companyId)
if (!topology) return null
const cache = new Map<string, CounterLegContext>()
return {
orphaned: topology.orphaned,
contextFor: (cashAccountId) => {
if (!cashAccountId) return { settlementLedger: null, twins: new Set() }
const cached = cache.get(cashAccountId)
if (cached) return cached
const own = topology.rows.find((row) => row.id === cashAccountId) ?? null
const context: CounterLegContext = own
? { settlementLedger: own.ledger_account, twins: twinLedgersOf(topology, own, own.ledger_account) }
: { settlementLedger: null, twins: new Set() }
cache.set(cashAccountId, context)
return context
},
}
}
/**
* Line-level counterpart of guardCounterLegs for the free-form booking
* dialog (POST /api/transactions/[id]/book), which submits explicit lines
* instead of a mapping result. Two shapes are covered:
* - Two or more distinct 19xx ledgers, one of them the transaction's own
* settlement ledger: a 19xx line that is a same-IBAN same-currency twin
* of the own row or an orphaned ledger is refused, since such a
* "transfer" books one physical account against itself or onto a junk
* ledger.
* - A single 19xx line that is NOT the own ledger (the user typed the bank
* leg where the money physically is, e.g. the live 1940 for a row
* stranded on the orphaned 1931): when that ledger is a sibling the row
* should move to (shouldRepointToSibling), the caller re-points
* transactions.cash_account_id there in the same locked UPDATE that
* links the voucher, exactly as manualLink does for the identical
* voucher reached through "Matcha mot befintlig verifikation". A twin
* the row may not move to (a dead or disabled twin) is refused, since
* posting would strand the only bank leg on a ledger no connection
* feeds (round 5); a non-twin foreign 19xx line posts as typed.
* An ordinary booking (single 19xx line on the own ledger) pays one PK read
* of the own row and nothing more; the topology is only loaded when a twin
* or foreign 19xx leg is present. Both fields are null when clean, not
* covered, or when the transaction has no cash_accounts row.
*/
export interface BookedLinesGuardResult {
/** The 19xx ledger the booking must not put in the counter position, or null. */
refusedLedger: string | null
/** cash_accounts row the transaction should be moved to on booking, or null. */
repointCashAccountId: string | null
}
const CLEAN_BOOKED_LINES: BookedLinesGuardResult = { refusedLedger: null, repointCashAccountId: null }
export async function guardBookedCounterLines(
supabase: SupabaseClient,
companyId: string,
accountNumbers: readonly string[],
settlementCashAccountId: string | null | undefined,
): Promise<BookedLinesGuardResult> {
const cashLegs = [...new Set(accountNumbers.filter((a) => /^19\d{2}$/.test(a)))]
if (cashLegs.length === 0 || !settlementCashAccountId) return CLEAN_BOOKED_LINES
if (cashLegs.length === 1) {
const { data: ownRow, error } = await supabase
.from('cash_accounts')
.select('ledger_account')
.eq('id', settlementCashAccountId)
.eq('company_id', companyId)
.maybeSingle()
if (error) {
log.warn('cash_accounts own-row lookup failed', { companyId, error: error.message })
return CLEAN_BOOKED_LINES
}
const ownLedger = (ownRow as { ledger_account?: string } | null)?.ledger_account ?? null
if (!ownLedger || ownLedger === cashLegs[0]) return CLEAN_BOOKED_LINES
const topology = await loadCashAccountTopology(supabase, companyId)
const own = topology?.rows.find((row) => row.id === settlementCashAccountId) ?? null
if (!topology || !own) return CLEAN_BOOKED_LINES
// A foreign 19xx line that is NOT a twin of the own row (a transfer to
// another physical account) posts as typed.
if (!twinLedgersOf(topology, own, ownLedger).has(cashLegs[0])) return CLEAN_BOOKED_LINES
const described = describeSiblingsFromTopology(topology, own)
const sibling = described.siblings.find((row) => row.ledger_account === cashLegs[0]) ?? null
if (sibling && shouldRepointToSibling(described, sibling)) {
return { refusedLedger: null, repointCashAccountId: sibling.id }
}
// A twin the row may not move to (a dead or disabled twin of a live or
// still-held row): posting would put the only bank leg on a ledger no
// connection feeds while the transaction stays here (issue #1643
// problem 4), the shape the categorize paths rewrite and manualLink
// refuses. Refuse it (round 5).
