non symetric graphicName symbols rendered symetric in IE

[pgiraud]

A rectangle graphicName symbol appears like a square in IE (VML). I think this is something related to coordorigin and coordsize. I'm digging into it, and hope I'll be able to provide a patch soon.

OpenLayers.Renderer.symbol.rectangle = [0,0, 10,0, 10,4, 0,4, 0,0];

[pgiraud]

patch with extended example

[pgiraud]

In the VML graphicName manager, we were calling drawCircle even if the symbol was not symetric. The size (width and height) and position (top and left) computation code was effectless as everything was recalculated in the drawCircle method.

Please review the patch.

I think there is still some work to do to factorize part of the code also used for externalGraphic.

[ahocevar]

The attached patch does the refactoring proposed by fredj. Visual tests pass. Thanks for any review.

[ahocevar]

I can confirm that fredj's patch does the right thing, but I refactored it to factor out the aspect ratio related stuff that external graphics have already been using into a separate function. So I think someone else should review this now.

[pgiraud]

Reassigned to me as I was the one who fist proposed a patch. Andreas, I'll look into it tomorrow morning. Thanks.

[pgiraud]

Your factorization patch looks fine. If you can confirm that tests are still OK, you can go ahead and commit. Thanks

[ahocevar]

(In [9026]) Preserve aspect ratio for custom symbols in the vml renderer. Original patch by fredj, refactored by me, r=pgiraud. (closes #1836)

[ahocevar]

This changeset breaks rotation of external graphics (see styles-rotation example).

[ahocevar]

leaner, faster and breaks nothing

[ahocevar]

aspect.path solves the issue with a different approach: symbols in the symbol cache will always have a square extent, not a rectangle. This saves us some calculation and therefore makes rendering even a bit faster.

[ahocevar]

Relevant tests still pass in IE. Please review.

[pgiraud]

OK, I checked the code out and this looks pretty good. I can confirm that this now works in IE6. You can commit.

[ahocevar]

(In [9217]) fixed aspect ratio rendering in IE to not break rotation any more. r=pgiraud (closes #1836)

[openlayers]

Looks like there is a typo in aspect.patch, within importSymbol function. When the diff is negative, the symbol's width is decreased, and it will never be equal to the symbol's height. This leads to incorrect sizing of symbols which height is larger than the width.

I think this code

            symbolExtent.left = symbolExtent.left + diff;
            symbolExtent.right = symbolExtent.right - diff;

is the correct solution.

Sorry, I don't (and shall not) provide any patch. I've no time for this.

[ahocevar]

I created a patch for this, with tests and visual tests to make sure that everything works as expected now.

Note that the failing test in IE is due to #2198, which also has a patch that is awaiting review.

Please review (and maybe also review #2198). Thanks!

[ahocevar]

And thanks anonymous for making us aware of this issue.

[ahocevar]

There are still issues. The following symbols render incorrectly in IE:

[2,0, 4,5, 0,5, 2,0]
[0.5,0, 10,8, 0,8, 0.5,0]

Better patches are welcome...

[ahocevar]

The patch is ok. According to http://openlayers.org/pipermail/users/2009-September/013826.html. IE just does not like floats and behaves better with larger numbers. So for the above definitions, equal symbols defined as follows will work:

[20,0, 40,50, 0,50, 20,0]
[5,0, 100,80, 0,80, 5,0]

Setting this to Review again.

[pgiraud]

Your patch looks good, you can commit.

[ahocevar]

(In [9665]) Fix for rendering custom symbols with aspect ratio differnt than 1. r=pgiraud (closes #1836)