Re: [Cdk-devel] [Cdk-commits] SF.net SVN: cdk:[15495] detgen/trunk/src/org/openscience/cdk

[Egon W. <ego...@gm...>]

Hi Gilleain!

On Thu, Mar 25, 2010 at 4:35 PM, gilleain torrance
<gil...@gm...> wrote:
> I might mention that the AtomContainer class is essentially a graph.

Almost: bonds are not restricted to two atoms.

> There are many different ways to represent a graph, each with
> advantages and disadvantages:
>
> a) Edge list (current implementation) - space efficient, but slow for
> getting connections
> b) Adjacency list - less space efficient, but quicker for getting connections
> c) Adjacency matrix - takes up lots of space for sparse graphs (like
> molecules) but useful for certain operations
>
> Unfortunately, this is a classic space/time tradeoff issue, where you
> can either have fast access to connections or use less memory, but not
> both.
>
> I don't know what would be best here; different AtomContainer
> implementations (AdjacencyMatrixAtomContainer, EdgeListAtomContainer,
> etc)

Alternative implementations was one of the reasons why we now have
interfaces, and should be programming against interfaces.

> or let the AtomContainer class internally chose between different
> ways to represent the molecular graph.

Yeah, that would be ideal :)

> Sorry to lecture the list on this topic, but thought I would anyway :)

Oh, please do not stop! That's what this list is for!

Egon


-- 
Post-doc @ Uppsala University
Proteochemometrics / Bioclipse Group of Prof. Jarl Wikberg
Homepage: http://egonw.github.com/
Blog: http://chem-bla-ics.blogspot.com/
PubList: http://www.citeulike.org/user/egonw/tag/papers