feat: add backfill service for historical rollup data

This commit is contained in:
Shivam Mishra
2026-03-13 10:55:10 +05:30
parent 13860b9163
commit 914e70568f
2 changed files with 232 additions and 0 deletions
@@ -0,0 +1,125 @@
# frozen_string_literal: true
class ReportingEvents::BackfillService
AGGREGATE_SELECTS = [
:name,
:user_id,
:inbox_id,
Arel.sql('COUNT(*)'),
Arel.sql('COALESCE(SUM(value), 0)'),
Arel.sql('COALESCE(SUM(value_in_business_hours), 0)')
].freeze
def self.backfill_date(account, date)
new(account, date).perform
end
def initialize(account, date)
@account = account
@date = date
end
def perform
delete_existing_rollups
start_utc, end_utc = date_boundaries_in_utc
rollup_rows = build_rollup_rows(start_utc, end_utc)
bulk_insert_rollups(rollup_rows) if rollup_rows.any?
end
private
def delete_existing_rollups
ReportingEventsRollup.where(account_id: @account.id, date: @date).delete_all
end
def date_boundaries_in_utc
tz = ActiveSupport::TimeZone[@account.reporting_timezone]
start_in_tz = tz.parse(@date.to_s)
end_in_tz = start_in_tz + 1.day
[start_in_tz.utc, end_in_tz.utc]
end
def build_rollup_rows(start_utc, end_utc)
aggregates = build_aggregates(start_utc, end_utc)
aggregates.map do |(dimension_type, dimension_id, metric), data|
{
account_id: @account.id,
date: @date,
dimension_type: dimension_type,
dimension_id: dimension_id,
metric: metric,
count: data[:count],
sum_value: data[:sum_value],
sum_value_business_hours: data[:sum_value_business_hours],
created_at: Time.current,
updated_at: Time.current
}
end
end
def build_aggregates(start_utc, end_utc)
aggregates = Hash.new { |h, k| h[k] = { count: 0, sum_value: 0.0, sum_value_business_hours: 0.0 } }
grouped_events(start_utc, end_utc).each { |grouped_event| accumulate_grouped_aggregates(aggregates, grouped_event) }
aggregates
end
def grouped_events(start_utc, end_utc)
@account.reporting_events
.where(name: ReportingEvents::MetricRegistry::EVENT_METRICS.keys, created_at: start_utc...end_utc)
.group(:name, :user_id, :inbox_id)
.pluck(*AGGREGATE_SELECTS)
.map { |grouped_row| grouped_event_attributes(grouped_row) }
end
def dimensions(grouped_event)
{
'account' => @account.id,
'agent' => grouped_event[:user_id],
'inbox' => grouped_event[:inbox_id]
}
end
def accumulate_grouped_aggregates(aggregates, grouped_event)
ReportingEvents::MetricRegistry.event_metrics_for_aggregate(
grouped_event[:event_name],
count: grouped_event[:count],
sum_value: grouped_event[:sum_value],
sum_value_business_hours: grouped_event[:sum_value_business_hours]
).each do |metric, metric_data|
accumulate_metric_aggregates(aggregates, dimensions(grouped_event), metric, metric_data)
end
end
def grouped_event_attributes(grouped_row)
event_name, user_id, inbox_id, count, sum_value, sum_value_business_hours = grouped_row
{
event_name: event_name,
user_id: user_id,
inbox_id: inbox_id,
count: count,
sum_value: sum_value,
sum_value_business_hours: sum_value_business_hours
}
end
def accumulate_metric_aggregates(aggregates, dimensions, metric, metric_data)
dimensions.each do |dimension_type, dimension_id|
next if dimension_id.nil?
