•  unittest
• based upon JUnit
  
• doctest
• light weight, text based
  

• py.test
• nose
  

•  Parametric tests
• does f(x) do the right thing over many parameters
• looping over large data set not a good solution, first failure kills the test
  
• doctest workfolw
• emacs doctest mode is very good
• Standalone scripts
• Test failures are a pain to isolate

•  Assign a probability to a path through the workflow
• get fine grandularity
• unittest feel
• if python code to test, write python code around testing code.
• Package
• jumbl - U. of Tennessee
• working on python bindings
• under active development
• jumble
  
• fuzzy testing
  

•  Markov analysis
  

• Performance, with long-time historical tracking?
• testing framework part of your build system and version control system
  
• use subversion repository and result is stored as a version number in the change log
  
• Easy acceptance of changes

• good systems?
• what is UI testing?
• GUI
  
• OS specific
• wing32 accessibility framework
• QT4 has testing built in
  
• kwwidgets (Kitt Wear is the maker)
  
• BSD license
• TK/TCL and C++
• only two cross platform toolkits
  
• GUI specific
• ParaView
• kww -> QT4
•  VTK
• generates imagery take diff threshold
• DART2
• automatic testing after commits
• RedHat "dogtail"
• Write code in a more declarative manner so unittest framework can mange/test state transitions.
• LDTP
• test everything accessibility enabled
• linux only
• KDE
• GNOME

• stat profile
• cprofile
  

•  20 -30 %
• retro-fitting code into tests is way more expensive than writing tests to begin with.

•  Regression testing is your only option?
• Converage analysis?
• No good tools in existence
• 
  

•  Low traffic
• unmoderated
  

•  List of all existing testing tools
  

•  Only first 2 or 3 grips have any kind of solution.
• Things that can be checked into version control for diff-ing would be ideal
• pylint
• can set it so that when you do a checkin you get a report on the syntax checking
• Something similar for testing would be useful
• checks run at commit time would be a good thing
• Levels of testing
• scipy does it
• numpy does it
• can be built into your unittest system
• levels only useful for big projects?
  
• "Dive into Python"  Book
• Good chapter on testing
• Nose
  
• supports tags, levels of testing, timing, plug-ins
• Nose better than. Py.Test
• Nose
• docs
• ease of use
• based upon uninttest
• can be run from setup.py
• Test collection of Nose is a big win
• Low overhead dependency
  
• Py.Test
• developer friendly
• multiple processes
• requires py.lib
• parallel testing
• May also have a test collection facility
• unittest is being extended - C. Winters?
  
• will have garbage collection
• orthogonal to Nose and py.test
• will not have test discovery
• Perl Test Anything Protocal
  
• TAP
  
• Format specification for saying that I ran something and this happened
• Code Coverage testing
• figleaf
• Developed by Titus
  
• coverage.py
• uses the actual AST - more comprehensible reports

•  This is lacking in all languages
• gmemprof for C
• getmemusage()
• enthought.util
• blunt instrument
• Valgrind for C++
• scan /proc/ for Linux
• Good for looking for growth issues
• could be all that getmemusage() does