Thanks for your prompt reply ! The problem is I DO KNOW there will be never overlapping. ----- Mail original -----

De: "Krzysztof Czainski" <1czajnik@gmail.com> À: boost-users@lists.boost.org Envoyé: Mercredi 19 Mars 2014 13:18:39 Objet: Re: [Boost-users] [ICL] Can't use an enum as map value

2014-03-19 13:11 GMT+01:00 Oodini < svdbg___@free.fr > :

Hello,

I declared an interval_map :

interval_map myMap;

Later, I do :

SomeEnumType enumValue = SOME_ENUM_VALUE; myMap.add(std::make_pair(disxrete_interval(10,17), SOME_ENUM_VALUE);

But that doesn't compile.

It seems that ICL wants the += operator must eb defined for value type (why ???).

What can I do ?

Hi,

If I remember correctly, you need to supply a combiner. The default combiner uses +=.

A combiner produces a new value for overlapping intervals. Consider (pseudocode):

myMap.add( [10,17), SOME_ENUM_VALUE ); myMap.add( [16,18), OTHER_ENUM_VALUE );

The combiner decides, what should be the value in [16,17). In this case, tries to += the OTHER_ENUM_VALUE to the existing SOME_ENUM_VALUE.

HTH, Kris _______________________________________________ Boost-users mailing list Boost-users@lists.boost.org http://lists.boost.org/mailman/listinfo.cgi/boost-users

That's run-time knowledge that the compiler can't infer or use. Investigate whether it's possible to avoid this behaviour via some policy

Probably, but the documentation is not very clear about the policies.

or if it might be acceptable to have a bogus operator+= defined for your type somewhere where it can be found.

By using the insert() method, instead of add(), I can get what I wanted (see the other thred I initiated), because the insert() method isn't annoyed by the enum type.

If none of those are acceptable, you could always consider patching the library to provide such a policy or trait.

I'd like to, but I dont' think my employer will agree... (time is money !).