numeric field with thousand separator - stats doesn't interpret as numeric, timechart does

[grundsch]

I stumbled on a very strange behavior of stats versus timechart when trying to interpret an extracted numerical field.
The field in question contains "," as 1000 separators, so stats consider it as text, and will not calculate any sum or avg.
On the other hand, timechart is happy to get rid of the "," and makes a nice sum.

here's a quick demo of the effect:

* |head 1| eval in="0;1,000;2,000;4,000"| makemv delim=";" in| mvexpand in | table _time in | streamstats sum(in)

_time                       in      sum(in)
1   12/09/2012 16:18:23.000 0       0
2   12/09/2012 16:18:23.000 1,000   0
3   12/09/2012 16:18:23.000 2,000   0
4   12/09/2012 16:18:23.000 4,000   0

and now with timechart:

* |head 1| eval in="0;1,000;2,000;4,000"| makemv delim=";" in| mvexpand in | table _time in | timechart bins=2 sum(in)

_time                       sum(in)
1   12/09/2012 00:00:00.000 7000

Funny, isn't it?

• Has anyone an explanation?
• Is there a way to tell splunk to interpret a field as numeric and get rid of 1000 separators? (tonumber() is not clever enough)

[Ayn · Accepted solution]

Remove the commas before piping to the stats command by using convert's rmcomma function.

rmcomma()
Syntax: rmcomma(<wc-field>) 
Description: Removes all commas from value, e.g. '1,000,000.00' -> '1000000.00' 

http://docs.splunk.com/Documentation/Splunk/latest/SearchReference/Convert

[grundsch]

rock'n roll, it's working! also with auto:
* |head 1| eval in="0;1,000;2,000;4,000"| makemv delim=";" in| mvexpand in | table _time in |convert auto(in) | streamstats sum(in)

I suppose that timechart is applying auto() by default.
Thanks!