
Boost : 
Subject: Re: [boost] [Review] GGL review starts today, November 5th
From: Bruno Lalande (bruno.lalande_at_[hidden])
Date: 20091113 08:16:49
> Well, maybe it is sufficient, but my point is only if someone checks the doxygen "classes" for example, he can be quite confused by all the templated elements. Maybe I am only not enough used to read (so much) templatedriven library documentation. The question is the needed level to read the doxygen document. Maybe some elements should not appear in the doxygen to make it clearer for the user ?
OK I see your point. We will check that, maybe some conventions of
syntactic shortcuts can be adopted to come up with something lighter
indeed.
> Yes completely. But I have a question about extension. What is the complexity of extending for a new kernel 'type' for example ? Because STL is known for its "number containers + number algorithms" system. Is it comparable ? Or is it more a "Sum for each element of ggl kernel of: "number of algorithms supported by the ggl::element"" ?
Well, I would say it's between both. The pointpoint distance
algorithm is available for each thing than can be seen as a point (by
adapted it using registration macros etc), but obviously if you decide
to add a new *concept* (let's say, an ellipse) it will hardly handle
it. So there's one algorithm for each concept, and all the possible
models of each concept (that is the actual type you adapt) are handled
by the corresponding version of the algorithm.
Does that answer your question?
Regards
Bruno
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk