Rails - How to group dates by time when they're in different timezones?

Question

I have a series of Appointments where the date and time is stored under start_date as DateTime. I'd like to categorize them as starting in the Morning, Daytime, or Evening. I created an array of hashes with labels and ranges I used for a SQL statement where I convert the start_date records into seconds using CAST(EXTRACT (EPOCH FROM start_time) AS INT)

TIME_RANGES = [
    {label: "Morning", min: 0, max: 32399},
    {label: "Daytime", min: 32400, max: 61199},
    {label: "Evening", min: 61200, max: 86399}
]

cases = TIME_RANGES.map do |r|
    "when CAST (EXTRACT (EPOCH FROM start_date) AS INT) % 86400 between '#{r[:min]}' and '#{r[:max]}' then '#{r[:label]}'"
end

time_ranges = Appointment.select("count(*) as n, case #{time_cases.join(' ')} end as time_range").group('time_range')

This takes the number of seconds in a day (86400) and labels the appointments based on the modulo of the start_date with 86400. However, a number of appointments take place in different timezones, but they're all stored as UTC. So an appointment at 08:00 AM EST is equivalent to one at 07:00 AM CST, but both are stored internally as 12:00 PM UTC. This would cause the appointments to be incorrectly labelled as "Daytime" when they're intended to be "Morning" (from the perspective of the User booking it).

Ideally, I would like some way to convert the start_date based on the User's timezone to make it look like it occurred in UTC. So I would want a 12:00 PM EST appointment to be labelled as if it were a 12:00 PM UTC appointment instead of 04:00 PM UTC. More specifically, I would like to subtract 14400 seconds from the converted start_date before performing the modulo.

I can join Appoinments to Users, which contains the User's timezone. How can I incorporate this information into my query above, so that a modified start_date is used for each record, depending on the User's timezone in that same record?

I know I could accomplish this with a loop of each timezone and adding/substracting a specific amount of seconds in each loop, then combining the results of all the loops, but I was wondering if there was a way to do it in one query.

Answer 1

Per my comment, I am assuming we have three tables: appointments, users, and preferences. In appointments we have start_date and user_id. In users we have preference_id. In preferences we have some column that names the time zone, so I'll call that tz_name.

Note: Postgres timezone functions are messy. I would highly recommend you read up on them. This excellent article is a good place to start.

It is possible to use pure SQL to generate the time ranges and return a grouped result. A pure SQL solution would be best if you need to label and group many records (thousands or more) at a time.

Assuming you are working with 1000 records or fewer at a time, you'll probably want to use Rails scopes, as this will give you an ActiveRecord result. Then you'll do your grouping using Ruby's Array methods.

That solution would look something like this:

# app/user.rb

class User < ApplicationRecord
  belongs_to :preference
  has_many :appointments
end

# app/appointment.rb

class Appointment < ApplicationRecord
  belongs_to :user

  scope :with_seconds, lambda {
    joins(user: :preference)
      .select('appointments.*, extract(epoch from timezone(tz_name, start_date::timestamptz)::time)::int as seconds')
  }

  # This method is optional. If it is excluded, calling #seconds 
  # on an appointment instance will raise an error if the 
  # .with_seconds scope has not been applied.
  def seconds
    attributes['seconds']
  end

  def time_range
    return nil unless seconds.present?

    if seconds.between?(0, 32_399)
      'Morning'
    elsif seconds.between?(32_400, 61_199)
      'Daytime'
    else
      'Evening'
    end
  end
end

The select portion of the scope probably deserves some explanation.

start_date::timestamptz: Take the start_date, which is stored as a Postgres timestamp, and convert it into a Postgres timestamp with time zone in the time zone of the Postgres server (presumably UTC).

timezone(tz_name, start_date::timestamptz): Convert the timestamp with time zone back into a timestamp type in the local time of the tz_name time zone.

timezone(tz_name, start_date::timestamptz)::time: Drop the date and keep the time component.

Then we extract epoch from that time component, which converts it into seconds.

Finally we convert the result to an integer to avoid anything falling through the cracks when we determine the time range.

Now you can do:

Appointment.all.with_seconds.group_by(&:time_range)

or

user = User.first
user.appointments.with_seconds.group_by(&:time_range)

For a pure SQL solution that will return ids grouped under the three time ranges, add this method to your Appointment model:

  def self.grouped_by_time_range
    current_scope = with_seconds.to_sql
    query = <<~SQL
      with converted_seconds as (#{current_scope})
      select array_agg(id) as ids, case when seconds < 32400 then 'Morning'
                                        when seconds < 61200 then 'Daytime'
                                        else 'Evening' end as time_range
      from converted_seconds
      group by time_range
    SQL

    result = ActiveRecord::Base.connection.execute(query.squish)
    result.to_a
  end

Answer 2

If you don't need strictly SQL based solution you might use Ruby's select method to extract this appointments as in this example: (I am assuming there is some kind of tz_name field in appointment model which holds timezone name)

morning_appointments = Appointment.all.select do |a|
  a.start_date.in_time_zone(a.tz_name).hour > 6 && a.start_date.in_time_zone(a.tz_name).hour < 20
end

Edit: Thanks @moveson for pointing out my mistake, I changed solution a bit.