return { refusedLedger: cashLegs[0], repointCashAccountId: null }
}
const topology = await loadCashAccountTopology(supabase, companyId)
if (!topology) return CLEAN_BOOKED_LINES
const own = topology.rows.find((row) => row.id === settlementCashAccountId) ?? null
if (!own) return CLEAN_BOOKED_LINES
const settlementAccount = own.ledger_account
if (!cashLegs.includes(settlementAccount)) return CLEAN_BOOKED_LINES
const twins = twinLedgersOf(topology, own, settlementAccount)
const counterLegs = cashLegs.filter((a) => a !== settlementAccount)
const twin = counterLegs.find((a) => twins.has(a)) ?? null
if (twin) return { refusedLedger: twin, repointCashAccountId: null }
return {
refusedLedger: findOrphanedCounterLedger(counterLegs, settlementAccount, topology.orphaned),
repointCashAccountId: null,
}
}
/**
* bank_connections.status for the given connection ids. Missing ids (and a
* failed lookup, which returns an empty map) read as "status unknown".
*/
async function getConnectionStatuses(
supabase: SupabaseClient,
companyId: string,
connectionIds: readonly string[],
): Promise<Map<string, string>> {
if (connectionIds.length === 0) return new Map()
const { data, error } = await supabase
.from('bank_connections')
.select('id, status')
.eq('company_id', companyId)
.in('id', [...connectionIds])
if (error) {
log.warn('bank_connections status lookup failed', { companyId, error: error.message })
return new Map()
}
return new Map(
((data ?? []) as Array<{ id: string; status: string }>).map((c) => [c.id, c.status]),
)
}
/**
* Of the given bank_connection ids, return the subset whose connection row has
* status 'revoked'. A revoked connection no longer holds a live claim on its
* cash_accounts rows: the allocator, the picker-save collision guard, and
* upsertFromPsd2's promote-in-place path all treat those rows like manual
* holders so a reconnect can land back on its original ledger account.
*
* On lookup failure this returns an empty set (treat every connection as
* active): the conservative pre-fix behavior.
*/
export async function getRevokedConnectionIds(
supabase: SupabaseClient,
companyId: string,
connectionIds: readonly string[],
): Promise<Set<string>> {
const statuses = await getConnectionStatuses(supabase, companyId, connectionIds)
return new Set([...statuses.entries()].filter(([, status]) => status === 'revoked').map(([id]) => id))
}
/**
* Find a free BAS class-19 slot for a new PSD2 cash account, respecting the
* UNIQUE (company_id, ledger_account) constraint. A bank returning N
* same-currency accounts must not map them all to the currency default —
* that's exactly the collision this prevents.
*
* Rules:
* - The currency default (1930/1932/1933/1934) is available when no
* PSD2-backed row holds it. A manual holder (the seeded 1930 row) does
* not block it — upsertFromPsd2 promotes that row in place.
* Rows held by a REVOKED connection count as manual too: disconnecting a
* bank releases its ledger claims, so reconnecting the same bank gets its
* original slot back instead of overflowing to 1939.
* - Overflow walks the free-use 1931–1959 sub-account slots, skipping the
* four currency defaults (reserved as suggestions for their currencies)
* and any slot held by ANY existing row — promoting an unrelated manual
* account (SIE-imported, kassa) would silently steal it.
* - Overflow ALSO skips 19xx numbers that already exist in the company's
* chart of accounts, even when no cash_accounts row holds them. A chart
* imported from SIE carries the company's real bank accounts by name
* ("1942 Nordnet", "1938 Danske eSett Settlement") without any PSD2 row
* behind them, and handing one of those out as "free" is how a SEK
* företagskonto ended up proposed as 1942 Nordnet. Only when every
* chart-free slot is exhausted do we fall back to chart-occupied numbers,
* so a company with a fully populated 19xx chart still gets an answer.
* - `exclude` carries slots already assigned earlier in the caller's loop
* but not yet visible in the table.
*
* Returns null when no slot is free (or the lookup fails) — callers fall back
* to their previous behavior and surface the error.
*/
export async function findFreeLedgerAccount(
supabase: SupabaseClient,
companyId: string,
currency: string,
exclude: ReadonlySet<string> = new Set(),
): Promise<string | null> {
const preferred = defaultLedgerForCurrency(currency)
const { data: rows, error } = await supabase
.from('cash_accounts')
.select('ledger_account, bank_connection_id')
.eq('company_id', companyId)
if (error) {
log.error('findFreeLedgerAccount lookup failed', { companyId, error: error.message })
return null
}
// Chart accounts are advisory here: a failed lookup must not block
// allocation, it just costs us the "don't steal a named bank account" guard.
