Can timestamp be conditional according to data content

[bowesmana]

My csv data contains a number of timestamps. I want the timestamp field to be conditional on the result of another field, say field 12. So if I compare field 12 in each row for "X" then I want to use the timestamp from field 15th and if field 12 contains "Y", I want to use field 10 for the timestamp.

Currently I am splitting my log file into two based on the field 12 value and then defining 2 different sourcetypes, each with a different TIME_PREFIX value in props.conf, but wondered if I could avoid that step and have a single sourcetype.

Following Ayn and Kristian Kolb's comments, here's more detail

These are the two standard lines of data

Course A,_scorm12_ar123,Custom Asset,username1,First,Last,Branch Sales,Group Id1,Group Path,0,2013-11-07,2013-11-07,2,0,,,In Progress,,,,35,14,Direct sales
Course B,_scorm12_cf411,Custom Asset,username2,First,Last,Group name,Group Id2,Group Path2,0,2013-11-16,2013-11-20,3,0,,2013-11-17,Completed,,100.00,100.00,35,62,Indirect sales

The first is a course in progress (field 17) and started on 2013-11-07 (field 11)
The second is a course completed (field 17) and started on 2013-11-16 (field 11) and completed on 2013-11-17 (field 16).

I tried two TIME_PREFIX settings

TIME_PREFIX=^(((\"[^\"]*\",)|([^,]*,)){15}|((\"[^\"]*\",)|([^,]*,)){10})
TIME_PREFIX=^(((\"[^\"]*\",)|([^,]*,)){10}|((\"[^\"]*\",)|([^,]*,)){15})

In the first case, _time is set to the completion date or the 'previous' record's time if the course in In Progress. In the second case, the _time is always set to the start date.

I don't suppose I can look ahead BEYOND the wanted time field to check for the value of field 17 to then extract the time from a field BEFORE...??

I guess if the Completed|In Progress text was before the time field I could include that in the pattern, but the docs for TIME_PREFIX say the time is extracted from the text following the regex match.

[kristian_kolb]

Not too sure how Splunk would treat this, but it might work. The (untested) example below assumes that you have space-separated fields, and that the fields themselves do NOT contain any spaces.

TIME_PREFIX = ^((\S+\s+){11}X\s+\S+\s+\S+\s+|(\S+\s+){9}(?=\S+\s+\S+\s+Y))

Also, depending on the size of the events, you may need to increase the MAX_TIMESTAMP_LOOKAHED.
If this approach does not work, it may be better to use your current strategy, with two different sourcetypes.

For more qualified help, you should consider posting a few sample events.

---

UPDATE:

Course A,_scorm12_ar123,Custom Asset,username1,First,Last,Branch Sales,Group Id1,Group Path,0,2013-11-07,2013-11-07,2,0,,,In Progress,,,,35,14,Direct sales

Course B,_scorm12_cf411,Custom Asset,username2,First,Last,Group name,Group Id2,Group Path2,0,2013-11-16,2013-11-20,3,0,,2013-11-17,Completed,,100.00,100.00,35,62,Indirect sales

Given that I've understood you correctly, the above date fields are the ones you want, based on the Completed/In Progress status.

I would try this with the following regex in TIME_PREFIX (but note that it may not work at all, I haven't tested it);

TIME_PREFIX = ^(([^,]*,){10}(?=([^,]*,){6}In Progress)|([^,]*,){15}(?=[^,]*,Completed))

Good luck,

/K

EDIT: typo
EDIT2: picked the wrong field in the first event - corrected now.

[kristian_kolb]

Did you try the updated TIME_PREFIX reges? It'd be fun to know if it worked or not.

/k

[kristian_kolb]

Well, to answer your question on wheter you can 'look beyond' the timestamp for a field value, yes you can. It's called a positive lookahead, and is seen in my example above.

The second regex (\S+\s+){9}(?=\S+\s+\S+\s+Y) will match a position after 9 sequences (i.e. just before field 10) of non-space-characters-followed-by-space. But only if this is followed by a Y in field 12.

Btw, I don't really understand your regexes, i.e. they dont make sense to me. And I'm quite sure that you can only specify TIME_PREFIX once per stanza in props.conf. One of them will be discarded.

/k

[bowesmana]

Thanks Kristian, see my updated description. I guess I'll have to use my existing strategy, but thanks for the great idea.

[kristian_kolb]

good luck.

[bowesmana]

Thanks, I should be able to map that to my data, so I'll try it in the morning!

[Ayn]

If you can specify a TIME_PREFIX that matches one at a time you can do this. Otherwise the answer is most likely no. But something like

TIME_PREFIX = ...,Y,...,|...,N,...

should work, where you give one regex that matches as far as you want in case the field is "Y" and another one for when the field is "N".

[bowesmana]

Thanks Ayn, see my updated description. I guess I'll have to use my existing strategy, but thanks for the great idea.

[bowesmana]

Thanks, I'll try this out in the morning, late here!

[kristian_kolb]

I'm too slow by a mile, I see. 🙂