Bug 581 ray crossing with 4d coord

[mukoki]

Test to demonstrate #581 bug and proposal to fix it

[dr-jts]

The fix looks good to me. Are you going to sign an ECA?

[dr-jts]

Actually I'm thinking a better fix might be to add a RayCrossingCounter.countSegment that accepts ordinate values instead of coordinates.

Thoughts? I can make a PR for this.

[mukoki]

The fix looks good to me. Are you going to sign an ECA?

Wow, lot of ceremony. I just filled the ECA form, but I don't know how to link my commits to this document.

[mukoki]

Actually I'm thinking a better fix might be to add a RayCrossingCounter.countSegment that accepts ordinate values instead of coordinates.

Thoughts? I can make a PR for this.

I think you're right. We don't really need to get coordinates with the exact coord dimension. We rather need x,y from both coordinates. In this case, you can also remove p1 and p2 coordinate creation and ring.getCoordinate(i, p1) call which is the place where the error occured.

[dr-jts]

Wow, lot of ceremony. I just filled the ECA form, but I don't know how to link my commits to this document.

Yes, the bureaucracy is annoying. AFAIK you can't amend this PR - you have to create another one.

[dr-jts]

Another possibility - just use CoordinateSequence.getCoordinate. Pretty sure modern JVMs optimize small object creation highly, so perhaps this is performant enough to use. Confirming this with a performance test would be good.

[mprins]

Wow, lot of ceremony. I just filled the ECA form, but I don't know how to link my commits to this document.

Yes, the bureaucracy is annoying. AFAIK you can't amend this PR - you have to create another one.

when you go to https://accounts.eclipse.org/legal/eca/validation/45250 you cab see that a signed commit is needed, use git commit -a -s

[mukoki]

Another possibility - just use CoordinateSequence.getCoordinate. Pretty sure modern JVMs optimize small object creation highly, so perhaps this is performant enough to use. Confirming this with a performance test would be good.

Do you mean as I did in the commit (bf88792) ?
Seems to work, but probably do more work than necessary compared to your idea to just use x,y ordinates. I'll try to test the later idea. Not sure I can easily compare performance though.

[dr-jts]

Another possibility - just use CoordinateSequence.getCoordinate. Pretty sure modern JVMs optimize small object creation highly, so perhaps this is performant enough to use. Confirming this with a performance test would be good.

Do you mean as I did in the commit (bf88792) ?

I'm seeing that commit use createCoordinate?

I mean using CoordinateSequence.getCoordinate(i). That results in a Coordinate being created for every access, but this should be fairly efficient.

[dr-jts]

Superseded by #589