const { data: chartRows, error: chartError } = await supabase
.from('chart_of_accounts')
.select('account_number')
.eq('company_id', companyId)
.like('account_number', '19%')
if (chartError) {
log.warn('findFreeLedgerAccount chart lookup failed', {
companyId,
error: chartError.message,
})
}
const chartTaken = new Set(
((chartRows ?? []) as Array<{ account_number: string }>).map(r => r.account_number),
)
const typedRows = (rows ?? []) as Array<{ ledger_account: string; bank_connection_id: string | null }>
const revokedConnectionIds = await getRevokedConnectionIds(
supabase,
companyId,
[...new Set(typedRows.map(r => r.bank_connection_id).filter((id): id is string => id !== null))],
)
const anyTaken = new Set<string>()
const connectedTaken = new Set<string>()
for (const row of typedRows) {
anyTaken.add(row.ledger_account)
if (row.bank_connection_id !== null && !revokedConnectionIds.has(row.bank_connection_id)) {
connectedTaken.add(row.ledger_account)
}
}
if (!exclude.has(preferred) && !connectedTaken.has(preferred)) return preferred
const reserved = new Set(Object.values(CURRENCY_LEDGER_DEFAULTS))
const candidates: string[] = []
for (let n = 1931; n <= 1959; n++) {
const candidate = String(n)
if (reserved.has(candidate)) continue
if (exclude.has(candidate) || anyTaken.has(candidate)) continue
candidates.push(candidate)
}
// First pass: slots the chart has never heard of, so we can create them
// cleanly. Second pass: chart-occupied slots, the pre-fix behavior, only
// once nothing unnamed is left.
const unnamed = candidates.find(c => !chartTaken.has(c))
if (unnamed) return unnamed
if (candidates.length > 0) {
log.warn('findFreeLedgerAccount fell back to a chart-occupied slot', {
companyId,
currency,
ledger: candidates[0],
})
return candidates[0]
}
log.warn('findFreeLedgerAccount exhausted 1931–1959', { companyId, currency })
return null
}
/**
* Allocate a ledger slot for a new PSD2 account AND make sure that account
* number exists in the company's chart of accounts — cash_accounts has no FK
* to the chart, but booking (and the AccountPicker, which only lists chart
* accounts) breaks on numbers the chart doesn't know. Sub-accounts outside
* the BAS reference (1931, …) are created with metadata derived from the
* account number; standard numbers get their BAS name.
*/
export async function allocatePsd2LedgerAccount(
supabase: SupabaseClient,
companyId: string,
userId: string,
// accountName is accepted for caller compatibility but no longer names the
// chart account: see the BAS-style naming note in the function body (#1643).
input: { currency: string; accountName?: string | null; exclude?: ReadonlySet<string> },
): Promise<string | null> {
const ledger = await findFreeLedgerAccount(supabase, companyId, input.currency, input.exclude ?? new Set())
if (!ledger) return null
// The CHART account always gets a BAS-style name: the BAS reference name
// when the slot is a standard account (1930 Företagskonto, 1940 Övriga
// bankkonton, ...), otherwise "Bankkonto <CUR>" for a free-use sub-account
// (1931, 1935, ...). ASPSPs report the account holder (i.e. the company) as
// the account name, and every failed reconnect used to persist another 19xx
// chart account named after the company (issue #1643 problem 3). The bank's
// display name still lands on cash_accounts.name via upsertFromPsd2, which
// is what the pickers show; input.accountName is deliberately ignored here.
const name = getBASReference(ledger)?.account_name ?? `Bankkonto ${input.currency.toUpperCase()}`
const sync = await syncMappedAccounts(
supabase,
companyId,
userId,
[
{
sourceAccount: ledger,
sourceName: name,
targetAccount: ledger,
targetName: name,
confidence: 1,
matchType: 'exact',
isOverride: false,
},
],
false,
)
if (sync.error) {
log.error('allocatePsd2LedgerAccount chart sync failed', {
companyId,
ledger,
error: sync.error,
})
return null
}
return ledger
}
export interface Psd2LedgerResolution {
ledgerAccount: string
/**
* Existing row to promote in place, when the IBAN was already known. Null
* when the ledger was freshly allocated.
*/
reuseCashAccountId: string | null
source: 'iban' | 'allocated'
}
/**
* Decide which BAS account a PSD2 account should book to, IBAN first.
*
* The IBAN identifies the physical bank account; the provider's account `uid`
* does not survive a re-authorization at every ASPSP, and a fresh connect to
* an already-connected bank mints a new bank_connection row regardless. Both
* cases used to look like "an account we have never seen", so the allocator
* handed out the next free 19xx slot and the user's mapping (1930/1940/1941)
* silently moved to 1942-1946 on every consent renewal.
