Benny Malengier wrote:
> 2009/10/2 Peter Landgren <peter.talken@...>:
>>> Yes, I know now how this happened. It comes from
>>> which I don't understand. There is an option in the dialog to select one of
>>> four types of arrows: ancestor -> descendant
>>> ancestor <- descendant
>>> ancestor <-> descendant
>>> ancestor -- descendant
>>> So the user has all possibilities to choose what's best.
>>> So, why this change?
> reopen the bug report and ask the people who gave the patch (that is
> rah) to have a look. Perhaps the strange way arrows where done was to
> avoid this, and on older graphviz this was needed to avoid seeing what
> you see, but on newer graphviz it is no problem....
Sorry I haven't been responsive; I'm back at university now so my time
is limited. Yes, the patch I submitted does reverse the graph direction.
The edges in the GV graphs are directional. This effects the appearance
of arrow heads but also the logic of what the 'direction' of the graph
means. The node parameters to the add_link function are named 'from'
and 'to'. However, without the patch, the nodes that plugins provide
end up on the page with the arrows originating from the 'to' node and
point at the 'from' node. It was obvious the plugins were written to
override the meaning in the API in order to make the graphs look right.
This illustrates the purpose of the patch; the meaning in: the API, the
plugins and the dot file are now all the same and correct.
The fact that the direction logic has changed doesn't make much of a
difference because the user interface for the graph direction doesn't
specify what the context is; there is nothing that says 'ancestor ==
top' or 'descendant == right'. The options could well be labelled
'Horizontal' and 'Horizontal (opposite direction)' and not lose any meaning.
This does change the behaviour of GRAMPS from the user perspective. I
don't think this is a particularly big issue though; the user just has
to change the direction of the graph in the dialog. If this is seen as
a problem though, I would suggest negating the user interface option for
the graph direction. But then that's again, adding more complexity
which is something this patch seeks to reduce.
The choice as far as I can see is to have an unnecessarily complex set
of graph file, API and plugins; an unnecessarily complex graph direction
widget; or have a change from the user's perspective. I'd go with the
change from the user's perspective; I reckon the user can probably
handle it :-)