Best practices with saving datetime & timezone info in database when data is dependant on datetime

by

Hugo’s answer is mostly correct, but I’ll add a few key points:

  • When you’re storing the customer’s time zone, do NOT store a numerical offset. As others have pointed out, the offset from UTC is only for a single point in time, and can easily change for DST and for other reasons. Instead, you should store a time zone identifier, preferably an IANA time zone identifier as a string, such as "America/Los_Angeles". Read more in the timezone tag wiki.

  • Your OrderDateTime field should absolutely represent the time in UTC. However, depending on your database platform, you have several choices for how to store this.

    • For example, if using Microsoft SQL Server, a good approach is to store the local time in a datetimeoffset column, which preserves the offset from UTC. Note that any index you create on that column will be based on the UTC equivalent, so you will get good query performance when doing your range query.

    • If using other database platforms, you may instead wish to store the UTC value in a timestamp field. Some databases also have timestamp with time zone, but understand that it doesn’t mean it stores the time zone or offset, it just means that it can do conversions for you implicitly as you store and retrieve values. If you intend to always represent UTC, then often timestamp (without time zone) or just datetime is more appropriate.

  • Since either of the above methods will store a UTC time, you’ll also need to consider how to perform operations that need an index of local time values. For example, you might need to create a daily report, based on the day of the user’s time zone. For that, you’d need to group by the local date. If you try to compute that at query time from your UTC value, you’ll end up scanning the entire table.

    A good approach to deal with this is to create a separate column for the local date (or perhaps even the local datetime depending on your needs, but not a datetimeoffset or timestamp). This could be a completely isolated column that you populate separately, or it could be a computed/calculated column based on your other column. Use this column in an index so you can filter or group by local date.

  • If you go for the computed-column approach, you’ll need to know how to convert between time zones in the database. Some databases have a convert_tz function built-in that understands IANA time zone identifiers.

    If you’re using Microsoft SQL Server, you can use the new AT TIME ZONE function in SQL 2016 and Azure SQL DB, but that only works with Microsoft time zone identifiers. To use IANA time zone identifiers, you’ll need a third party solution, such as my SQL Server Time Zone Support project.

  • At query time, avoid using the BETWEEN statement. It is fully inclusive. It works ok for whole dates, but when you have time involved you’re better off doing a half-open range query, such as:

    ... WHERE OrderDateTime >= @t1 AND OrderDateTime < @t2

    For example, if @t1 were the start of today, @t2 would be the start of tomorrow.

Regarding the scenario discussed in comments where the user’s time zone has changed:

  • If you choose to calculate the local date in the database, the only scenario you need to worry about is if a location or business switches time zones without a “zone split” occurring. A zone split is when a new time zone identifier is introduced which covers the area that changed, including their old and new rules.

    For example, the latest zone added to the IANA tzdb at the time of writing this is America/Punta_Arenas, which was a zone split when the southern part of Chile decided to stay at UTC-3 when the rest of Chile (America/Santiago) went back to UTC-4 at the end of DST.

    However, if a minor locality on the border of two time zones decides to change which side they follow, and a zone split wasn’t warranted, then you’d potentially be using the rules of their new time zone against their old data.

  • If you store the local date separately (computed in the application, not the DB), then you’ll have no problems. The user changes their time zone to the new one, all old data is still intact, and new data is stored with the new time zone.

More Related Contents:

  1. Daylight saving time and time zone best practices [closed]
  2. What is the difference between UTC and GMT?
  3. DateTime vs DateTimeOffset
  4. Convert UTC date time to local date time
  5. How do I get the UTC time of “midnight” for a given timezone?
  6. pytz localize vs datetime replace
  7. Is it always a good idea to store time in UTC or is this the case where storing in local time is better?
  8. Parsing of Ordered Timestamps in Local Time (to UTC) While Observing Daylight Saving Time
  9. php convert datetime to UTC
  10. .NET DateTime.Now returns incorrect time when time zone is changed
  11. Convert UTC DateTime to another Time Zone
  12. Handling dates when we switch to daylight savings time and back
  13. Python: Figure out local timezone
  14. How to tell if a timezone observes daylight saving at any time of the year?
  15. How to detect Ambiguous and Invalid DateTime in PHP?
  16. How to get timezone, Language and County Id in flutter by the location of device in flutter?
  17. What are the “standard” timezone abbreviations?
  18. How to convert local time string to UTC?
  19. Creating a DateTime in a specific Time Zone in c#
  20. How to store date/time and timestamps in UTC time zone with JPA and Hibernate
  21. get UTC time in PHP
  22. python – datetime with timezone to epoch
  23. Using DateTime properties in Code-First Entity Framework and SQL Server
  24. Where’s the DateTime ‘Z’ format specifier?
  25. Python time to age, part 2: timezones [duplicate]
  26. How precise is the internal clock of a modern PC?
  27. Get UTC offset from time zone name in python
  28. Twig date difference
  29. How to convert UTC datetime to another timezone?
  30. How to control appearance of ‘:’ in time zone offset when parsing/formatting Datetime

Leave a Comment