*
* Matching on the IBAN instead means a known account keeps its ledger, its
* cash_accounts row id and therefore its linked transactions. The previous
* holder's connection status is deliberately NOT considered: one IBAN is one
* physical account, so the connection that just authorized it is the one that
* owns it. This matters for the case that motivated the fix, where the old
* connection still reads 'active' because its session was killed bank-side
* without telling us.
*
* Returns null only when allocation itself fails; callers keep their existing
* fallback.
*/
export async function resolvePsd2LedgerAccount(
supabase: SupabaseClient,
companyId: string,
userId: string,
input: {
iban?: string | null
currency: string
accountName?: string | null
exclude?: ReadonlySet<string>
},
): Promise<Psd2LedgerResolution | null> {
const exclude = input.exclude ?? new Set<string>()
const wanted = normalizeIban(input.iban)
if (wanted) {
const { data, error } = await supabase
.from('cash_accounts')
.select('id, iban, ledger_account')
.eq('company_id', companyId)
.not('iban', 'is', null)
if (error) {
// Fall through to allocation: a failed lookup must not block the
// connection, it just costs us the reuse.
log.warn('resolvePsd2LedgerAccount iban lookup failed', {
companyId,
error: error.message,
})
} else {
const match = ((data ?? []) as Array<{ id: string; iban: string | null; ledger_account: string }>)
.find(row => normalizeIban(row.iban) === wanted)
// A ledger already claimed earlier in the caller's loop cannot be handed
// out twice, even on an IBAN hit: two rows on one ledger violate the
// (company_id, ledger_account) UNIQUE constraint.
if (match && !exclude.has(match.ledger_account)) {
return {
ledgerAccount: match.ledger_account,
reuseCashAccountId: match.id,
source: 'iban',
}
}
}
}
const allocated = await allocatePsd2LedgerAccount(supabase, companyId, userId, {
currency: input.currency,
accountName: input.accountName,
exclude,
})
if (!allocated) return null
return { ledgerAccount: allocated, reuseCashAccountId: null, source: 'allocated' }
}
/** Max transaction ids per `.in()` filter when rebinding: keeps the request URL short. */
const REBIND_ID_CHUNK_SIZE = 100
/**
* Rebind the MOVABLE transactions of one cash_accounts row onto another.
*
* Movable mirrors PATCH /api/transactions/[id]/cash-account (#1570): NOT
* booked (journal_entry_id), NOT confirmed-matched (invoice_id /
* supplier_invoice_id) and NOT anchored to a verifikat through
* transaction_voucher_links or an invoice/supplier-invoice payment row (both
* anchor without setting journal_entry_id on the transaction; see
* lib/transactions/is-booked.ts). An anchored row's voucher carries the 19xx
* line that records which ledger the money moved on, so its binding stays.
*
* The anchor tables cannot be expressed as a NOT EXISTS in a PostgREST update
* filter, so they are consulted in a pre-check; the column gate is re-asserted
* on the UPDATE itself against a concurrent book or auto-match.
*
* Runs before the target row is promoted and none of rebind / demote-or-delete
* / promote is transactional (PostgREST calls). Safe because the two rows share
* (bank_connection_id, external_uid), so their transactions already carry the
* currency the promote is about to write onto the target.
*
* @returns the number of rows rebound
*/
async function rebindMovableTransactions(
supabase: SupabaseClient,
companyId: string,
fromCashAccountId: string,
toCashAccountId: string,
): Promise<number> {
const candidates = await fetchAllRows<{ id: string }>(({ from, to }) =>
supabase
.from('transactions')
.select('id')
.eq('company_id', companyId)
.eq('cash_account_id', fromCashAccountId)
.is('journal_entry_id', null)
.is('invoice_id', null)
.is('supplier_invoice_id', null)
.order('id', { ascending: true })
.range(from, to),
)
if (candidates.length === 0) return 0
const candidateIds = candidates.map((row) => row.id)
const anchored = new Set<string>()
for (const chunk of chunkIds(candidateIds, REBIND_ID_CHUNK_SIZE)) {
const anchorRows = await Promise.all([
supabase
.from('transaction_voucher_links')
.select('transaction_id')
.eq('company_id', companyId)
.in('transaction_id', chunk),
supabase
.from('invoice_payments')
.select('transaction_id')
.eq('company_id', companyId)
.in('transaction_id', chunk),
supabase
.from('supplier_invoice_payments')
.select('transaction_id')
.eq('company_id', companyId)
.in('transaction_id', chunk),
])
for (const { data, error } of anchorRows) {
if (error) throw new Error(error.message)
for (const row of (data ?? []) as Array<{ transaction_id: string | null }>) {
if (row.transaction_id) anchored.add(row.transaction_id)
}
}
}
const movableIds = candidateIds.filter((id) => !anchored.has(id))
let moved = 0
for (const chunk of chunkIds(movableIds, REBIND_ID_CHUNK_SIZE)) {
const { data, error } = await supabase
.from('transactions')
.update({ cash_account_id: toCashAccountId })
.eq('company_id', companyId)
.eq('cash_account_id', fromCashAccountId)
.in('id', chunk)
.is('journal_entry_id', null)
.is('invoice_id', null)
.is('supplier_invoice_id', null)
.select('id')
if (error) throw new Error(error.message)
moved += (data ?? []).length
}
return moved
}
/** One account's refreshed balance snapshot, as the sync loop stores it. */
export interface SyncedBalanceInput {
external_uid: string
balance?: number | null
available_balance?: number | null
balance_updated_at?: string | null
}
/**
* Mirror freshly-synced balances from bank_connections.accounts_data into
* cash_accounts. Before this, cash_accounts.balance was written only at
* connect/selection-save time and then drifted: the transactions-page source
* picker (which reads cash_accounts) showed a connect-time snapshot as if it
* were current.
