Run II Physics Analysis Software Requirements Checklist : ROOT

DATA ACCESS

• Access rates (online):  ????
• upports shared memory option
• supports sockets
• Access rates (offline) : 5-6 MBytes/sec demonstrated  on  Linux/Pentium w/o serious optimization
• Serial vs random access: supports both modes
• Granularity of access:        leaf of the tree
• Foreign Input and Output Formats:
• h2root (PAW ntuple --> ROOT tree)
• g2hroot: conversion of GEANT3 hit banks into ROOT format
• Specialized output formats:  several standard objects:
• ROOT Tree (ntuple),
• ROOT histogram
• ROOT containers (TClonesArray, for example)

DATA ANALYSIS

Scripting Language: C/C++ interpreted language (CINT)
 
• full featured scripting language :  YES
• available,  its major weakness is the opposite side of its strongest advantage C++ is much more difficult to interpret than  other languages like Python or  Java,  so not all the features of C++ are supported at the interactive level
• analysis tool's object model :
• data are stored in the external files in ROOT Trees,
•  primary interface: command line (interactive scripts, shared libraries)
•  GUI interface / GUI builder
• extract data from events :
•  TTree::GetEvent(i)
• express complex mathematical expressions : YES
•  scripting language is C+(+), so all  the math  is available (TMath class)
• debugging facilities: YES
•  CINT  supports primitive level of debugging, one can step through the interpreted code
•  insert cout/printf  statements
• interface the scripting language to dynamically linked compiled high level languages
•  TSystem::Load(char* filename)   loads in an  arbitrary shared library. An interface should be provided though. Automatic tool for generating the interfaces (rootcint) exists
User Control:
 
• control functions: YES
• interface to operating system
• Mathematical operations: YES
• Results of analysis available to users: YES
• ROOT file
• .ps, .eps, .gif, .C (ROOT macro)
• command line recall and interactive command line editing: YES

Data Selection:
 

• program selection criteria using extracted data: YES
• display selection criteria as text : YES
•  normally stored in CINT scripts
 

Input/Output:

 
•  support its own object I/O format: YES
• non-templated object format, objects are supposed to be derived from the base class TObject
• allow its own format object files to be read or written from compiled programs: YES
• read or write object files in foreign formats: YES
• write selected event objects to one or more output streams: YES
• supported, several output files may be opened at a time:

     TFile f1("a.root","RECREATE")
     TFile f2("b.root","RECREATE")
• object definition language and/or be able to define new object formats programmatically: YES

- interpreted language it C++
• read events in one format, convert and write them out in a different format: YES
• virtual streaming: YES

 

Numeric and Mathematical Functionality:

 
• accurate and precise numerical functionality, including double precision: YES

- all  math operations implemented in double precision
• Analysis capabilities applied to fetched data as well as subsequent renditions
• Functions operating on multiple data sets
• fit, parameterize, and calculate statistical quantities from data : YES

- uses  MINUIT for fitting (the package tested and polished by the HEP community
    over the decades
- all the functionality of HBOOK package including histogram fitting is already available
• user control of fitting algorithms: YES

- MINUIT  has an extensive interactive interface
- source codes are available

Offline Compatibility:
 

•  tailor the sequence of mathematical operations : YES
• ability to include external software in their analysis:  YES
• also export CINT scripts into offline code
• functionality of the analysis package linked into user defined code: YES

Prototyping:
 

• prototyping of simple versions which can later be expanded upon: YES
• Prototyped sequences contain the full interface of an arbitrarily complex version: YES
•  prototyped methods could be immediately used  in the offline code

DATA PRESENTATION

USABILITY

• Batch vs. interactive processing:  YES
• wnteractive session: >root
•  batch mode                :  > root -b  script
• Sharing data structures: YES
• Shared access by several clients: YES
• Parallel processing (using distinct data streams): YES

• Debugging and profiling:
• debugging of the scripts
• profiling of the compiled code
• Modularity (user code): YES
• Modularity (system code): YES
• Access to source code: YES
• http://root.cern.ch
• Robustness: to be improved
• Web based documentation: YES
• http://root.cern.ch
• Use of standards:
• Portability: YES
• all UNIX flavours
• Windows (95/NT)
• Mac OS
• Scalability: YES
•  based on the consideration of scalability CDF has decided to write event data in ROOT format
• Performance: better than HBOOK
• User Friendliness: to be improved
• this is our first experience with large scale C++ product