Re: [PyX-devel] Re: [PyX-checkins] pyx/pyx trafo.py,1.53,1.54

[Andre W. <wo...@us...>]

Hi,

On 14.09.05, Joerg Lehmann wrote:
> > I'm aware that we need the epsilon in the constructor only, but still
> > I would do it very much like to keep it in the trafo as an instance
> > variable and inherit it to newly created trafos (like by rotated) ...
> 
> What do you want to do in the case of __mul__, when there are two
> trafos.

Hmmm. I don't know. We do have the same problem in normpath. In some
sense I would suggest the lower value for epsilon (and None, if one of
the epsilons is None). I don't think its a problem at the normsubpath,
that we increase the precision by that. What do you think?

(Yes, I would like to do the same in normsubpath and for the trafos.)

> > PS: I think we should set the default for trafo._epsilon to 1e-10 (or
> >     even lower!), since a determinate for a 2x2 matrix has the
> >     dimension of its items to the power of two. We should be able to
> >     scale by a factor of 0.001 without any problems.
> 
> Definitely, let's go for (1e-5)**2.

Ok. Just to remember, that epsilon in normpath is measured in pts. To
scale down a one in our default unit cm, the factor is:

    andre@pb:~/python/pyx$ python
    Python 2.3.3 (#1, Aug 23 2004, 20:06:57)
    [GCC 3.3 20030304 (Apple Computer, Inc. build 1640)] on darwin
    Type "help", "copyright", "credits" or "license" for more information.
    >>> from pyx import normpath, unit
    >>> (normpath._epsilon/unit.topt(1))**2
    1.2445216049382721e-13
    >>>

I think 1e-10 is fine, but we should keep in mind, that this threshold
is still not yet generous. And I don't want to say, that the epsilon
in normpath and in trafo is directly related. But its fine as an
educated guess ... ;-)


André

-- 
by  _ _      _    Dr. André Wobst
   / \ \    / )   wo...@us..., http://www.wobsta.de/
  / _ \ \/\/ /    PyX - High quality PostScript and PDF figures
 (_/ \_)_/\_/     with Python & TeX: visit http://pyx.sourceforge.net/