AIDA Abstract Interfaces for Data Analysis

1
AIDAAbstract Interfaces for Data Analysis
Andreas Pfeiffer CERN IT/API andreas.pfeiffer_at_cer
n.ch
2
Outline

• What is AIDA
• History/Collaboration/Documentation
• Some Details
• Examples
• Ongoing work
• Summary

3
What is AIDA

• Abstract Interfaces for Data Analysis (AIDA)
• The goals of the AIDA project are to define
  abstract interfaces for common physics analysis
  objects, such as histograms, ntuples, fitters, IO
  etc.The adoption of these interfaces should make
  it easier for developers and users to select to
  use different tools without having to learn new
  interfaces or change their code. In addition it
  should be possible to exchange data (objects)
  between AIDA compliant applications.

4
Motivation

• Advantages
• The user needs to learn only one set of
  interfaces
• Same user code can be used with different
  AIDA-compliant analysis applications
• Pool experience of different developer teams
• LHC, OpenScientist, JAS
• Different analysis tools can exchange analysis
  objects
• same storage format, use functionality from other
  tools
• Two versions of AIDA interfaces
• One for C
• One for Java
• As identical as possible

5
Abstract Interfaces

• Abstract Interfaces
• only pure virtual methods, inheritance only from
  other A.I.
• components use other components only through
  their A.I.
• defines a kind of a protocol for a component
• Maximize flexibility and re-use of packages
• allow each component to develop independently
• re-use of existing packages to implement
  components reduces start-up time significantly
• De-couple implementation of a package from its use

6
AIDA Example

• Use same code with any AIDA-compliant analysis
  tool.

7
Architectural issue Components (I)

• Identify components by functionality
• Define protocol using Abstract Interfaces
• Emphasize separation of different aspects for
  each component
• Example Histogram
• statistical entity (density distribution of a
  physics quantity)
• view of a collection of data points (which can
  be a density distribution but also a detector
  efficiency curve)
• command to manipulate/store/plot/fit/...
• Users view is different from implementors
  (developers) view
• separate Abstract Interfaces for both aspects

8
Use of Components withAbstract Interfaces

• User Code uses only Interface classes
• IHistogram1D hist histoFactory-gt
  create1D(track quality, 100, 0., 10.)
• Actual implementations are selected at run-time
• loading of shared libraries
• No change at all to user code but keep freedom
  to choose implementation

Histo- Impl. 2
9
Across the languages

• JAida C access to Java libs
• using C proxies implementing the C Abstract
  Interfaces to the Java interfaces

CUserCode
AIDA-IF C
AIDA-IF Java
JAida
Java Lib
10
XML standards

• Started with 1D and 2D Histograms
• aim easy transfer between applications
• Will extend to other data types
• other histos, fits, ntuples,
• Comments/contributions welcome !

11
History

• Initial idea formed during discussion at
  HepVis-99 workshop at Orsay
• Informal AIDA discussions at CERN in 2000
• AIDA workshops
• January 2001 - Paris/Orsay
• April 2001 - Boston (preceding HepVis 2001)
• Informal meetings (e.g during Geant4 meetings and
  video conferences)
• June 2002 CERN
• Interfaces have been designed by discussion and
  (eventual) consensus
• Takes some time, but result is well though out
  and robust

12
Organization - Developers

• No formal collaboration/author list.
• Some people who have contributed (ideas, code,
  etc).
• Guy Barrand, Pavel Binko, Mark Donszelmann,
  Wolfgang Hoschek, Tony Johnson, Emmanuel
  Medernach, Dino Ferrero Merlino, Lorenzo Moneta,
  Jakub Moscicki, Ioannis Papadopoulos, Andreas
  Pfeiffer, Max Sang, Victor Serbo, Max Turri
• Apologies to people accidentally missed

13
Organization Code, Documentation

• AIDA open source project
• CVS repository cvs.freehep.org
• anonymous download available
• Web page http//aida.freehep.org
• General information, relevant links
• Tutorial, users guide, examples
• Downloads and web-browsable source code
• Test cases (coming soon)

14
Current Status

• AIDA Version 2.2 released (December 2001)
• First End User release
• Three implementations of AIDA exist
• Anaphe/Lizard (C)
• http//anaphe.web.cern.ch/anaphe
• Open Scientist (C)
• http//www.lal.in2p3.fr/OpenScientist
• JAIDA/JAS (Java) AIDA-JNI 1.0 (C)
• http//java.freehep.org/lib/freehep/doc/aida
• GEANT4 adopted AIDA for analysis
• AIDA 3 Currently under discussion
• Release foreseen for Sep 2002

15
AIDA Interfaces Summary

• AIDA Factories
• ITuple
• IHistogram
• ICloud
• ITree

16
AIDA Design
17
Aida design (details)
File /data/pfeiffer/Rose/AIDA-2.2.mdl Sun Jun
30 171839 2002 Class Diagram Logical View /
Main Page 1
18
Example Program (Java)

• Create, fill, and view
• 1D and 2D histograms

19
ITuple

• ITuple - interface to the Data
• get/set methods for double, float, int,
• Information about columns min, max, mean, rms
• Navigating start(), next(), skip(int nRows)
• Project ITuple into 1D, 2D, 3D histogram
• New features for AIDA 2.3
• Support for complex internal structures
  (subfolders)
• Merging and chaining of ITuples under discussion

20
Details - ITuple

• Interface to
• the Data

21
IHistogram (1D-3D)

• Binned histogram IHistogram1D, 2D, 3D
• fill methods (with/without weight)
• Histogram info entries, mean, rms, axis
• Bin info centre, entries, height, error
• Histogram arithmetic add, multiply, divide
• Convenience methods, like coordinate-to-index
  conversion

22
Details IHistogram (1D-3D)
23
ICloud

• Unbinned collection of points ICloud1D, 2D, 3D
• Can represent scatter plot, dynamically
  rebinnable histogram
• Can be converted to a binned histogram
• Standard get/set methods for entries
• Collection info lower, upper, mean, rms

24
Details - ICloud
25
IFunction and Fitting

• Fitting IFunction, IFitFunction
• IFunction simple interface, allows to set
  parameters and get function value
• IFitFunction fit function to a histogram
• Extends IFunction
• Various fit control methods step size, bounds,
  etc.
• Allows to perform fit and get results
• AIDA 2.2 fitting functionality fairly limited
• AIDA 2.3 (Under discussion) extended
  functionality

26
ITree

• ITree
• directory-like structure (Unix directory
  convention)
• Methods like cd, ls, mkdir, etc.
• AIDA analysis objects (tuples, histograms,
  clouds, ets.) exist within ITree directories
• save/restore functionality, hides storage
  details from the user
• Compatible with database or file storage
• Can support multiple file formats
• Mount/Unmount functionality (like unix) allows
  multiple stores to be seamlessly merged
• AIDA XML format is defined for data interchange

27
Details - ITree

• Directory-like
• structures ITree

28
June 2002 Developers Workshop

• 2 Day Users Workshop
• 3 Day Developer Workshop
• Items under discussion
• Fitting (two proposals, from SLAC, CERN)
• Similar but different, differences need to be
  resolved
• Improved plotting (IPlotter)
• Graph (XYData)
• Tuple chaining, merging
• Small updates/extensions to API
• Input/Participation from people always welcome !

29
Ongoing work

• Refining Fitter and Plotter components
• IFitter, IOptimizer, IFitResult
• Iplotter, IStyles for plotting
• Developer-level interfaces
• Code sharing
• More robust operation
• Put AIDA-based utilities in CVS
• Utility to test AIDA implementations
• User contributions
• Common binary storage format

30
Summary

• Abstract Interfaces de-couple components of
  frameworks
• Weakly coupled components and frameworks have
  large advantages
• User code needs no change if changing
  implementation
• Even across language boundaries (JAIDA)
• Ease of re-use of a component
• Flexibility through independence of
  implementation
• Maintainability through independent evolution of
  components
• Example using Geant-4 and AIDA compliant analysis
  tools (see tutorial)