Store time between records - mysql

Question

I have a table called asset_usages which records the viewing of an asset by a viewer. The relevant fields are
id (int)
asset_id (int)
viewer_type (string)
viewer_id (int)
viewed_at (datetime)
I have a new field i just added called time_between_viewings, which is an int field representing seconds. I want to set this to the time, in seconds, since that asset was last viewed. So, if i had these four records:
+-----+----------+-----------+-------------+---------------------+-----------------------+
| id | asset_id | viewer_id | viewer_type | viewed_at | time_between_viewings |
+-----+----------+-----------+-------------+---------------------+-----------------------+
| 506 | 7342 | 1182 | User | 2009-01-05 11:10:01 | NULL |
| 509 | 7342 | 1182 | User | 2009-01-05 11:12:47 | NULL |
| 514 | 6185 | 1182 | User | 2009-01-05 11:14:28 | NULL |
| 524 | 6185 | 1182 | User | 2009-01-05 11:28:18 | NULL |
| 618 | 1234 | 1182 | User | 2009-01-05 11:29:03 | NULL |
| 729 | 1234 | 1182 | User | 2009-01-05 11:29:01 | NULL |
+-----+----------+-----------+-------------+---------------------+-----------------------+
then time_between_viewings should be set as follows:
+-----+----------+-----------+-------------+---------------------+-----------------------+
| id | asset_id | viewer_id | viewer_type | viewed_at | time_between_viewings |
+-----+----------+-----------+-------------+---------------------+-----------------------+
| 506 | 7342 | 1182 | User | 2009-01-05 11:10:01 | NULL |
| 509 | 7342 | 1182 | User | 2009-01-05 11:12:47 | 166 |
| 514 | 6185 | 1182 | User | 2009-01-05 11:14:28 | NULL |
| 524 | 6185 | 1182 | User | 2009-01-05 11:28:18 | 830 |
| 618 | 1234 | 1182 | User | 2009-01-05 11:29:03 | 2 |
| 729 | 1234 | 1182 | User | 2009-01-05 11:29:01 | NULL |
+-----+----------+-----------+-------------+---------------------+-----------------------+
where 166 and 830 are the time difference between each pair, in seconds.
What would be the sql to populate this field? I can't quite figure it out.
IMPORTANT NOTE: the data is not always inserted into the db in chronological order. Ie, you could have two records A and B, where B has a higher id but A has a later value for viewed_at. So, looking for the first matching record with a lower id would not necesarily give you the previous viewing by the same person - you'll need to examine all the records in the database.
thanks! max
EDIT - stated that time_between_viewings is an int field representing seconds.
EDIT - added a couple of rows as an example of a row with a higher id but earlier timestamp
EDIT - i just realised that i didn't stipulate the question properly. The time_between_viewings should be equal to the time since the asset was last viewed by the same viewer, ie the time between the record and the previous (based on viewed_at) record that has the same asset_id, viewer_id and viewer_type. The example data i gave still holds, but i could have put in some different viewer_id and viewer_type values to flesh the example out a bit.

Answer 1

If would be helpful if you prepared sample table and data inserts.
Read this link to learn why it is so important if you want to get help : http://tkyte.blogspot.com/2005/06/how-to-ask-questions.html
This time I created it for you, click on this link: http://sqlfiddle.com/#!2/9719a/2
And try this query (you will find this query together with sample data under the above link) :
select alias1.*,
timestampdiff( second, previous_viewed_at, viewed_at )
as time_between_viewings
from (
select alias.*,
(
select viewed_at from (
select
( select count(*) from asset_usages y
where x.asset_id = y.asset_id
and y.viewed_at < x.viewed_at
) as rn,
x.*
from asset_usages x
) xyz
where xyz.asset_id = alias.asset_id
and xyz.rn = alias.rn - 1
) previous_viewed_at
from (
select
( select count(*) from asset_usages y
where x.asset_id = y.asset_id
and y.viewed_at < x.viewed_at
) as rn,
x.*
from asset_usages x
) alias
) alias1;

Answer 2

This SELECT statement will give you the right data. You might need to do the update in chunks, though.
You can drop the ORDER BY clause for the UPDATE statement. As written, the derived data doesn't depend on the order of rows in the outer SELECT statement.
select asset_id, viewer_id, viewer_type, viewed_at,
prev_viewed_at, timestampdiff(second, prev_viewed_at, viewed_at) elapsed_sec
from (select asset_id, viewer_id, viewer_type, viewed_at,
(select max(t2.viewed_at)
from Table1 t2
where t2.viewed_at < Table1.viewed_at
and t2.asset_id = Table1.asset_id
and t2.viewer_id = Table1.viewer_id
) prev_viewed_at
from Table1
)t3
order by asset_id, viewer_id, viewed_at;