#164 closed defect (fixed)
HepMC output to file differs from internal Rivet
| Reported by: | hoeth | Owned by: | hoeth |
|---|---|---|---|
| Priority: | major | Milestone: | |
| Component: | Unknown | Version: | 0.trunk |
| Keywords: | Cc: |
Description
When writing HepMC output to a file, the event weights are set to 1 (at least for unweighted events), while internal Rivet analyses see the cross-section as weights. Upshot is: Running Rivet internal or external yields different results.
Attachments (0)
Change History (7)
comment:1 Changed 13 years ago by
comment:2 Changed 13 years ago by
It breaks debugging. That's how Korinna and I found it. Trouble with a Rivet analysis, decided to produce a HepMC file to speed up debugging, and the aida files coming from Sherpa and from stand-alone Rivet were different -- with the same events. I consider such inconsistencies a bug, not a feature.
comment:3 Changed 13 years ago by
If the aida files are really different then I would be worried. But if the histogramming works correctly, a constant weight should not make a difference (e.g. because histograms are finalised by scale(xs/sumOfWeights)). So I'd like to understand this better, if it really is a problem.
comment:4 Changed 13 years ago by
If everything is scaled by 1/sumOfWeights(), the output is the same. If not, well, then not. Whether or not you think it is a problem that we present two equivalent ways of running a Rivet analysis to the user that then turn out not to be equivalent, is up to you. I think it is.
comment:5 Changed 13 years ago by
| Owner: | changed from support@sherpa-mc.de to hoeth |
|---|
I still don't understand whether this is an academic discussion or whether there really is a problem that you think should be solved. Is there any use case for using a histogram without scaling by 1/sumOfWeights()??? That way your histogram will depend on the number of events that you feed in (beyond just improvements in the statistics), so I don't see how that would make sense.
Anyway, I don't mind either way (since in my view they are 100% equivalent), so if you want to scale the weight in the Rivet interface similarly as we do in the HepMC interface (for unrelated obscure reasons), then feel free to do it.
comment:6 Changed 13 years ago by
| Resolution: | → fixed |
|---|---|
| Status: | new → closed |
Closing this ticket for now, as the issue seems to be resolved. Please reopen if further discussion becomes necessary.
In my opinion this is not a bug, or do you have any case where this will make a difference or even break something?