Re: [boost] [Review] Boost Graphs Library (GGCL)

From: Beman Dawes (beman_at_[hidden])
Date: 2000-08-30 10:46:40

Jens Maurer wrote:

 > - The library introduces some non-member functions to remain extensible.
 >An extension scenario would be that a user implements his own graph
 >data type implementing e.g. AdjacenyMatrix.  Then I can call edge(u,v,g)
 >on an object g of that graph type from anywhere and Koenig lookup will
 >find the correct function.  What happens if some other subsystem wants
 >to be extensible as well and defines a non-member edge() function with
 >a different meaning?  Hm... Either the function signature disambiguates,
 >or there's a problem.  I must think about this separately, and it's not
 >a specific problem with GGCL.

Name pollution is a serious problem with non-member functions.

It is a problem technically because of the disambiguation issue Jens 
identifies above, and because humans may become confused more easily by a 
non-member name than a member name with its improved context.

--Beman