I'm failing at translating the following PHP/ImageMagick code into Ruby RMagick (to make it more manageable for future users and to understand what it's really doing):
$output = array();
$returnValue = 0;
$pngFiles = $myDir->find("/.png$/i");
foreach($pngFiles as $pngFile) {
$cmd = 'convert '.$pngFile->path.' -resize 1x1 -alpha on -channel o -format "%[fx:u.a]" info:'
exec($cmd, $output, $returnValue);
if($output[0] != 1) {
logMessage("PNG file contains some alpha transparency and will not be modified");
}
}
By now I thought I more or less understood what the convert-command is doing, but translating it to RMagick makes me rethink that.
For example: Why is $output[0] != 1 sometimes true on PNGs in $myDir, but RMagick's Image.alpha? is always true on PNGs in $myDir? Am I missing something?
I think the best way to get me back on track would be, if anyone could explain to me what the convert-command is exactly doing (including the expression %[fx:u.a]).
Update: In the meantime I've written the script I needed this information for. You can check it out at Github if it's to any help to you.
The code checks to see if a particular image contains transparency.
-format '%[fx:u.a]' info:
This instructs image magick to check the first image u, the alpha channel of that a and output info on it, it will return 0 if the top-left pixel is transparent and non-zero if not I think. That is why the image is being resized to 1x1, so that only a single pixel needs to be consulted. The -channel o is the opacity channel.
So the code in English would read, cycle through all PNG files, look at the alpha channel (opacity) only, resize to a single pixel and see if it's transparent. Hence the echo message.
Unfortunately I do not know Ruby, or RMagick, but a quick look at the API seems to suggest using image.channel(AlphaChannel) to get the alpha channel (AlphaChannel is a ChannelType value, not sure if you have to specify ChannelType.AlphaChannel), then follow with .resize(1,1) to get the size down, and finish with either .pixel_color(0,0) or .get_pixels(0,0,1,1) to get the Pixel object back (get_pixels() returns an array), which I believe has an opacity attribute. However, the channel() command changes the RGB values to the value of the channel selected, and I'm not sure it preserves the opacity channel so you might just need to look at red forinstance, or omit the .channel() call entirely - though I do not know if that would disrupt the result offhand.
Perhaps if Ruby supports decent functional programming approaches.
image.channel(AlphaChannel).resize(1,1).pixel_color(0,0).red
or this if pixel_color() does not return the opacity for some reason
image.channel(AlphaChannel).resize(1,1).get_pixels(0,0,1,1)[0].red
Without the channel() calls it would be:
image.resize(1,1).pixel_color(0,0).opacity
or
image.resize(1,1).get_pixels(0,0,1,1)[0].opacity
Again, my Ruby is non-existent, so you may have to rearrange those a lot, but the primitives are there.
References