key = [dimension_type, dimension_id, metric]
aggregates[key][:count] += metric_data[:count]
aggregates[key][:sum_value] += metric_data[:sum_value].to_f
aggregates[key][:sum_value_business_hours] += metric_data[:sum_value_business_hours].to_f
end
end
def bulk_insert_rollups(rollup_rows)
# rubocop:disable Rails/SkipsModelValidations
ReportingEventsRollup.insert_all(rollup_rows)
# rubocop:enable Rails/SkipsModelValidations
end
end
@@ -0,0 +1,107 @@
require 'rails_helper'
describe ReportingEvents::BackfillService do
describe '.backfill_date' do
let(:account) { create(:account, reporting_timezone: 'America/New_York') }
let(:date) { Date.new(2026, 2, 11) }
let(:user) { create(:user, account: account) }
let(:inbox) { create(:inbox, account: account) }
let(:conversation) { create(:conversation, account: account, inbox: inbox, assignee: user) }
it 'treats nil metric values as zero during backfill' do
reporting_event = create(
:reporting_event,
account: account,
name: 'first_response',
value: 100,
value_in_business_hours: 50,
user: user,
inbox: inbox,
conversation: conversation,
created_at: Time.utc(2026, 2, 11, 15)
)
# Simulate a legacy row that already exists in the database with nil metrics.
# rubocop:disable Rails/SkipsModelValidations
reporting_event.update_columns(value: nil, value_in_business_hours: nil)
# rubocop:enable Rails/SkipsModelValidations
expect do
described_class.backfill_date(account, date)
end.not_to raise_error
rollup = ReportingEventsRollup.find_by!(
account_id: account.id,
date: date,
dimension_type: 'account',
dimension_id: account.id,
metric: 'first_response'
)
expect(rollup.count).to eq(1)
expect(rollup.sum_value).to eq(0)
expect(rollup.sum_value_business_hours).to eq(0)
end
it 'aggregates grouped rows without instantiating reporting events' do
second_user = create(:user, account: account)
second_inbox = create(:inbox, account: account)
second_conversation = create(:conversation, account: account, inbox: second_inbox, assignee: second_user)
create_backfill_event(name: 'first_response', value: 100, value_in_business_hours: 60, user: user,
inbox: inbox, conversation: conversation, created_at: Time.utc(2026, 2, 11, 14))
create_backfill_event(name: 'first_response', value: 40, value_in_business_hours: 20, user: user,
inbox: inbox, conversation: conversation, created_at: Time.utc(2026, 2, 11, 15))
create_backfill_event(name: 'conversation_resolved', value: 200, value_in_business_hours: 80, user: second_user,
inbox: second_inbox, conversation: second_conversation, created_at: Time.utc(2026, 2, 11, 16))
create_backfill_event(name: 'reply_time', value: 500, value_in_business_hours: 300, user: user,
inbox: inbox, conversation: conversation, created_at: Time.utc(2026, 2, 12, 5))
reporting_event_instantiations = count_reporting_event_instantiations do
described_class.backfill_date(account, date)
end
expect(reporting_event_instantiations).to eq(0)
first_response_rollup = find_rollup('agent', user.id, 'first_response')
expect(first_response_rollup.count).to eq(2)
expect(first_response_rollup.sum_value).to eq(140)
expect(first_response_rollup.sum_value_business_hours).to eq(80)
resolution_time_rollup = find_rollup('agent', second_user.id, 'resolution_time')
expect(resolution_time_rollup.count).to eq(1)
expect(resolution_time_rollup.sum_value).to eq(200)
expect(resolution_time_rollup.sum_value_business_hours).to eq(80)
end
def create_backfill_event(**attributes)
create(
:reporting_event,
account: account,
**attributes
)
end
def find_rollup(dimension_type, dimension_id, metric)
ReportingEventsRollup.find_by!(
account_id: account.id,
date: date,
dimension_type: dimension_type,
dimension_id: dimension_id,
metric: metric
)
end
def count_reporting_event_instantiations(&)
instantiation_count = 0
subscriber = lambda do |_name, _start, _finish, _id, payload|
next unless payload[:class_name] == 'ReportingEvent'
instantiation_count += payload[:record_count]
end
ActiveSupport::Notifications.subscribed(subscriber, 'instantiation.active_record', &)
instantiation_count
end
end
end