notae / cpfd Goto Github PK

• Seamless multiple representation type:
  • Set
  • Range
  • etc.
• Probabilistic Domain

module layout:
Control.CPFD -- for user import
Control.CPFD.Solver
Control.CPFD.PrimitiveConstraint
etc.

should be executed by cabal test.

Pool (list of Var) should be stored into StateT state to avoid to take it around.
Then I can hide Pool from user.

Extend label functions.

• Consider domain change : pass old and new domain to propagation rules

During a domain propagation, if other propagations modified the domain, the preceding propagation may expand the domain. Domains are normally reduced monotonously.
An example scenario:
step 1: preceding planed [1,2,3] -> [1,2]
step 2: another reduced [1,2,3] -> [3]
step 3: preceding try [1,2,3] -> [1,2] but expand [3] -> [1,2] actually

Test Items:

• Container laws (function cmapA, cup, crown)
• Variable domain propagation
• Variable domain labeling
• etc.

• Move data (ex. Domain) to Data.*
• Separate Control.CPFD into sub modules.

Support soft constraints (ex. FCSP).
This is related to #4, #14.

Statistics information:

• Counts of variable domains updated
• Counts of branches in variable labeling
• etc.

Interface / Implementation:

• Store information into StateT state
• Initialize by runFD
• Update by setX functions, labelling functions and etc.
• Read by some function for reading provided to user
• Dump on internal errors

Debug information:

• Dump of solver state
  • Variables: ID (given by new), Domain, Domain Stack
  • Propagation Queue
• Trace
  • Constraint Addition
  • Domain Changed
  • Domain Propagation Enqueue
  • Domain Propagation Dequeue (Execute)

should be executed by cabal test.