Mateusz Loskot — Sat, 23 Nov 2013 10:19:33 GMT

I see you've diagnosed the problem well, would you be able to provide patch? It would speed things up a lot.

yuyoyuppe — Sat, 23 Nov 2013 12:28:31 GMT

Replying to mloskot:

I see you've diagnosed the problem well, would you be able to provide patch? It would speed things up a lot.

I did a quick debugging and discovered that the reason of the first case behavior was initializing transformation matrix(svg_mapper.hpp:248) with invalid values, because it uses bounding box, which has zero-length side if only one point was supplied to svg_mapper instance. Which is why division by zero happenes at map_transformer.hpp:119-120.[[BR]]

Since the situation where svg_mapper object'll have only one point is a degenerated case, I think it's more like a design issue, than a bug. However it's possible to include this case by adding something like this

BOOST_ASSERT_MSG(+std::numeric_limits<type>::infinity() != sx && -std::numeric_limits<type>::infinity() != sx, "svg_mapper cannot contain only one point");

at map_transformer.hpp:121.

And regarding second point, I was wrong because of my own error - not paying enough attention:).

Thank you.

yuyoyuppe — Sat, 23 Nov 2013 12:50:52 GMT

Also there is more... When using difference in case of geometry1 object is the same as output_collection object, difference will produce incorrect output. May be it needs some assert or documentation note too?

Boost C++ Libraries: Ticket #9427: svg_mapper bugs