Re: BrValueObject in CVS

From: Bjorn H Samset (bjornhs@rcf.rhic.bnl.gov)
Date: Tue Nov 21 2000 - 03:46:47 EST

  • Next message: I. G. Bearden: "Re: BrValueObject in CVS"

    On Mon, 20 Nov 2000, Christian Holm Christensen wrote:
    
    > Damn. I guess that's the price you've gotta pay for taking a short
    > vacation.
    >
    > I DO NOT LIKE THE IDEA OF BrValueObject!!!!!  
    
    Hi Christian et al. Vacation? Yes, I recall that word ;-)
    
    I agree with most or all of your sentiments - that's why I was sceptical
    of adding it in the first place. I never advocated a heavy use of this
    object - it was only made as a handy tool for an application-program. I
    have a few comments in the opposite direction, though:
    
    Since I am not a programming-expert, I see two reasons for accepting an
    object if this size (which I agree is huge for the purpouse)
    1) It is indeed a BrDataObject. This may be, as you comment, an
    unneceseary step taken to "hide" ROOT, but the way BRAT is structured
    right now BrDataObjects are a lot easier to handle - at least for the
    non-expert.
    2) Using this object is an easy way of reucing file-size in other areas,
    since I can keep one or two variables and then throw away large chunks of
    raw data from the analyzed files. This in my view makes the rare use of a
    large object more than acceptable, since my computer is allready more than
    crammed with 100+Mb reco-files.
    
    > 2) Specialised classes is prefered. If you really need a summary
    >    object of some sort, then you should make such a class.
    
    I agree this is best, but I fear we will end up with either classes that
    are huge due to keeping too many variables of infinately many small
    "utility"-classes for moving this and that information. 
    
    > 4) There's a slightly odd but very simple way to store a simple number
    >    in ROOT. Take a look at what Rene had to say on this:
    >    
    >       http://root.cern.ch/root/roottalk/roottalk00/2311.html
    
    Yup - this is more or less what I wanted. But:
    
    >    His suggestion is perfect for debugging stuff and so on, but
    >    probably not for production
    
    Agreed. It's also a bit non-intuitive for the unexperienced ROOTed BRAT...
    
    I still advocate having some capabitily within BRAT for handily moving
    small pieces of data, and I welcome a discussion as to how we should do
    this. For instance, referring to the example I gave in an earlier mail,
    storing the number of clusters in a TPC is not easily fittable in any of
    the classes. I could of course store the entire cluster-table, but that
    was what I was trying to aviod in the first place.
    
    Anyway - I agree that BrValueObject is not the best implementation 
    of this. Learning from my previous mistake ;-) though, I won't remove
    it right away, but wait a bit to allow for further comments. Please
    respond, good people - I am sure there are quite a lot of opinions on this
    ;-)
    
    ------------------------------------------------
    Bjorn H. Samset
    Master-student in Heavy Ion physics
    Mob: +47 92 05 19 98  Office: +47 22 85 77 62  
    Adr: Kri 2A709 Sognsveien 218 0864 Oslo
    



    This archive was generated by hypermail 2b29 : Tue Nov 21 2000 - 03:47:56 EST