Itl Collectors, behave like Sets. This can be understood easily, if we consider, that every map of sets can be transformed to an equivalent set of pairs. For instance in the pseudocode below map m

itl::map<int,set<int> >  m = {(1->{1,2}), (2->{1})}; 

is equivalent to set s

itl::set<pair<int,int> > s = {(1,1),(1,2),   //representing 1->{1,2}
                              (2,1)       }; //representing 2->{1}

Also the results of add, subtract and other operations on map m and set s preserves the equivalence of the containers almost perfectly:

m += (1,3); 
m == {(1->{1,2,3}), (2->{1})}; //aggregated on collision of key value 1
s += (1,3);
s == {(1,1),(1,2),(1,3),   //representing 1->{1,2,3}
      (2,1)             }; //representing 2->{1}

The equivalence of m and s is only violated if an empty set occurres in m by subtraction of a value pair:

m -= (2,1); 
m == {(1->{1,2,3}), (2->{})}; //aggregated on collision of key value 2
s -= (2,1);
s == {(1,1),(1,2),(1,3)   //representing 1->{1,2,3}
                       }; //2->{} is not represented in s

This problem can be dealt with in two ways.

Solution (1) led to the introduction of map traits, particularly trait partial_absorber, which is the default setting in all itl's map templates.

Solution (2), is applied to check the semantics of itl::Maps for the partial_enricher trait that does not delete value pairs that carry neutrons. Protonic equality is implemented by a non member function called is_protonic_equal. Throughout this chapter protonic equality in pseudocode and law denotations is denoted as =p= operator.

The validity of the sets of laws that make up Set semantics should now be quite evident. So the following text shows the laws that are validated for all Collector types C. Which are itl::map<D,S,T>, interval_map<D,S,T> and split_interval_map<D,S,T> where CodomainT type S is a model of Set and Trait type T is either partial_absorber or partial_enricher.

Laws on set union, set intersection and set difference

Associativity<C,+,== >: C a,b,c; a+(b+c) == (a+b)+c
Neutrality<C,+,== >   : C a;       a+C() == a
Commutativity<C,+,== >: C a,b;       a+b == b+a

Associativity<C,&,== >: C a,b,c; a&(b&c) ==(a&b)&c
Commutativity<C,&,== >: C a,b;       a&b == b&a

RightNeutrality<C,-,== >: C a;   a-C() ==  a
Inversion<C,-,=v= >     : C a;   a - a =v= C()

All the fundamental laws could be validated for all itl Maps in their instantiation as Maps of Sets or Collectors. As expected Inversion only holds for protonic equality, if the map is not a partial_absorber.

                             +    &    -
Associativity                ==   == 
Neutrality                   ==        ==
Commutativity                ==   ==
Inversion partial_absorber             ==
          partial_enricher             =p=

Distributivity Laws

     Distributivity<C,+,&,=v= > : C a,b,c; a + (b & c) =v= (a + b) & (a + c)
     Distributivity<C,&,+,=v= > : C a,b,c; a & (b + c) =v= (a & b) + (a & c)
RightDistributivity<C,+,-,=v= > : C a,b,c; (a + b) - c =v= (a - c) + (b - c)
RightDistributivity<C,&,-,=v= > : C a,b,c; (a & b) - c =v= (a - c) & (b - c)

Results for the distributivity laws are almost identical to the validation of sets except that for a partial_enricher map the law (a & b) - c == (a - c) & (b - c) holds for lexicographical equality.

                                                   +,&    &,+
     Distributivity  joining                       ==     ==
                     splitting   partial_absorber  =e=    =e=
                                 partial_enricher  =e=    ==                     
                     
                                                   +,-    &,-
RightDistributivity  joining                       ==     ==
                     splitting                     =e=    ==

DeMorgan's Law and Symmetric Difference

DeMorgan<C,+,&,=v= > : C a,b,c; a - (b + c) =v= (a - b) & (a - c)
DeMorgan<C,&,+,=v= > : C a,b,c; a - (b & c) =v= (a - b) + (a - c)

                         +,&     &,+
DeMorgan  joining        ==      ==
          splitting      ==      =e=

SymmetricDifference<C,== > : C a,b,c; (a + b) - (a * b) == (a - b) + (b - a)

Reviewing the validity tables above shows, that the sets of valid laws for itl Sets and itl Maps of Sets that are neutron absorbing are exactly the same. As expected, only for Maps of Sets that represent empty sets as associated values, called neutron enrichers, there are marginal semantical differences.