Improve straight skeleton robustness

[strk]

This is done by shifting data near the origin
while computing the skeleton.

Fixes #89

[strk]

I'm looking at adding a testcase for this, meanwhile if you see an opportunity to avoid a copy, that'd be nice (but performance isn't bad as it is)

[strk]

Writing the testcase I hit another bug, making "Covers" return true when instead should return false ...
See #93

[strk]

So #93 was finally found to be an invalid bug. But I'm still having problems reproducing the robustness issue when reading the polygon from WKT, as it slightly moves vertices to not represent a problem anymore.

[Remi-C]

If I remember well, WKT limits the precision.
WKB on the opposite is more precise (about 2 digits more).

Maybe you should test using wkb as input, and not wkt.

Cheers,
Rémi-C

2015-06-17 11:10 GMT+02:00 Sandro Santilli notifications@github.com:

So #93 #93 was finally found
to be an invalid bug. But I'm still having problems reproducing the
robustness issue when reading the polygon from WKT, as it slightly moves
vertices to not represent a problem anymore.

—
Reply to this email directly or view it on GitHub
#92 (comment).

[strk]

Right, but I can't find a way for SFCGAL to take WKB input.

[vmora]

This may be worth an enhancement ticket

[strk]

Filed: #94

[strk]

Pushed as c9e3cff -- testcase will need to wait (will file another ticket)