[metapost] another question

[Daniel H. Luecking]

The intersectionpoint macro, based on the intersectiontimes

primitive, works by succesively subdividing the two paths

and determining whether the parts are close together. It stops

when the range of the time parameter has been narrowed down

to some preset tiny parameter.


It is mainly luck if this produces the mathematically exact

intersection point. This seems to be especially true if the

point of intersection is a node of both paths. In that

case, the algorithm can even falsely conclude that there is

no intersection.


Coming within a distance less than .008 is good enough for

drawing purposes, but one should not rely on the

intersectionpoint even lying exactly on the paths that supposedly

intersect there.


Dan,


Daniel H. Luecking [luecking at uark.edu]
Department of Mathematical Sciences
1 University of Arkansas
Fayetteville, AR, USA 72701-1201


________________________________
From: metapost <metapost-bounces+luecking=uark.edu at tug.org> on behalf of Walt Burkhard <burkhard at cs.ucsd.edu>
Sent: Tuesday, October 2, 2018 12:37 PM
To: Metapost List
Subject: [metapost] another question

Hello,

I have another question for  the program attached below.  The program is to round out the corner of two intersecting  paths.
When it runs,  the intersectionpoint does not seem to work correctly.   The program makes three calls to intersectionpoint,  the first two seem to work as expected.    The third is to return  the pair (120 ,300) but
instead returns something close  (119.99693. 299.99229).
 Using  numbersystem = double does not resolve this issue.    What is going on here?

Any suggestions would be appreciated.   Many thanks in advance.

Walt Burkhard



-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://tug.org/pipermail/metapost/attachments/20181002/d01e1d03/attachment.html>