Hi,
talking a bit with Concezio resulted in a "maybe" new idea about timestamp
killing.
The idea is:
move the scanning of the timestamp at Ibu level and doing it once for
all...
It is possible the following?
a) create the list of to_be_killed ts
b) the the list in CompTool or Ibu
c) when dumping the hbook/root add a flag (0/1) corresponding to the
"to_be or not_to_be killed" for each event
d) at anaRecoil level one only need to read the ntuple variable and
discriminate the event on this basis
This avoid the problem of comparing each event at anaRecoil level with the
list of to_be_killed and speedup the analysis.
We only throw away the bad events once for all.
Let me know your comments on this idea.
I don't know if what I am proposing is easy to do, but we are doing
something similar in CompTool with the integrated purity so maybe someone
can teach me on how to do it in this case...
Cheers,
Alessio
______________________________________________________
Alessio Sarti
Universita' & I.N.F.N. Ferrara
tel +39-0532-781928 Ferrara
"Quod non fecerunt barbari, fecerunt Berlusconi"
"Che il bianco sia bianco, e che il nero sia nero
che uno e uno fanno due e che la scienza dice il vero....
DIPENDE !"
On Wed, 6 Feb 2002, Urs Langenegger wrote:
>
> Hoi Alessio,
>
> > upper and lower Id are not the only thing that I've encountered dealing
> > with timestamp:
> > partition and platform number, run Number, a char that is computed as a
> > check sum on the other quantities....
>
> Yes, but the platform is always the same (the BABAR detector), the
> checksum does not add any information per se (it's just a checksum),
> and the runnumber is not needed for event identification. In fact, we
> (babar) are lucky that we even have runnumbers (there were real plans
> to get rid of this 'ancient' concept). Oh yes, the partition is also
> always the same, because we need all of the detector to do the
> measurement (since I have not yet convinced the rest of you to do the
> measurement only with electrons ;-)
>
> So all you need to decide whether an event is unique is to compare
> upper and lower.
>
> Convinced?
>
> Cheers,
> --U.
>
|