yams #9933 change relation definition syntax [open]

no more relation defined inside the class defining an Entity no more SubjectRelation and ObjectRelation no more Entity.add_relation() only usage becomes from yams.buildobjs import Relation, N, composite) # define Person, Company, Order, WorkOrder works_for = Relation(Person[0,N], Company[0,1]) split_into = Relation(Order[1][composite], WorkOrder[1,N])
priority	normal
type	enhancement
done in	1.0.0
load	0.500
load left	0.500
closed by	<not specified>

Comments

2009/08/17 14:14, written by sthenault
also:
- other parameters still given using named arguments
- deprecates usage of RelationDefinition
- we still have to distinguish relation type / definition, so we keep RelationType class as today
- 2010/01/27 07:23, written by adimascio
  I think that split_into = Relation(Order[1][composite], WorkOrder[1,N]) is a bad syntax and requires additional complexity in the code for little (well, zero from my POV) gain.
  
  After some discussions with Alexandre and Sylvain, we've agreed on the following agenda :
  - remove ObjectRelation and BothWayRelation
  - split cardinality in two distinct properties subjcard and objcard and use the standard UML notation (i.e 1..n, 0..1, etc.)
  - rename SubjectRelation into Relation and continue to allow relation definitions inside the entity class
  - keep class-style kind of declaration (as in class MyRelation(RelationDefinition): ...)
  Any other thoughts ?
  - 2010/01/27 07:29, written by sthenault
    
    use the same Relation class for relation defined in entity definition and for parametrized classes?
    - 2010/02/10 20:35, written by nchauvat
      
      please disallow defining relation within entities.
      - 2010/02/11 07:48, written by sthenault
        
        huum, this sugar seems valuable to me but I won't fight to keep it if everyone
        is fine with removing it.
      - 2010/02/11 08:08, written by adimascio
        
        I agree with Sylvain. I find that 80% of the time, it's simpler and convenient to have the definition of the relation inside the entity type class.
        
        2010/02/11 08:43, written by sthenault
        
        I've one argument though in favor of dropping this ability: when using this
        people rely (without knowing it) on an implicitly created relation type, and the
        relation added that way isn't importable. This is problematic if one want to
        import it to modify permissions or such.
        
        2010/02/11 08:49, written by adimascio
        
        Good point. Also, should we consider renaming RelationType and RelationDefintion ? I've got the feeling some persons get confused.
        
        2010/02/11 09:05, written by sthenault
        
        I think we still need to differentiate the concepts, and imo this is essentially
        a question of vocabulary that should be properly explained (it's in the book) and
        understood. What would you propose to make things clearer?
        
        2010/02/11 09:35, written by acampeas
        
        I find these perfect names and never encountered someone confused by them.
        
        What confuses people is inline (in Entity) relation definition (the Object/Subject thing), and possibly the implicit relation type creation happening there, and (thing of the past) impossibility to set permissions there.
  - 2010/01/27 12:34, written by nchauvat
    
    please provide examples... and reconsider allowing the definition of relations within the definition of entities.
    - 2010/01/27 13:12
      
      you mean you don't agree with "continue to allow relation definitions inside the entity class"
      - 2010/02/10 20:34, written by nchauvat
        
        yes, that's what I mean.