*
* Balance-only by design: routing fields (ledger_account, enabled, name) are
* owned by the picker-save and callback paths via upsertFromPsd2. Rows are
* matched on (company_id, bank_connection_id, external_uid); accounts without
* a timestamped balance are skipped (never null out a stored balance because
* one refresh was skipped or failed). Mirror failures are logged, not thrown:
* a failed mirror must not fail the sync that produced the data.
*/
export async function updateBalancesFromSync(
supabase: SupabaseClient,
companyId: string,
bankConnectionId: string,
accounts: SyncedBalanceInput[],
): Promise<void> {
for (const account of accounts) {
if (account.balance == null || !account.balance_updated_at) continue
// Manual sync and cron are not serialized per connection: an older run
// finishing later must not overwrite a newer mirror (the timestamp would
// visibly move backwards). Only rows with an older-or-missing timestamp
// accept the write. Two literal predicates instead of one .or(), and the
// payload inlined twice: the schema guard cannot resolve dynamically-built
// logical expressions or payload variables.
const { error: staleError } = await supabase
.from('cash_accounts')
.update({
balance: account.balance,
available_balance: account.available_balance ?? null,
balance_updated_at: account.balance_updated_at,
})
.eq('company_id', companyId)
.eq('bank_connection_id', bankConnectionId)
.eq('external_uid', account.external_uid)
.lt('balance_updated_at', account.balance_updated_at)
const { error: nullError } = await supabase
.from('cash_accounts')
.update({
balance: account.balance,
available_balance: account.available_balance ?? null,
balance_updated_at: account.balance_updated_at,
})
.eq('company_id', companyId)
.eq('bank_connection_id', bankConnectionId)
.eq('external_uid', account.external_uid)
.is('balance_updated_at', null)
const error = staleError ?? nullError
if (error) {
log.error('updateBalancesFromSync failed', {
companyId,
bankConnectionId,
externalUid: account.external_uid,
error: error.message,
})
}
}
}
/**
* Upsert a PSD2-sourced cash account during connection callback / sync. Keyed on
* (company_id, bank_connection_id, external_uid). When the row exists, balance
* and ledger_account are refreshed; the rest of the metadata stays put.
*
* Never sets is_primary: that's owned by the user via the AccountPicker or by
* the initial-backfill migration.
*/
export async function upsertFromPsd2(
supabase: SupabaseClient,
companyId: string,
input: UpsertFromPsd2Input,
): Promise<void> {
const payload = {
company_id: companyId,
bank_connection_id: input.bank_connection_id,
external_uid: input.external_uid,
iban: input.iban ?? null,
// Only overwrite when the bank sent one: a typed BBAN must survive a
// sync from an ASPSP that reports IBAN only.
...(input.bban ? { bban: input.bban } : {}),
name: input.name ?? null,
currency: input.currency.toUpperCase(),
ledger_account: input.ledger_account,
balance: input.balance ?? null,
available_balance: input.available_balance ?? null,
balance_updated_at: input.balance_updated_at ?? null,
enabled: input.enabled ?? true,
source: 'enable_banking' as CashAccountSource,
}
// create_company_with_owner and the seed_default_cash_account migration plant
// a manual (bank_connection_id IS NULL) row on the same ledger_account so
// reconciliation routes work before any PSD2 connection exists, and the
// disconnect handler demotes a revoked connection's rows to manual the same
// way. Rows still pointing at a REVOKED connection (orphans from before the
// disconnect handler released claims) no longer hold a live claim either.
