[Rivet] [Rivet-svn] r1860 - trunk/src/Analyses

[Andy Buckley]

Frank Siegert wrote:
> Andy Buckley, Wednesday 07 October 2009:
>> Frank Siegert wrote:
>>> Cool, that's a nice addition to our LEP analyses.
>>> Have you considered doing something similar as I have done (recently)
>>> for the JADE_OPAL analysis, such that actually only the E_cms
>>> appropriate histograms get initialised and written out? Otherwise
>>> lots of empty histograms get written to the aida file, wouldn't they?
>>> (and then also get plotted)
>> I'll take a look at what you've done: it's certainly not so nice to
>> have 104 plots from each run, 76 of which are empty, but fortunately
>> they can be selectively plotted with e.g. "make-plots *-y02.dat"
> 
> If you do all the plotting "manually", then yes.
> 
>> Directionless mental wander:
>>    I'm still not so fond of the putting histo bookings in the analyze
>> loop, since it's a bit non-intuitive: maybe we could instead augment
>> the tests on the first event to make the beams and E_cms available via
>> the AnalysisHandler/Analysis classes? Or is there a reason why we might
>> expect a single event stream to contain events at several CMS energies?
>> While I quite like the idea of "lazy booking", it concerns me that if
>> Analysis::bookHistogram1D() is silent about attempted duplicate
>> bookings (which would be the way to make this neat), then
>> copy-and-paste errors in the booking code will be harder to find...
> 
> I agree, but didn't want to make such a big change: After all you would 
> have to scan the first event without analyzing it yet and then come back 
> to it. But if we have more and more analyses where this would be useful (I 
> also recall that there is a 630/1800 GeV analysis?) this might be 
> worthwhile.

As you'll have noticed, this functionality is now being used: the first 
event is read before the analyses are initialised and used to set a 
sqrt(s) and beam pair for the run. These can be accessed in the analysis 
classes with the sqrtS() and beams() methods. The run class will exit 
with an error if any event fails to have the expected beams and sqrt(s).

I really like this, because most multi-energy analyses do the same thing 
at each energy: now the init() phase can be used to associate the 
appropriate histo path with a single histo object and usually all the 
energy comparisons can be removed from the execute() and finalize() 
loops. Simplification is good!

I'll reply to Frank's email about the API for this in a moment... I 
agree that we should probably move it out of the Run class, since it 
still applies when Rivet is not reading its HepMC events from a stream.

Andy

-- 
Dr Andy Buckley
SUPA Advanced Research Fellow
Particle Physics Experiment Group, University of Edinburgh