Re: [Tracker] Updated To do for next version

[Jamie McCracken <jamiemcc blueyonder co uk>]

Renà Stadler wrote:
> Am Dienstag, den 03.10.2006, 12:50 +0100 schrieb Jamie McCracken:
> [...]
> > I think for image files (gif, tiff and others) we might want to use 
> > imageMagick's identify command to get the dimensions (especially if 
> > libexif and libpng are not present as we can fallback on this too)
> >
> > EG
> >
> > identify -format "%wx%h"   sample.jpg
> > 2112x2720
> >
> >
> > this will allow us to get the Image.Width and Image.Height by parsing 
> > stdout. We also dont need a hard dependency on ImageMagicks stuff.
> >
> > It can also be used to get comments embedded and other stuff like 
> > resolutions in image :
> >
> > identify -format "%c"   sample.jpg
> [...]
>
> Using identify might be a bad idea.  Last time I checked, it had the bad
> habit of performing all of its extraction and analysis capabilities
> (counting colors etc.) even if you just request things as simple as the
> geometry or the embedded comment (which should become available after
> just parsing a few bytes of most image formats).  That is, it is very
> slow/wastes CPU cycles.  For getting the geometry, I found that it is
> much faster and efficient to use a GdkPixbufLoader and connect to the
> "size-prepared" signal.  Not that I think this is the perfect solution
> either (I think GdkPixbuf has already allocated memory to hold the
> complete decoded image when it emits the signal), but it tells about the
> deficiencies of Imagemagick.

yes you might be right there (identify does seem a little slow on large 
images)

We cant depend on GDK either.

I only suggested it as a fallback for images not supported by libexif or 
libpng (which means stuff like tiff, gif, xpm et all). If we have time 
we can extract libextractor's plugins for these other formats which 
should be more efficient.

--
Mr Jamie McCracken
http://jamiemcc.livejournal.com/