// In all three cases the PSD2 sync claiming that BAS slot has to promote the
// holder row in place: a plain upsert on (company_id, bank_connection_id,
// external_uid) wouldn't match it and the INSERT path then trips the
// (company_id, ledger_account) UNIQUE constraint. Promoting (instead of
// inserting) keeps the row id stable so transactions.cash_account_id links
// and the ledger's history stay attached.
const { data: holderRow, error: holderLookupError } = await supabase
.from('cash_accounts')
.select('id, bank_connection_id')
.eq('company_id', companyId)
.eq('ledger_account', input.ledger_account)
.maybeSingle()
if (holderLookupError) {
log.error('upsertFromPsd2 holder lookup failed', {
companyId,
bankConnectionId: input.bank_connection_id,
externalUid: input.external_uid,
error: holderLookupError.message,
})
throw new Error(`cash_accounts upsert failed: ${holderLookupError.message}`)
}
const typedHolder = holderRow as { id: string; bank_connection_id: string | null } | null
let promotableRowId: string | null = null
if (typedHolder) {
if (typedHolder.id === input.reuse_cash_account_id) {
// Matched by IBAN upstream: this row IS this account, whoever held it
// last. Promoting keeps its id (transactions.cash_account_id stays
// linked) and re-points it at the connection that just authorized.
promotableRowId = typedHolder.id
} else if (typedHolder.bank_connection_id === null) {
promotableRowId = typedHolder.id
} else if (typedHolder.bank_connection_id !== input.bank_connection_id) {
const revoked = await getRevokedConnectionIds(supabase, companyId, [
typedHolder.bank_connection_id,
])
if (revoked.has(typedHolder.bank_connection_id)) {
promotableRowId = typedHolder.id
}
}
// Holder owned by the input connection itself (or by another ACTIVE
// connection): fall through to the plain upsert. For the former the upsert
// matches on (company_id, bank_connection_id, external_uid) and updates in
// place; for the latter the UNIQUE constraint rejects the write and the
// error surfaces to the caller (the picker-save collision guard should
// have caught it earlier).
}
if (promotableRowId) {
// Promoting the holder makes it THE row for this (bank_connection_id,
// external_uid). If this connection + uid already has a row on another
// ledger (the reconnect callback mirrored it onto an overflow slot while
// the target slot was still wrongly blocked by a revoked connection), that
// duplicate must be resolved first or the promote trips the UNIQUE
// (company_id, bank_connection_id, external_uid) constraint.
const { data: ownRow, error: ownLookupError } = await supabase
.from('cash_accounts')
.select('id, is_primary')
.eq('company_id', companyId)
.eq('bank_connection_id', input.bank_connection_id)
.eq('external_uid', input.external_uid)
.neq('id', promotableRowId)
.maybeSingle()
if (ownLookupError) {
log.error('upsertFromPsd2 duplicate lookup failed', {
companyId,
bankConnectionId: input.bank_connection_id,
externalUid: input.external_uid,
error: ownLookupError.message,
})
throw new Error(`cash_accounts upsert failed: ${ownLookupError.message}`)
}
const typedOwn = ownRow as { id: string; is_primary: boolean } | null
let transferPrimary = false
if (typedOwn) {
// The duplicate is the overflow mirror (e.g. 1931) a broken reconnect
// left behind; the promoted holder (e.g. 1930) is the ledger the user
// mapped. Movable transactions (unbooked, unmatched, not anchored to a
// verifikat) rebind onto the promoted row so categorize/booking proposes
// that ledger. Booked or anchored rows keep their binding: their
// vouchers carry the old 19xx line.
let movedCount = 0
try {
movedCount = await rebindMovableTransactions(
supabase,
companyId,
typedOwn.id,
promotableRowId,
)
} catch (rebindError) {
const message = rebindError instanceof Error ? rebindError.message : String(rebindError)
log.error('upsertFromPsd2 duplicate transaction rebind failed', {
companyId,
bankConnectionId: input.bank_connection_id,
externalUid: input.external_uid,
error: message,
})
throw new Error(`cash_accounts upsert failed: ${message}`)
}
if (movedCount > 0) {
// Behandlingshistorik (BFNAR 2013:2 kap 8): light-touch for a
// pre-verifikat staging binding, the same weight as the single-row
// move in PATCH /api/transactions/[id]/cash-account.
