opencog-dev team mailing list archive

[Gustavo Machado Campagnani Gama <gama@xxxxxxxxxxxxx>]

Hi Joel,

I like your proposal in general. Just  but I have two caveats: 1. I have
a feeling that we will eventually end up with a few cases where the name
will be too long and unfriendly (getHandleSetByTypeOrTargetOrName...);
2. it'll make porting code from other novamente-derived trees a lot
harder.

But I still think the benefits outweight the 

On Thu, 2008-06-26 at 14:02 +1200, Joel Pitt wrote:
> Hi all,
> 
> There are an awful lot of overloaded getHandleSet methods in AtomTable
> that return HandleEntry pointers and in AtomSpace that return
> OutputIterators.
> 
> Is there a reason for making them overloaded instead of making them
> explicit in describing what type of Handle set they return?
> 
> i.e. getHandleSetByImportance(AttentionValue::sti_t,
> AttentionValue::sti_t upperBound = 32767)
> instead of
> getHandleSet(AttentionValue::sti_t, AttentionValue::sti_t upperBound = 32767)
> 
> getHandleSetByTargets(Type*, bool*, Arity, Type type = ATOM, bool
> subclass = true)
> instead of
> getHandleSet(Type*, bool*, Arity, Type type = ATOM, bool subclass = true)
> 
> For the latter example, there are several overloaded methods that get
> a handle set by what atoms have certain targets, and it'd make sense
> to leave these overloaded. Unfortunately, this change would take a
> while to differentiate which of the overloaded method each call refers
> to, but I think it'd be worth it to make the interface friendlier to
> developers.
> 
> I guess whether to do this depends on more for_each iterators like
> Linas created should be made for each equivalent getHandleSet...
> 
> Any thoughts?
> 
> J