可以将文章内容翻译成中文,广告屏蔽插件可能会导致该功能失效(如失效,请关闭广告屏蔽插件后再试):
问题:
I have a text field "presence_changed_at" with text values i.e. '2014/12/17 08:05:28 +0000
. I need to convert this into timestamp. In postgreSQL there is function TO_TIMESTAMP(), however in redshift this does not seem to be supported. I can get the date without time by
TO_DATE("presence_changed_at",'YYYY/MM/DD HH24:MI:SS')
which produces
2014-12-12
but i can't find any way to get TIMESTAMP format.
Thanks in advance for solving this
回答1:
try convert function:
convert(timestamp,presence_changed_at)
回答2:
It's surprisingly horrible to get a datetime from a unixtime in redshift without a subselect. However, you can do this:
select timestamptz 'epoch' + YOURTIME * interval '1 second'
回答3:
Please use TO_Timestamp("presence_changed_at",'YYYY/MM/DD HH24:MI:SS')
to get the desired output
回答4:
You can cast it to a timestamp. This supports quite a few reasonable formats without having to specify one.
select '2014/12/17 08:05:28 +0000'::timestamp; timestamp --------------------- 2014-12-17 08:05:28 select '2014-12-02T05:00:00'::timestamp; timestamp --------------------- 2014-12-02 05:00:00 select '2014-12-02T05:00:00PM'::timestamp; timestamp --------------------- 2014-12-02 17:00:00
回答5:
As of today, to_timestamp is supported by redshift. So, your SQL query should work. Check http://docs.aws.amazon.com/redshift/latest/dg/r_TO_TIMESTAMP.html
回答6:
I recently worked on a database where a date & time variable was stored as text in a VARCHAR type, in multiple different formats (don't ask...), and had to convert it to a TIMESTAMP type. Since there is no TO_TIMESTAMP()
function in Redshift, I used the trick suggested by Yiyu Jia on his [blog][1]. In a nutshell, the trick is to
- use TO_DATE() to get the date converted
- append the time portion of the input text to the above
- CAST the resulting string to a TIMESTAMP
For example here's the snippet to deal with a field named myDate with dates in either of the following formats
- "Feb 8 2013 10:06PM"
- "25/09/2007 16:21:00"
It is rather heavy, but works. A regex test is used to test if the date corresponds to the format handled on a given line. (that is only necessary when dealing with multiple possible formats)
The 'Feb 0 2013' case is a bit more complicated because I remove the time portion of the text, before submitting it to TO_DATE(), and because another regex is used to extract the time portion that is appended (as opposed the simpler SUBSTRING() used for the same purpose, in the other case).
... , CASE -- Special date indicating "date not available": replaced by NULL WHEN myDate = '31/12/9999 23:59:59' OR myDate = 'Dec 31 9999 11:59PM' THEN NULL -- 'Feb 8 2013 10:06PM' case WHEN myDate ~ '^[JFMASOND][a-z]{2}' THEN CAST(TO_DATE(REGEXP_REPLACE(myDate , '\\s[0-9]{1,2}:[0-9]{2}[AP]M$', ''), 'Mon FMDD YYYY') || REGEXP_REPLACE(myDate , '[JFMASOND][a-z]{2}\\s+[0-9]{1,2}\\s+[0-9]{4}\\s+', ' ') AS TIMESTAMP) -- '25/09/2007 16:21:00' case WHEN myDate ~ '^[0-9]{2}/[0-9]{2}/[0-9]{4} ' THEN CAST(TO_DATE(myDate , 'DD/MM/YYYY HH24:MI:SS') || SUBSTRING(myDate FROM 11) AS TIMESTAMP) ELSE NULL END AS MyNiceTimeStamp, ... [1]: http://yiyujia.blogspot.com/2014/04/redshift-convert-integer-to-timestamp.html
回答7: