Weeks are special because they don't fit evenly into months, years or anything bigger. Weeks and days use a slightly different approach. Quarters is effectively the same as months (since DATE_PART gives us quarters as an integer from 1-4). Subtract the two and add in the year expressed in months to handle rollover.Ģ021/2 - 2020/11 would be (2 - 11) + 1 * 12 = 3 Months – just use DATE_PART to get the month as an integer (1-12). It computes weeks and below by truncating. It computes year, quarter, and month boundaries by subtracting integers. This isn't consistent across databases – Redshift uses Sunday, while in Snowflake it's configurable. Similarly DateDiff('week', '05-02-2021', '05-03-2021') is 1, because 5/02/21 is a Sunday and 5/03/21 is a Monday Note that Postgres uses ISO 8601 week numbering, so weeks will always start on Mondays. So what does counting date boundaries mean? It's best illustrated with an exampleĭATEDIFF('year', '12-31-2020', '01-01-2021') returns 1 because even though the two dates are a day apart, they've crossed the year boundary. It will always return an integer, so it's very useful for grouping date differences together. This function take a time unit and two dates, and counts the number of date boundaries crossed between them. This gist creates a function in Postgres that implements the DATEDIFF function found in Snowflake, BigQuery, and Redshift. I'll jump straight to the code for those who like to see the answer first, and further down explain how it works ![]() Why is this even a function? Why not just use date_part? Because time rolls over: the last month of the year is 12 and the first is 1, so naively using date part would give us -11. A great blog by sqlines suggests an implementation, but I found it didn't quite match the functionality of most data warehouses. If you want to use Postgres as a data warehouse you'll probably want it. ![]() For operational data it's probably not often used, but if you do analytical queries it can be pretty helpful. Unfortunately Postgres simply doesn't have it. This function can be used to bucket times together, like when doing a cohort analysis. Data warehouses like Redshift and Snowflake have a super useful DATEDIFF function – given two timestamps and a date part (hour, year, week, etc) it'll return how far apart they are.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |