[Rivet] Rivet 2.4 and xsec/event counter information

[Andy Buckley]

On 10/11/15 09:02, David Grellscheid wrote:
>
>
> On 09/11/15 21:47, Andy Buckley wrote:
>> On 09/11/15 11:20, Leif Lönnblad wrote:
>>> On 2015-11-09 10:44, David Grellscheid wrote:
>>>> That's a neat idea! I assume you would do this just at the level of the
>>>> HepMC converter?
>>>
>>> Yes. It would have to be mentioned in the specs for HepMC3 that the
>>> event number should be interpreted as the number of events generated so
>>> far, and that zero-weight events may be left out.
>>
>> Which would require a lot of care in usage like ATLAS' generator
>> filters, where events in the HepMC are discarded before writing to disk
>> based on kinematic (etc.) selections... compatibility with that needs to
>> be considered, at least.
>
> Won't that still work? Let's say you originally have events numbered
>
> E 1 ...
> E 452 ...
> E 1642 ...
>
> and you leave out the middle one, you still get the correct "number of
> attempts" from the third event.

Ah, I see. I was somehow thinking that the individual deltas between 
event numbers would be important.

If the max event number is taken to mean the number of attempts, this is 
fine. If it's the difference between the min and max, then it can be 
distorted by filters with discard the first event (or a whole, 
potentially very large) block of events at the start of the run.

Because of this, maybe the min event number should be written into a 
header, or runs always declared to start at 1. I'm not sure if the 
latter would work for experiment procedures, but I *think* it is 
currently the case in ATLAS.

Andy

-- 
Dr Andy Buckley, Lecturer / Royal Society University Research Fellow
Particle Physics Expt Group, University of Glasgow