Files
sure/app/models/recurring_transaction/identifier.rb
Guillem Arias Fauste 7c06fe6296 feat(recurring): allow marking transfers as recurring (#895) (#1589)
Refs #895, discussion #1224.

Adds a "Mark as recurring" entry point on the transfer detail drawer
that creates a `RecurringTransaction` carrying both source and
destination accounts. The recurring index, settings toggle
(`recurring_transactions_disabled`), and projected upcoming feed all
light up automatically once the data shape is there.

Schema:

* `destination_account_id` nullable FK to accounts. `on_delete: :cascade`
  matches #20251030172500's precedent for accounts FKs. The existing
  `account_id` FK is widened to cascade in the same migration so
  Family destruction with a recurring transfer doesn't FK-violate.
* Two predicate-partitioned partial unique indexes per shape:
  non-transfer rows (`destination_account_id IS NULL`, original
  5-column shape preserved) and transfer rows (6-column shape
  including the destination). Postgres treats NULLs as distinct in
  unique indexes, so widening would have broken non-transfer dedupe.
* Two CHECK constraints enforcing transfer invariants in PostgreSQL:
  `chk_recurring_txns_transfer_requires_source` (destination implies
  source) and `chk_recurring_txns_transfer_distinct_accounts`
  (destination cannot equal source). Per CLAUDE.md "Enforce null
  checks, unique indexes, and simple validations in the database
  schema for PostgreSQL".
* `Account` gains an `inbound_recurring_transfers` inverse so the
  destroy chain reaches both ends.

Controller / behaviour:

* `transfers#mark_as_recurring` mirrors `transactions#mark_as_recurring`:
  i18n flashes (4 new keys: transfer_marked_as_recurring,
  transfer_already_exists, transfer_creation_failed,
  transfer_feature_disabled), `respond_to format.html`,
  `redirect_back_or_to transactions_path`, server-side gate on
  `recurring_transactions_disabled?`, and rescue both `RecordInvalid`
  and `RecordNotUnique` for the race window between the dedupe
  `find_by` and `create_from_transfer`. The `StandardError` rescue
  now logs the exception (class, message, transfer/family/user ids)
  before surfacing the generic flash so production failures aren't
  context-less.
* `RecurringTransaction.accessible_by(user)` now requires
  destination_account_id (when present) to be in the user's
  accessible set, so a recurring transfer never leaks to a user
  without access to BOTH endpoints.
* Model validation gains a `destination_account.blank?` branch in
  `transfer_endpoints_consistent` so a dangling
  `destination_account_id` (referenced row destroyed) surfaces as a
  normal validation error instead of an FK exception on save.
* `Identifier` filter for transfer-kind transactions moved into SQL.

UI:

* Recurring index table and projected feed render transfer rows with
  the existing letter-avatar and the row's `name` field
  ("Transfer to {destination}"). No special pill or icon -- every row
  in `/recurring_transactions` is recurring by definition. Amount
  column on transfers uses `text-secondary` (muted-but-live) instead
  of the income/expense colour, since transfers are zero-net for the
  family.

Out of scope (called out in the PR body):

* Auto-creation of future Transfer rows on a schedule
  (discussion #1224's primary ask). Behaviour change vs the
  current projection-only model.
* Auto-identification of recurring transfer pairs in `Identifier`.
* Frequency model richer than `expected_day_of_month`.
* `Cleaner` for recurring transfers (issue #1590 tracks this).

Tests:

* `RecurringTransaction#transfer?` predicate (with / without
  destination).
* `transfer_endpoints_consistent`: rejects same source and
  destination, rejects dangling destination_account_id, rejects
  cross-family destination.
* `RecurringTransaction.create_from_transfer` happy path;
  multi-currency variant stores source-side currency.
* `projected_entry` exposes source / destination on transfer rows.
* `Identifier` skips transfer-kind transactions; creates a pattern
  from expense halves while ignoring co-resident transfer halves.
* Destroying the destination account cascades to inbound recurring
  transfers (FK + AR association).
* Unique partial index still de-duplicates non-transfer rows after
  the destination_account_id widening.
* `transfers#mark_as_recurring` happy path, idempotent on second
  call, rejected when `recurring_transactions_disabled`.

Suite: 3261 / 0 / 0 / 24 on the latest upstream/main. Lint clean.
Brakeman clean.

