Re: [PyX-devel] Re: [PyX-checkins] pyx/pyx normpath.py,1.10,1.11

[Michael S. <m-s...@us...>]

On 28.09.05, Andre Wobst wrote:
> Hi,
> 
> On 28.09.05, André Wobst wrote:
> > allow pathitem append on empty normpath
> > ...
> > !             if self.normsubpaths:
> > !                 context = path.context(*(self.normsubpaths[-1].atend_pt() +
> > !                                          self.normsubpaths[-1].atbegin_pt()))
> > !                 item.updatenormpath(self, context)
> > !             else:
> > !                 self.normsubpaths = item.createnormpath(self).normsubpaths
> 
> There is an interesting misfeature in the normpath: We can descripe a
> path without a currentpoint, which is not empty. This is done by an
> empty normsubpath (neither having any normsubpathitems nor a
> skippedline) used as the last normsubpath in a normpath. In PostScript
> you can't describe such a path. Only a completely empty path doesn't
> have a currentpoint.
> 
> There are several options here:
> 
> 1) We can ignore that. (But the code fragment shown above will fail
>    when adding a pathitem to such a normpath (=self).)
> 
> 2) We can forbid to have an empty normsubpath as the last normpath
>    except when we have only a single normsubpath, i.e. the path is
>    still completely empty.
> 
> 3) Any normsubpath should have a currentpoint (or a movetopoint or a
>    staringpoint or whatever you name it). We would not need the trick
>    with the zero-length skippedline for descriping a moveto anymore.

> I'm +1 for option 3. Note that this implies, that an empty path cannot
> contain any normsubpath anymore. Not even an empty one, since there
> are no real "empty" normsubpaths anymore, since they at least have
> a movetopoint.

A clear -1 from me for option 3. Just because of what you say:

> This also makes it impossible to build a normsubpath by
> first creating an empty one and then filling it with normsubpathitems.
> This might be unhandy.

Especially for deformers, ... where new normpaths are built. It is a
very convenient feature of the normpaths that the can be built in a
loop without always testing if it has been created already. This is
what made the smoother e.g. unreadable.

On the level of normsubpaths, which are used internally I would like
to keep the creation procedure with an emtpy normsubpath.
On the level of normpaths, this could be different.
What about forbidding to _add_ an empty normsubpath to a normpath?

Then, the question
  if self.normsubpaths:

really does what you expect it to do.
The creation of a normpath reads then:
1. create an empty normpath
2. create the normsubpaths in a loop
2a add them if they are not empty

Alltoghether, this is a little weaker than your option 2
which gets a +1 from me.

4) Another possiblity would be a self.good() which tests each normsubpath for
   filling (-1 from me because it will be sloooooow)

> Last but not least we should take into account,
> that the epsilon is a feature of a normsubpath. In order to set the
> epsilon we would also need to set the starting point either by passing
> some normsubpathitems or by setting it explicitly (as a moveto would
> do it instead of using the skippedline-trick).

Michael.

-- 
"A mathematician is a device for turning coffee into theorems"
  Paul Erdös.