Re: [Sketch-list] About filling open (sub)paths

[Antoon P. <Ant...@re...>]

On Sat, May 04, 2002 at 08:45:20PM +0200, Bernhard Herzog wrote:
> Antoon Pardon <Ant...@re...> writes:
>=20
> > On Sun, Apr 28, 2002 at 09:31:46PM +0200, Bernhard Herzog wrote:
> > > The most importand being that it's what PostScript, PDF, SVG,
> > > Illustrator (I haven't verified this now but it should follow from
> > > PostScript's behavior) and XFig (there are probably more).
> >=20
> > I  don't think this is a good reason.
>=20
> Well, I think it's a good reason to support this behavior in Sketch.

Well, you are the final arbiter.

> > CGM doesn't do it this way.
> > If I start to search I can probably find more too. Do  you  think
> > this  should  be  decided  by  a poll?
>=20
> Depends. It's definitely not a good idea to change Sketch's treatment o=
f
> fills in open paths. Adding some other way to handle them os possible,
> but I'm not very keen on it as it makes the code quite a bit more
> complicated (more case distinctions in many places) for comparatively
> little gain.

In this case I'm wearing my usercap. I don't care about implemen=AD
tation problems, I want a user interface that is as consistent as
possible.   I  understand  that implementation problems sometimes
stand in the way.  But I consider changes because of that as  de=AD
sign concessions, not as good arguments for designing it as such.

> All the programs I mentioned don't fill paths with open sub-paths at
> all, but that's not exactly what you seem to propose. If I understood
> you correctly you want open sub-paths to be ignored when filled but not
> closed sub-paths. Doing this differently than all other programs is not
> a good idea IMO.

Well doing it differently has it's drawbacks, I agree with  that.
However that is an argumument that doesn't weights the merrits of
each approach.  Sometimes the most popular approach can have  se=AD
vere drawbacks, but nobody wants to deviate from it because it is
treated as the standard to follow. That's why I personnaly  don't
like it that much.

> How exactly does this work in CGM?

Well  I'm for the moment only familiar with version 1.  There you
have, for the purpose of this discussion, three different curves.
A  polyline,  this  is  an open path, a polygon, this is a single
closed path and a polygonset, witch is a multiple closed path  of
which  each  vertex  can  be made visible or invisible.  Only the
polygon and polygonset has a fill attribute.

> > > Early unpublished Sketch versions actually followed Corel's example=
 but
> > > I changed that because being able to fill open paths actually gives=
 you
> > > more flexibility and it makes data exchange with other programs eas=
ier.
> >
> > Well if you want more flexibility why not allow color, dashpatern
> > or at least the visibility of  each  edge  seperately?
>
> Because I'm too lazy :-)
>
> The increased flexibility is not about making one edge invisible or
> something. That edge doesn't exist! It's only virtually there for
> filling.

Well I would say if it doesn't exist it can't be there as a  bor=AD
der  for  the filling. The problem is that most times the way you
add a virtual edge it isn't the one I want. Take  to  open  paths
that  are combined into one curve. The way you add a virtual path
is by closing each subpath. But that is not how I expected it the
first  time  I tried it. I expected the two open paths to be seen
as one path so that the fill area would be the polygon within the
points of the two paths combined.

> With the ability to fill open paths you can create some shapes
> that would be more awkward to create if you couldn't,

Sure  but  to the user you also create certain expectations. As a
user I don't see open or closed paths. I see either a line or  an
area  of  which  not  all edges may be visible. If I see one edge
that is not visible I expect I can  make  other  edges  invisible
too.

> OTOH, all shapes
> you can create when open paths can't be filled can be easily created if
> you can fill them by simply setting the fill to empty.
>
>
> I think we have to distinguish two things here:
>
> 1. What's the area to be filled
>
>    Here Sketch currently uses the even-odd rule assuming a straight lin=
e
>    between start and endpoint for open sub-paths. Support for
>    winding-rule fills would also be good, though, and in fact required
>    for full support of postscript/AI/PDF.

Well I think that one can argue that for an open path there is no
such area.

> > That  would certainly add more flexibilty. One thing you may con=AD
> > sider then, is that paths can only be filled or stroked  and  the
> > possibilty  to create layerd paths, each layer can then be filled
> > or stroked or otherwise pattered in a differend  way.  But  maybe
> > that would make things to complicated.
>
> For the time being, yes.

Well I think I am closing the discussion here. My main purpose is
accomplished. I 've given all the arguments I have but the  final
decision  is  yours.  I'll still use sketch whatever the decision
will be.

--=20
Antoon Pardon