fill-opacity different from SVG

[rmagick]

I'm trying to reproduce this example from the SVG spec (http://www.w3.org/TR/SVG11/masking.html) but I see a discrepency. Notice the 2nd pair of circles from the left on the bottom.

I believe this MVG script is the equivalent of the SVG:

Code: Select all

push graphic-context
 scale 0.355,0.355
 push graphic-context
  fill '#00000000FFFF'
  rectangle 100,100 1100,250
 pop graphic-context
 push graphic-context
  fill-opacity 0.5
  push graphic-context
   fill '#FFFF00000000'
   fill-opacity 1
   circle 382.5,250 432.5,250
  pop graphic-context
  push graphic-context
   fill '#000080800000'
   fill-opacity 1
   circle 417.5,250 467.5,250
  pop graphic-context
 pop graphic-context
pop graphic-context

However, that MVG script produces this pair of entirely opaque circles instead of the .5 opaque circles shown in the SVG example.

[magick]

ImageMagick can produce MVG from SVG for those SVG graphic elements ImageMagick supports. For your example, type

convert msvg:opacity01.svg opacity01.mvg

You now have the MVG required to produce the image:

Code: Select all

push graphic-context
  viewbox 0 0 340 99
  affine 0.283333 0 0 0.282857 0.0 0.0
  #
  #  Example opacity01 - opacity property
  push graphic-context
    fill 'none'
    stroke 'blue'
    rectangle 1,1 1199,349
  pop graphic-context
  push graphic-context
    fill '#0000ff'
    rectangle 100,100 1100,250
  pop graphic-context
  push graphic-context
    fill 'red'
    opacity '1'
    circle 200,100 200,150
  pop graphic-context
  push graphic-context
    fill 'red'
    opacity '.8'
    circle 400,100 400,150
  pop graphic-context
  push graphic-context
    fill 'red'
    opacity '.6'
    circle 600,100 600,150
  pop graphic-context
  push graphic-context
    fill 'red'
    opacity '.4'
    circle 800,100 800,150
  pop graphic-context
  push graphic-context
    fill 'red'
    opacity '.2'
    circle 1000,100 1000,150
  pop graphic-context
  push graphic-context
    opacity '1'
  push graphic-context
    fill 'red'
    opacity '1'
    circle 182.5,250 182.5,300
  pop graphic-context
  push graphic-context
    fill 'green'
    opacity '1'
    circle 217.5,250 217.5,300
  pop graphic-context
  pop graphic-context
  push graphic-context
    opacity '.5'
  push graphic-context
    fill 'red'
    opacity '1'
    circle 382.5,250 382.5,300
  pop graphic-context
  push graphic-context
    fill 'green'
    opacity '1'
    circle 417.5,250 417.5,300
  pop graphic-context
  pop graphic-context
  push graphic-context
    opacity '1'
  push graphic-context
    fill 'red'
    opacity '.5'
    circle 582.5,250 582.5,300
  pop graphic-context
  push graphic-context
    fill 'green'
    opacity '.5'
    circle 617.5,250 617.5,300
  pop graphic-context
  pop graphic-context
  push graphic-context
    opacity '1'
  push graphic-context
    fill 'green'
    opacity '.5'
    circle 817.5,250 817.5,300
  pop graphic-context
  push graphic-context
    fill 'red'
    opacity '.5'
    circle 782.5,250 782.5,300
  pop graphic-context
  pop graphic-context
  push graphic-context
    opacity '.5'
  push graphic-context
    fill 'red'
    opacity '.5'
    circle 982.5,250 982.5,300
  pop graphic-context
  push graphic-context
    fill 'green'
    opacity '.5'
    circle 1017.5,250 1017.5,300
  pop graphic-context
  pop graphic-context
pop graphic-context

[rmagick]

Apparently I was not clear about the difference I am questioning. In the ImageMagick output the pair of red & green circles 2nd from the left on the bottom row look like they're entirely opaque, but in the SVG example they are .5 opaque.

Per your advice I downloaded the SVG example and converted it to MVG (the result is of course identical to the MVG file in your reply) and then to PNG with these commands:

Code: Select all

convert msvg:test/.test_opacity01.svg im_opacity01.mvg
convert -size 340x100 im_opacity01.mvg im_opacity01.png

The resulting PNG is:

The SVG example (here) shows the circles to be .5 opaque.

This is ImageMagick 6.5.4-10 2009-08-10 Q16.

[magick]

We can reproduce the problem you reported and have a patch for it available within the next day or two. Thanks.

[rmagick]

Thanks!

[rmagick]

Should I be seeing the fix in 6.5.5-2? My test still demonstrates the problem.

[magick]

Post the image you are generating and mention what is wrong with it. In our tests we appear to be getting correct results.

[rmagick]

We're both right. I guess your test uses the opacity primitive. Since I'm using the MagickWand API, my test uses fill-opacity. The fill-opacity primitive does not produce the desired result.

Here are two versions of the same test. This was created using the DrawSetFillOpacity in the MagickWand API: http://home.roadrunner.com/~foxhunter/f ... city01.mvg. In this version, the 2nd pair of red & green circles on the bottom row is entirely opaque but should be .5 opaque.

Here's the output:
.

This version is identical except I changed all uses of the fill-opacity primitive to opacity: http://home.roadrunner.com/~foxhunter/opacity01.mvg. The circles are now .5 opaque as expected. Here's the output:

[magick]

Looks like we'll need to add DrawSetOpacity() to the API. Give us a day or two. Thanks.

[rmagick]

I see DrawSetOpacity in 6.5.5-5 and now I have a question. I'm confused about the difference between opacity and fill-opacity so I looked at magick/draw.c. I see that opacity changes the fill opacity as well as the stroke opacity and fill-opacity changes only the fill opacity. However it appears that the opacity value is computed differently for the two. For opacity, it's

Code: Select all

graphic_context[n]->opacity=RoundToQuantum((MagickRealType)
   QuantumRange*(1.0-((1.0-QuantumScale*graphic_context[n]->opacity)*
   factor*atof(token))));
graphic_context[n]->fill.opacity=graphic_context[n]->opacity;

For fill-opacity, it's

Code: Select all

graphic_context[n]->fill.opacity=RoundToQuantum((MagickRealType)
   QuantumRange*(1.0-factor*atof(token)));

That is, the opacity primitive modifies the current opacity value, but fill-opacity just computes a new opacity value w/o referring to the current value. This is what I'd expect just by looking at the results of my tests (above). Is this difference deliberate? I would expect them to be computed the same way.

[magick]

See http://wiki.services.openoffice.org/wik ... ity_in_SVG which differentiates opacity, fill-opacity and stroke-opacity. If after reading you think ImageMagick's implementation is buggy, let us know.