Signed-off-by: Guillem Arias Fauste <gariasf@proton.me>
2026-05-12 00:37:47 +02:00

305 lines
12 KiB
Ruby

class RecurringTransaction
class Identifier
attr_reader :family
def initialize(family)
@family = family
end
# Identify and create/update recurring transactions for the family
def identify_recurring_patterns
three_months_ago = 3.months.ago.to_date
# Skip transfer-kind transactions: they're one half of a Transfer pair, so grouping them
# under their single account would produce incoherent recurring "patterns" that don't
# represent the underlying account-pair flow. Recurring transfers are tracked on a
# different shape (RecurringTransaction with destination_account_id). Filtering at the
# SQL level avoids loading and discarding transfer entries for a busy family.
entries_with_transactions = family.entries
.joins("INNER JOIN transactions ON transactions.id = entries.entryable_id")
.where(entryable_type: "Transaction")
.where("entries.date >= ?", three_months_ago)
.where.not("transactions.kind": Transaction::TRANSFER_KINDS)
.includes(:entryable)
.to_a
# Group by merchant (if present) or name, along with amount (preserve sign) and currency.
grouped_transactions = entries_with_transactions
.select { |entry| entry.entryable.is_a?(Transaction) }
.group_by do |entry|
transaction = entry.entryable
# Use merchant_id if present, otherwise use entry name
identifier = transaction.merchant_id.present? ? [ :merchant, transaction.merchant_id ] : [ :name, entry.name ]
[ identifier, entry.amount.round(2), entry.currency, entry.account_id ]
end
recurring_patterns = []
grouped_transactions.each do |(identifier, amount, currency, account_id), entries|
next if entries.size < 3 # Must have at least 3 occurrences
# Check if the last occurrence was within the last 45 days
last_occurrence = entries.max_by(&:date)
next if last_occurrence.date < 45.days.ago.to_date
# Check if transactions occur on similar days (within 5 days of each other)
days_of_month = entries.map { |e| e.date.day }.sort
# Calculate if days cluster together (standard deviation check)
if days_cluster_together?(days_of_month)
expected_day = calculate_expected_day(days_of_month)
# Unpack identifier - either [:merchant, id] or [:name, name_string]
identifier_type, identifier_value = identifier
pattern = {
amount: amount,
currency: currency,
account_id: account_id,
expected_day_of_month: expected_day,
last_occurrence_date: last_occurrence.date,
occurrence_count: entries.size,
entries: entries
}
if identifier_type == :merchant
pattern[:merchant_id] = identifier_value
else
pattern[:name] = identifier_value
end
recurring_patterns << pattern
end
end
# Create or update RecurringTransaction records
recurring_patterns.each do |pattern|
# Build find conditions based on whether it's merchant-based or name-based
find_conditions = {
amount: pattern[:amount],
currency: pattern[:currency],
account_id: pattern[:account_id]
}
if pattern[:merchant_id].present?
find_conditions[:merchant_id] = pattern[:merchant_id]
find_conditions[:name] = nil
else
find_conditions[:name] = pattern[:name]
find_conditions[:merchant_id] = nil
end
begin
recurring_transaction = family.recurring_transactions.find_or_initialize_by(find_conditions)
# Handle manual recurring transactions specially
if recurring_transaction.persisted? && recurring_transaction.manual?
# Update variance for manual recurring transactions
update_manual_recurring_variance(recurring_transaction, pattern)
next
end
# Set the name or merchant_id on new records
if recurring_transaction.new_record?
if pattern[:merchant_id].present?
recurring_transaction.merchant_id = pattern[:merchant_id]
else
recurring_transaction.name = pattern[:name]
end
# New auto-detected recurring transactions are not manual
recurring_transaction.manual = false
end
recurring_transaction.assign_attributes(
expected_day_of_month: pattern[:expected_day_of_month],
last_occurrence_date: pattern[:last_occurrence_date],
next_expected_date: calculate_next_expected_date(pattern[:last_occurrence_date], pattern[:expected_day_of_month]),
occurrence_count: pattern[:occurrence_count],
status: recurring_transaction.new_record? ? "active" : recurring_transaction.status
)
recurring_transaction.save!
rescue ActiveRecord::RecordNotUnique
# Race condition: another process created the same record between find and save.
# Retry with find to get the existing record and update it.
recurring_transaction = family.recurring_transactions.find_by(find_conditions)
next unless recurring_transaction
# Skip manual recurring transactions
if recurring_transaction.manual?
update_manual_recurring_variance(recurring_transaction, pattern)
next
end
recurring_transaction.update!(
expected_day_of_month: pattern[:expected_day_of_month],
last_occurrence_date: pattern[:last_occurrence_date],
next_expected_date: calculate_next_expected_date(pattern[:last_occurrence_date], pattern[:expected_day_of_month]),
occurrence_count: pattern[:occurrence_count]
)
end
end
# Also check for manual recurring transactions that might need variance updates
update_manual_recurring_transactions(three_months_ago)
recurring_patterns.size
end
