unnormal memory usage (x4)

[a-shafaat]

Describe the bug
reading images uses a lot more memory than it should. reading a 2 MB image using GD uses only 160 MB but the library uses 545mb!

Code Example
$manager = new ImageManager(new Driver());
$image = $manager->read($filePath, FilePathImageDecoder::class); // 545mb memory peak
vs
$image = imagecreatefromjpeg($filePath); // 160mb memory peak

Expected behavior
lower memory usage (near 160MB)

Images

Environment (please complete the following information):

• PHP Version: 8.2
• OS: alpine (k8s)
• Intervention Image Version: 3.3.0
• GD or Imagick: GD

[olivervogel]

Thanks for your report. This is already addressed in #1269

[olivervogel]

reading a 2 MB image using GD uses only 160 MB

Reading a 2 MB image with GD used 5.70 MB in a example I just created. How did you calculate these values?

[a-shafaat]

I'm using symfony framework. The symfony framework has a web GUI profiler with a performance tab that shows peak memory usage and some other related info.

[a-shafaat]

the image i have used:
https://cdn.tizbam.ir/600dpi_Page_1.jpg

[olivervogel]

Thanks. The topic will be pursued further here.

[olivervogel]

One more note: The decoding process of Intervention Image involves more than just opening the image. The following tasks are performed:

• Extraction of EXIF data for appropriate formats
• Adjust the image rotation according to the EXIF data.

In your example image, the image rotation is set to 6 in the EXIF data. This means that the image is rotated once by 270 degrees to make so it is aligned correctly. This can take a lot of memory for large image sizes, especially with GD.

The examples from your code samples are therefore not 1:1 comparable.

[a-shafaat]

thanks for your notes. I think any further processing must be optional and developers should be able to ignore them with a flag or some other way.
in my case, as you may guessed already, I want to do some OMR, and most of the work is done in another environment (python + opencv + ...). In php part of my environment, I only want to resize the picture if it's larger than 3000 pixels in any dimension and convert it to webp.
The Python part has some computer vision capabilities and can recognize the rotations. so exif rotation is a redundancy for me.