[Patch] Simplify calculation of SVG size in ImageProvider

[floscher]

This is a small patch improving how the ImageProvider calculates the x and y scaling factors for an SVG:

• the scaled width and height as well as the scaling factors are now final variables that are assigned only once.
• all negative numbers in the desired image dimensions are treated the same (that dimension is scaled automatically). Before only -1 was treated this way, other negative values were treated as the desired width/height.
• the logo sizes are now square (128×129 → 128×128 and 256×258 → 256×256)

[floscher]

ping

[Don-vip]

Replying to floscher:

• the logo sizes are now square (128×129 → 128×128 and 256×258 → 256×256)

I don't remember why we (I?) did that but it was intentional and wasn't a typo. Did you change them to fix a specific issue?

[floscher]

I think (can't check right now), the logo was not exactly square until I modified the SVG logo and amongst other things made the size of the image exactly square by adjusting the blank space around the logo.
I'll check that later.

[floscher]

Ok, I see. You added the 128×129 dimension in https://josm.openstreetmap.de/changeset/7768/josm . At that time the PNG version of the logo was 256×258, but the SVG version was 1000×1000px in size at all times.

So it should be ok to make the dimensions square, now that only the SVG version is used.

[Don-vip]

In 13723/josm:

fix #16201 - Simplify calculation of SVG size in ImageProvider (patch by floscher)

[Don-vip]

This breaks org.openstreetmap.josm.gui.mappaint.MapCSSRendererTest.testRender[way-repeat-image]:

jenkins/job/JOSM/4141/jdk=JDK8/testReport/junit/org.openstreetmap.josm.gui.mappaint/MapCSSRendererTest/testRender_way_repeat_image_/

[floscher]

Follow-up patch for the previous patch: Fixes calculation of scaling factor which caused functional test to fail.

[floscher]

Thank you for the notification. The issue was, that I calculated the scaling factor based on the width/height of the resulting raster image (integer) and the size of the unscaled SVG (floating point number). It's now calculated from the size of the scaled SVG as floating point number and the size of the unscaled SVG also as floating point number.

The test org.openstreetmap.josm.gui.mappaint.MapCSSRendererTest.testRender[way-repeat-image] still fails, but the differences are really minimal (30 pixels differ by 18 or less). These are probably rounding artifacts, because changed the calculation a bit around so rounding happens at different places than before.

New output:

Diff to reference image:

[floscher]

Output of the way-repeat-image test after the second patch

[floscher]

Difference between the reference image and the output of way-repeat-image test after the second patch

[Don-vip]

In 13737/josm:

fix #16201 - Fixes calculation of scaling factor (patch by floscher)