Last update:
Apr 1, 2000
|
|
|
The CDF Run II Event Data Model
|
Robert Kennedy1,
Dan Amidei2,
Liz Buckley-Geer1,
Paolo Calafiura3,
Chris Green4,
Todd Huffman5,
Jim Kowalkowski1,
Al Lee6,
Kevin McFarland7,
Pasha Murat1,
Simona Rolli8,
Liz Sexton-Kennedy1,
Marge Shapiro3,
Prem Singh9,
Rick Snider1,
Peter Tamburello6
- Fermilab
- University of Michigan
- Lawrence Berkeley National Laboratory
- Purdue University
- Oxford University
- Duke University
- Rockefeller University
- TUFTS University
- University of Illinois
Speaker:
Robert Kennedy
The CDF Event Data Model describes how CDF C++ and F77 software may
access event data in CDF Offline software. The CDF Event Data Model Working
Group has reviewed the Run I Event Data Model which was based on YBOS data
banks stored in a global array. While this model was successful for Run I
analyses, it did not take advantage of the Object-Oriented design of the
CDF Run II Offline software.
In this paper, we describe the CDF Run II Event Data Model. Event
data is passed via an EventRecord amongst user-written software modules
whose execution is coordinated by the AC++/Framework package. C++ classes
meeting a few criteria can have instances stored in the EventRecord.
Objects in the EventRecord are assigned a unique identifier and become
write-locked, key changes in behavior from the old model. Support for
storable links, collections, YBOS banks, and data access by Fortran
software are accommodated. The ROOT Object I/O system is used to read/write
the EventRecord to diskfiles. This model has simplified the effort to make
event data persistent, without requiring an overwhelming effort to retrofit
our large, existing code base written to the old event data model.
Presentation: | Short Paper: |
|