Question

I am very new to BigQuery by google

I want to parse time stamp (yyyy/mm/dd:hh:mm:ss) based on the day and the month wish to bucket days into weeks.
I didn’t find any BigQuery function which does this.

Hence, I was wondering if there was a way in which I can write a UDF and then access it in a BigQuery query

Answer 1

There are two questions here, so two answers:

• BigQuery does support UDFs: docs. (It didn’t when I first answered this.)

• Even without UDFs, the date bucketing is still doable. BigQuery has one time parsing function, PARSE_UTC_USEC, which is expecting input in the form YYYY-MM-DD hh:mm:ss. You’ll need to use REGEXP_REPLACE to get your date into the right format. Once you’ve done that, UTC_USEC_TO_WEEK will block things into weeks, and you can group by that. So tying all that together, if your table has a column called timestamp, you could get counts by week via something like

  SELECT week, COUNT(week)
  FROM (SELECT UTC_USEC_TO_WEEK(
                 PARSE_UTC_USEC(
                   REGEXP_REPLACE(
                     timestamp, 
                     r"(\d{4})/(\d{2})/(\d{2}):(\d{2}):(\d{2}):(\d{2})", 
                     r"\1-\2-\3 \4:\5:\6")), 0) AS week
        FROM mytable) 
  GROUP BY week;
  

  Note that the 0 here is the argument for which day of the week to use as the “beginning”; I’ve used Sunday, but for “business”-y things using 1 (i.e. Monday) would likely make more sense.

Just in case you need it, the section on timestamp functions in the docs is helpful.