* fix(reconciliation): exclude opening balance from unmatched-1930 set (#443) Reconciliation incorrectly counted IB (source_type='opening_balance') vouchers on 1930 as unmatched bank transactions and included them in the GL period total. Result: an SIE-imported book always showed a phantom unmatched voucher and a difference equal to the IB amount, even when every real bank transaction was matched. This was the root of the "reconciliation says broken but numbers look right" complaint. The fix is two-sided: - get_unlinked_1930_lines RPC now skips source_type='opening_balance' rows. IB has no counterpart in the bank feed by definition. - getReconciliationStatus splits glOpeningBalance out of the period movement and computes difference = bankTotal − glPeriodMovement (was bankTotal − glBalance). gl_1930_balance is preserved unchanged for back-compat; new gl_1930_period_movement and gl_1930_opening_balance fields let the UI show both perspectives. - BankReconciliationView shows period movement vs bank in the diff and folds IB into a small caption ("Ingående balans … räknas inte i avstämningen") only when non-zero. Mirrors Fortnox's "Ingående saldo vid avstämningsstart" pattern — IB is verified separately, never surfaced as unmatched. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(reconciliation): use IS DISTINCT FROM for NULL-safe source_type filter `source_type <> 'opening_balance'` evaluates to NULL (not TRUE) when `source_type IS NULL` under SQL three-valued logic. journal_entries.source_type is NOT NULL today so the practical difference is zero, but if that constraint is ever relaxed `<>` would silently drop NULL rows from the unmatched-1930 set, making them invisible to reconciliation. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * fix(test): use valid source_type 'bank_transaction' in pg-real fixture CI pg-real failed because 'bank_import' is not in the journal_entries source_type CHECK constraint. The valid value for bank-imported transactions is 'bank_transaction' (per migration 20260513170001 + earlier). Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * docs(reconciliation): deprecate gl_1930_balance on ReconciliationStatus Mark the IB-inclusive balance field as @deprecated on both the server-side ReconciliationStatus interface and the component-side mirror. Downstream consumers that read this field expecting it to drive the diff will be off by the IB amount whenever a SIE-imported opening balance exists on 1930. The new gl_1930_period_movement field is the correct basis for the diff and is what the difference field is computed against. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> * refactor(reconciliation): tighten review feedback on PR #485 Three tweaks from the Swedish-accounting bot review: 1. Drop the gl_1930_period_movement ?? gl_1930_balance fallback in BankReconciliationView. gnubok has no persisted reconciliation snapshots and Vercel deploys atomically, so the old-server/new-client case can't occur — the fallback was defensive code for an impossible state. Per CLAUDE.md: don't add back-compat shims when you can just change the code. gl_1930_period_movement and gl_1930_opening_balance are now required on the type. 2. Clarify in the migration comment that source_type='opening_balance' is reserved for the fiscal-year IB voucher (always at period_start). The bot questioned whether the unconditional filter could hide a mid-year corrective IB entry; document the invariant — mid-year corrections use source_type='correction' or 'manual', never 'opening_balance'. 3. Test fixture 2099 → 2091. 2099 is årets resultat (current year); 2091 is balanserad vinst/förlust, the realistic counterpart for a carried-forward bank IB. RPC filters on 1930 only so this is cosmetic but matches what a real SIE import would produce. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
79 lines
2.8 KiB
PL/PgSQL
79 lines
2.8 KiB
PL/PgSQL
-- Exclude opening_balance vouchers from the unmatched-1930 set.
|
|
--
|
|
-- IB lines (source_type = 'opening_balance', typically created by SIE import
|
|
-- or year-end carry-over) post to 1930 on period_start. They have no
|
|
-- counterpart in the bank feed by definition — the bank statement starts at
|
|
-- IB and accumulates from there. Counting them as "unmatched" produces a
|
|
-- phantom voucher in the reconciliation UI and a difference equal to the IB
|
|
-- amount, even when every real bank transaction is matched.
|
|
--
|
|
-- Supersedes the prior definitions in:
|
|
-- - supabase/migrations/20240101000030_bank_reconciliation.sql
|
|
-- - supabase/migrations/20260401100000_fix_unlinked_1930_lines_company_id.sql
|
|
-- - supabase/migrations/20260415000000_schema_sync.sql
|
|
|
|
DROP FUNCTION IF EXISTS public.get_unlinked_1930_lines(uuid, date, date);
|
|
|
|
CREATE FUNCTION public.get_unlinked_1930_lines(
|
|
p_company_id UUID,
|
|
p_date_from DATE DEFAULT NULL,
|
|
p_date_to DATE DEFAULT NULL
|
|
)
|
|
RETURNS TABLE (
|
|
line_id UUID,
|
|
journal_entry_id UUID,
|
|
debit_amount NUMERIC,
|
|
credit_amount NUMERIC,
|
|
line_description TEXT,
|
|
entry_date DATE,
|
|
voucher_number INT,
|
|
voucher_series TEXT,
|
|
entry_description TEXT,
|
|
source_type TEXT
|
|
)
|
|
LANGUAGE sql
|
|
STABLE
|
|
SECURITY DEFINER
|
|
SET search_path = public
|
|
AS $$
|
|
SELECT
|
|
jel.id AS line_id,
|
|
je.id AS journal_entry_id,
|
|
jel.debit_amount,
|
|
jel.credit_amount,
|
|
jel.line_description,
|
|
je.entry_date,
|
|
je.voucher_number,
|
|
je.voucher_series,
|
|
je.description AS entry_description,
|
|
je.source_type
|
|
FROM public.journal_entry_lines jel
|
|
JOIN public.journal_entries je ON je.id = jel.journal_entry_id
|
|
WHERE jel.account_number = '1930'
|
|
AND je.company_id = p_company_id
|
|
AND je.status = 'posted'
|
|
-- Unconditional exclusion. By gnubok's data model 'opening_balance' is
|
|
-- reserved for the fiscal-year IB voucher (always posts on period_start);
|
|
-- mid-year corrective entries use 'correction' or 'manual'. So this filter
|
|
-- can't accidentally hide a legitimate mid-period unmatched entry — there
|
|
-- is no such thing as a mid-period opening_balance.
|
|
--
|
|
-- IS DISTINCT FROM is NULL-safe. Today journal_entries.source_type is
|
|
-- NOT NULL, so the only behavioural difference vs `<>` is defensive: if the
|
|
-- NOT NULL constraint is ever relaxed, `<>` would silently drop NULL rows
|
|
-- (NULL <> 'x' evaluates to NULL, not TRUE), making them invisible to
|
|
-- reconciliation. IS DISTINCT FROM treats NULL as a distinct value.
|
|
AND je.source_type IS DISTINCT FROM 'opening_balance'
|
|
AND (p_date_from IS NULL OR je.entry_date >= p_date_from)
|
|
AND (p_date_to IS NULL OR je.entry_date <= p_date_to)
|
|
AND NOT EXISTS (
|
|
SELECT 1
|
|
FROM public.transactions t
|
|
WHERE t.journal_entry_id = je.id
|
|
AND t.company_id = p_company_id
|
|
)
|
|
ORDER BY je.entry_date, je.voucher_number;
|
|
$$;
|
|
|
|
NOTIFY pgrst, 'reload schema';
|