Performance of python subprocess vs Wand?

Question

I'm looking for a compute- and memory- efficient way to use ImageMagick in a Python program to retrieve the dimensions of batches of photos.

I first used the current favorite ImageMagick-Python package, Wand. I tried it and it felt slow. So I rigged up a test of it vs simply exec-ing ImageMagick using subprocess.check_output(). Subprocess was more than 10 TIMES faster.

My question: is that other people's experience? Is there a way to use Wand faster?

Wand took 1.2 seconds:

1.jpg: 3264x2448
2.jpg: 1600x1200
3.jpg: 1700x1101
4.jpg: 1600x1200
5.jpg: 3648x2736
6.jpg: 2789x1980
7.jpg: 2400x1600
8.jpg: 3648x2736
processed 8 files in 1.236s

Subprocess only took 0.1 seconds to retrieve the same files:

1.jpg: 3264x2448
2.jpg: 1600x1200
3.jpg: 1700x1101
4.jpg: 1600x1200
5.jpg: 3648x2736
6.jpg: 2789x1980
7.jpg: 2400x1600
8.jpg: 3648x2736
processed 8 files in 0.102s

The Wand code:

for filename in files:
  with wand.image.Image(filename=filename) as img: return img.width, img.height

The subprocess code:

subprocess.check_output(['identify', '-format', '%f:%w:%h\\n', 'path/to/imgs/*.jpg'])
# parse output by splitting each line on the ':'

I realize it's not apples to apples because I'm invoking ImageMagick once per file in Wand. However, I don't see any batch option in Wand, so I believe this is the best Wand can do for my scenario.

Thanks!

Davidadavidde

The problem may be that Wand is using ctypes to provide the interface to imagemagick, which may be slow in this particular situation. Did you try to use PythonMagick instead? – Biased

PythonMagick was last revved Sept 2012 and only has a readme as docs, whereas Wand is extremely recent -- Sept 2013 -- and has beautiful docs. And PythonMagic requires Boost, which I doubt is available in my shared hosting environment. – Davidadavidde