# Update variance for existing manual recurring transactions.
#
# Transfer rows (destination_account_id present) are skipped: their
# variance / occurrence tracking would need pair-detection across
# both endpoints rather than the single-account name/merchant match
# the helper performs. Issue #1590 tracks the proper Cleaner-aware
# matching for recurring transfers.
def update_manual_recurring_transactions(since_date)
family.recurring_transactions
.where(manual: true, status: "active", destination_account_id: nil)
.find_each do |recurring|
# Find matching transactions in the recent period
matching_entries = RecurringTransaction.find_matching_transaction_entries(
family: family,
merchant_id: recurring.merchant_id,
name: recurring.name,
currency: recurring.currency,
expected_day: recurring.expected_day_of_month,
lookback_months: 6,
account: recurring.account
)
next if matching_entries.empty?
# Extract amounts and dates from all matching entries
matching_amounts = matching_entries.map(&:amount)
last_entry = matching_entries.max_by(&:date)
# Recalculate variance from all occurrences (including identical amounts)
recurring.update!(
expected_amount_min: matching_amounts.min,
expected_amount_max: matching_amounts.max,
expected_amount_avg: matching_amounts.sum / matching_amounts.size,
occurrence_count: matching_amounts.size,
last_occurrence_date: last_entry.date,
next_expected_date: calculate_next_expected_date(last_entry.date, recurring.expected_day_of_month)
)
end
end
# Update variance for a manual recurring transaction when pattern is found
def update_manual_recurring_variance(recurring_transaction, pattern)
# Check if this transaction's date is more recent
if pattern[:last_occurrence_date] > recurring_transaction.last_occurrence_date
# Find all matching transactions to recalculate variance
matching_entries = RecurringTransaction.find_matching_transaction_entries(
family: family,
merchant_id: recurring_transaction.merchant_id,
name: recurring_transaction.name,
currency: recurring_transaction.currency,
expected_day: recurring_transaction.expected_day_of_month,
lookback_months: 6,
account: recurring_transaction.account
)
# Update if we have any matching transactions
if matching_entries.any?
matching_amounts = matching_entries.map(&:amount)
recurring_transaction.update!(
expected_amount_min: matching_amounts.min,
expected_amount_max: matching_amounts.max,
expected_amount_avg: matching_amounts.sum / matching_amounts.size,
occurrence_count: matching_amounts.size,
last_occurrence_date: pattern[:last_occurrence_date],
next_expected_date: calculate_next_expected_date(pattern[:last_occurrence_date], recurring_transaction.expected_day_of_month)
)
end
end
end
private
# Check if days cluster together (within ~5 days variance)
# Uses circular distance to handle month-boundary wrapping (e.g., 28, 29, 30, 31, 1, 2)
def days_cluster_together?(days)
return false if days.empty?
# Calculate median as reference point
median = calculate_expected_day(days)
# Calculate circular distances from median
circular_distances = days.map { |day| circular_distance(day, median) }
# Calculate standard deviation of circular distances
mean_distance = circular_distances.sum.to_f / circular_distances.size
variance = circular_distances.map { |dist| (dist - mean_distance)**2 }.sum / circular_distances.size
std_dev = Math.sqrt(variance)
# Allow up to 5 days standard deviation
std_dev <= 5
end
# Calculate circular distance between two days on a 31-day circle
# Examples:
# circular_distance(1, 31) = 2 (wraps around: 31 -> 1 is 1 day forward)
# circular_distance(28, 2) = 5 (wraps: 28, 29, 30, 31, 1, 2)
def circular_distance(day1, day2)
linear_distance = (day1 - day2).abs
wrap_distance = 31 - linear_distance
[ linear_distance, wrap_distance ].min
end
# Calculate the expected day based on the most common day
# Uses circular rotation to handle month-wrapping sequences (e.g., [29, 30, 31, 1, 2])
def calculate_expected_day(days)
return days.first if days.size == 1
# Convert to 0-indexed (0-30 instead of 1-31) for modular arithmetic
days_0 = days.map { |d| d - 1 }
# Find the rotation (pivot) that minimizes span, making the cluster contiguous
# This handles month-wrapping sequences like [29, 30, 31, 1, 2]
best_pivot = 0
min_span = Float::INFINITY
(0..30).each do |pivot|
rotated = days_0.map { |d| (d - pivot) % 31 }
span = rotated.max - rotated.min
if span < min_span
min_span = span
best_pivot = pivot
end
end
# Rotate days using best pivot to create contiguous array
rotated_days = days_0.map { |d| (d - best_pivot) % 31 }.sort
# Calculate median on rotated, contiguous array
mid = rotated_days.size / 2
rotated_median = if rotated_days.size.odd?
rotated_days[mid]
else
# For even count, average and round
((rotated_days[mid - 1] + rotated_days[mid]) / 2.0).round
end
# Map median back to original day space (unrotate) and convert to 1-indexed
original_day = (rotated_median + best_pivot) % 31 + 1
original_day
end
# Calculate next expected date
def calculate_next_expected_date(last_date, expected_day)
next_month = last_date.next_month
begin
Date.new(next_month.year, next_month.month, expected_day)
rescue ArgumentError
# If day doesn't exist in month, use last day of month
next_month.end_of_month
end
end
end
end