[Rivet] Different binning of reference data and Rivet output

[Andy Buckley]

Holger Schulz wrote:
> Hi,
> 
> I just noted that running Rivet with the UA5_1989 analysis
> the binning of the resulting histos differs from those of the
> reference histos in most of the cases (9 out of 12 observables).
> The binning difference shows up in the AIDA files already.
> 
> The only obvious differences I could find are the encoding (UTF-8
> for Rivet output and ISO-8859-1 for the ref data) and different
> aida versions as well as different implementations.
> 
> Also the values are stored as "floats" in the case of the Rivet output,
> while the ref data is stored as exponentials.
> 
> I guess none of the above should really be a problem, so I am a bit
> puzzled at the moment. Do you have any idea where to look for
> root of this problem?

Are there any gaps in the binning? Rivet determines bins by removing 
duplicates from the list of bin edges taken from the ref AIDA files, and 
LWH/AIDA force the bins to be consecutive, i.e. the upper edge of bin n 
must be the lower edge of bin n+1. If you want to debug how the ref file 
is translated into LWH bin edges, I find that putting cout statements in 
the appropriate bits of src/Tools/RivetAIDA.cc and src/Analysis.cc works 
best. Good luck!

Andy