Welcome to OGeek Q&A Community for programmer and developer-Open, Learning and Share
Welcome To Ask or Share your Answers For Others

Categories

0 votes
676 views
in Technique[技术] by (71.8m points)

sql - What is the fastest way to truncate timestamps to 5 minutes in Postgres?

Postgres can round (truncate) timestamps using the date_trunc function, like this:

date_trunc('hour', val)
date_trunc('minute', val)

I'm looking for a way to truncate a timestamp to the nearest 5-minute boundary so, for example, 14:26:57 becomes 14:25:00. The straightforward way to do it is like this:

date_trunc('hour', val) + date_part('minute', val)::int / 5 * interval '5 min'

Since this is a performance-critical part of the query, I'm wondering whether this is the fastest solution, or whether there's some shortcut (compatible with Postgres 8.1+) that I've overlooked.

See Question&Answers more detail:os

与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…
Welcome To Ask or Share your Answers For Others

1 Reply

0 votes
by (71.8m points)

I was wondering the same thing. I found two alternative ways of doing this, but the one you suggested was faster.

I informally benchmarked against one of our larger tables. I limited the query to the first 4 million rows. I alternated between the two queries in order to avoid giving one a unfair advantage due to db caching.


Going through epoch/unix time

SELECT to_timestamp(
    floor(EXTRACT(epoch FROM ht.time) / EXTRACT(epoch FROM interval '5 min'))
    * EXTRACT(epoch FROM interval '5 min')
) FROM huge_table AS ht LIMIT 4000000

(Note this produces timestamptzeven if you used a time zone unaware datatype)

Results

  • Run 1: 39.368 seconds
  • Run 3: 39.526 seconds
  • Run 5: 39.883 seconds

Using date_trunc and date_part

SELECT 
    date_trunc('hour', ht.time) 
    + date_part('minute', ht.time)::int / 5 * interval '5 min'
FROM huge_table AS ht LIMIT 4000000

Results

  • Run 2: 34.189 seconds
  • Run 4: 37.028 seconds
  • Run 6: 32.397 seconds

System

  • DB version: PostgreSQL 9.6.2 on x86_64-pc-linux-gnu, compiled by gcc (Ubuntu 4.8.2-19ubuntu1) 4.8.2, 64-bit
  • Cores: Intel? Xeon?, E5-1650v2, Hexa-Core
  • RAM: 64 GB, DDR3 ECC RAM

Conclusion

Your version seems to be faster. But not fast enough for my specific use case. The advantage of not having to specify the hour makes the epoch version more versatile and produces simpler parameterization in client side code. It handles 2 hour intervals just as well as 5 minute intervals without having to bump the date_trunc time unit argument up. On a end note, I wish this time unit argument was changed to a time interval argument instead.


与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…
OGeek|极客中国-欢迎来到极客的世界,一个免费开放的程序员编程交流平台!开放,进步,分享!让技术改变生活,让极客改变未来! Welcome to OGeek Q&A Community for programmer and developer-Open, Learning and Share
Click Here to Ask a Question

1.4m articles

1.4m replys

5 comments

57.0k users

...