log.info('upsertFromPsd2 rebound movable transactions to promoted cash account', {
companyId,
fromCashAccountId: typedOwn.id,
toCashAccountId: promotableRowId,
movedCount,
})
}
// Rows still bound after the rebind are booked or anchored: the
// duplicate then survives as a demoted manual row (deleting it would SET
// NULL those transactions' cash_account_id links; the #1643 orphan
// guards handle the released twin). With nothing bound it is a leftover
// mirror and is deleted outright so its overflow slot frees up.
const { data: linkedTx, error: linkedTxError } = await supabase
.from('transactions')
.select('id')
.eq('company_id', companyId)
.eq('cash_account_id', typedOwn.id)
.limit(1)
if (linkedTxError) {
log.error('upsertFromPsd2 duplicate transaction check failed', {
companyId,
bankConnectionId: input.bank_connection_id,
externalUid: input.external_uid,
error: linkedTxError.message,
})
throw new Error(`cash_accounts upsert failed: ${linkedTxError.message}`)
}
if ((linkedTx ?? []).length > 0) {
const { error: demoteError } = await supabase
.from('cash_accounts')
.update({ bank_connection_id: null, external_uid: null })
.eq('id', typedOwn.id)
if (demoteError) {
throw new Error(`cash_accounts upsert failed: ${demoteError.message}`)
}
} else {
const { error: deleteError } = await supabase
.from('cash_accounts')
.delete()
.eq('id', typedOwn.id)
if (deleteError) {
throw new Error(`cash_accounts upsert failed: ${deleteError.message}`)
}
}
// A primary duplicate must hand the flag to the promoted row either way:
// deleted, it would leave the __PRIMARY_SEK__ sentinel unresolvable;
// demoted, the sentinel would keep resolving to the stale manual row.
transferPrimary = typedOwn.is_primary
}
// .select() so we can detect a 0-row UPDATE: Supabase's update().eq() returns
// { error: null, data: [] } if the row was deleted between the SELECT above
// and this UPDATE (rare but theoretically possible under concurrent ops).
// If that happens, fall through to the normal upsert path instead of
// silently returning success without persisting anything.
const { data: promoted, error: promoteError } = await supabase
.from('cash_accounts')
.update(payload)
.eq('id', promotableRowId)
.select('id')
if (promoteError) {
log.error('upsertFromPsd2 promote-holder failed', {
companyId,
bankConnectionId: input.bank_connection_id,
externalUid: input.external_uid,
error: promoteError.message,
})
throw new Error(`cash_accounts upsert failed: ${promoteError.message}`)
}
if (promoted && promoted.length > 0) {
if (transferPrimary) {
try {
await setPrimary(supabase, companyId, promotableRowId)
} catch (primaryError) {
// The promote itself succeeded; losing the primary flag is
// recoverable via the AccountPicker, so log instead of unwinding.
log.error('upsertFromPsd2 primary transfer failed', {
companyId,
cashAccountId: promotableRowId,
error: primaryError instanceof Error ? primaryError.message : String(primaryError),
})
}
}
return
}
// Holder row vanished between SELECT and UPDATE: fall through to upsert.
// Any rows the rebind above already moved onto it were SET NULL by the
// transactions.cash_account_id FK when it went; the next sync re-ingests
// them under the fresh row (the same self-heal as a deleted twin).
}
const { error } = await supabase
.from('cash_accounts')
.upsert(payload, { onConflict: 'company_id,bank_connection_id,external_uid' })
if (error) {
log.error('upsertFromPsd2 failed', {
companyId,
bankConnectionId: input.bank_connection_id,
externalUid: input.external_uid,
error: error.message,
})
throw new Error(`cash_accounts upsert failed: ${error.message}`)
}
}
/**
* Find (or create) a manual cash account for a BAS ledger slot, so transactions
* ingested outside the PSD2 flow (create_transactions / CSV) can carry a real
* cash_account_id instead of NULL. Without the link, reconciliation 404s on the
* account and the match dialog falls back to 1930 (issue #1016).
*
* Manual rows (source='manual', bank_connection_id=null) are already first-class:
* every company is seeded a manual 1930 the same way, and upsertFromPsd2 promotes
* a manual holder in place if a bank later claims the slot. So pre-creating one
* here does NOT race the PSD2 sync (the concern noted in lib/transactions/ingest.ts):
* the worst case is a later connection promoting this row, which is the intended flow.
*
* Keyed on the (company_id, ledger_account) UNIQUE constraint: a concurrent
* insert surfaces as 23505, which we treat as "someone else won the race" and
* re-read. The row's currency follows the first transaction that created it; a
* ledger account holds one currency by that same constraint.
*/
export async function ensureManualCashAccount(
supabase: SupabaseClient,
companyId: string,
ledgerAccount: string,
currency: string,
name?: string | null,
): Promise<string> {
const existing = await supabase
.from('cash_accounts')
.select('id, currency')
.eq('company_id', companyId)
.eq('ledger_account', ledgerAccount)
.maybeSingle()
if (existing.error) {
throw new Error(`ensureManualCashAccount lookup failed: ${existing.error.message}`)
}
if (existing.data) {
const row = existing.data as { id: string; currency: string | null }
// (company_id, ledger_account) is UNIQUE, so a ledger holds exactly one
// currency. A different-currency transaction pointing at the same ledger is
// a real conflict (e.g. a SEK row landing on a ledger already claimed for
// USD): fail loudly instead of binding it to the wrong-currency account.
if (row.currency && row.currency.toUpperCase() !== currency.toUpperCase()) {
throw new Error(
`Cash account ${ledgerAccount} is denominated in ${row.currency}, not ${currency.toUpperCase()}`,
)
}
return row.id
}
const insert = await supabase
.from('cash_accounts')
.insert({
company_id: companyId,
ledger_account: ledgerAccount,
currency: currency.toUpperCase(),
name: name?.trim() || `Bankkonto ${currency.toUpperCase()}`,
enabled: true,
is_primary: false,
source: 'manual' as CashAccountSource,
})
.select('id')
.single()
if (insert.error) {
// Lost the (company_id, ledger_account) race: re-read the winner's row.
if (insert.error.code === '23505') {
const reread = await supabase
.from('cash_accounts')
.select('id')
.eq('company_id', companyId)
.eq('ledger_account', ledgerAccount)
.maybeSingle()
if (reread.data) return (reread.data as { id: string }).id
}
log.error('ensureManualCashAccount insert failed', {
companyId,
ledgerAccount,
error: insert.error.message,
})
throw new Error(`ensureManualCashAccount insert failed: ${insert.error.message}`)
}
return (insert.data as { id: string }).id
}
/**
* Toggle a cash account's enabled flag. Used by the AccountPicker when a user
* opts in or out of syncing a particular PSD2 account.
*/
export async function setEnabled(
supabase: SupabaseClient,
companyId: string,
cashAccountId: string,
enabled: boolean,
): Promise<void> {
const { error } = await supabase
.from('cash_accounts')
.update({ enabled })
.eq('company_id', companyId)
.eq('id', cashAccountId)
if (error) throw new Error(`cash_accounts setEnabled failed: ${error.message}`)
}
/**
* Remap a cash account to a different BAS ledger account. Triggers RLS + the
* (company_id, ledger_account) UNIQUE constraint: surface conflict errors so
* the UI can prompt the user to resolve.
*/
export async function setLedgerAccount(
supabase: SupabaseClient,
companyId: string,
cashAccountId: string,
ledgerAccount: string,
): Promise<void> {
const { error } = await supabase
.from('cash_accounts')
.update({ ledger_account: ledgerAccount })
.eq('company_id', companyId)
.eq('id', cashAccountId)
if (error) throw new Error(`cash_accounts setLedgerAccount failed: ${error.message}`)
}
/**
* Set or clear the verifikationsserie override for a cash account. null means
* "follow the per-source-type default"; the engine reads this via
* resolveCashAccountVoucherSeries() when it books from the account.
*/
export async function setVoucherSeries(
supabase: SupabaseClient,
companyId: string,
cashAccountId: string,
voucherSeries: string | null,
): Promise<CashAccount | null> {
const { data, error } = await supabase
.from('cash_accounts')
.update({ voucher_series: voucherSeries })
.eq('company_id', companyId)
.eq('id', cashAccountId)
.select('*')
.maybeSingle()
if (error) throw new Error(`cash_accounts setVoucherSeries failed: ${error.message}`)
return (data as CashAccount | null) ?? null
}
/**
* Mark a cash account as the primary for its company. Delegates to the
* `set_cash_account_primary` RPC so the clear-old-primary and set-new-primary
* updates happen inside a single transaction. The intermediate "no primary"
* state is never visible to concurrent readers: important because
* skattekonto-booking's __PRIMARY_SEK__ resolver runs through getPrimary() and
* would otherwise see null in the gap and mis-route the counter account.
*/
export async function setPrimary(
supabase: SupabaseClient,
companyId: string,
cashAccountId: string,
): Promise<void> {
const { error } = await supabase.rpc('set_cash_account_primary', {
p_company_id: companyId,
p_cash_account_id: cashAccountId,
})
if (error) {
throw new Error(`cash_accounts setPrimary failed: ${error.message}`)
}
}