DeprecationWarning for file_get and capture_preview when not passing a CameraFile

[mgineer85]

Hi @jim-easterbrook ,

in latest changelog I found I should not pass a CameraFile to file_get and capture_preview.
But when I do not, I get DeprecationWarnings which claim I should pass a CameraFile.

Example:

NOT passing a CameraFile like this example

camera_file = self._camera.file_get(file_to_download[0], file_to_download[1], gp.GP_FILE_TYPE_NORMAL)  # pyright: ignore [reportAttributeAccessIssue]
camera_file.save(str(filepath))

results in a deprecation warning

src/tests/tests/backends/test_linux_gphoto2.py::test_get_images_gphoto2
  /media/michael/Dev/dev/photobooth/photobooth-app/src/photobooth/services/backends/gphoto2.py:322: DeprecationWarning: In function Camera_file_get a CameraFile object should be passed as an in/out parameter.
    camera_file = self._camera.file_get(file_to_download[0], file_to_download[1], gp.GP_FILE_TYPE_NORMAL)  # pyright: ignore [reportAttributeAccessIssue]

When passing a CameraFile in agreement with the changelog for v2.6.1, I get no deprecation warning:

camera_file = gp.CameraFile()  # pyright: ignore [reportAttributeAccessIssue]
self._camera.file_get(file_to_download[0], file_to_download[1], gp.GP_FILE_TYPE_NORMAL, camera_file)  # pyright: ignore [reportAttributeAccessIssue]
camera_file.save(str(filepath))

I am using v2.6.1 - am I missing something obvious here?

[mgineer85]

Also the automated testing check for emitting the deprecated warning, but it is in contrast to the changelog. Which is the correct way?

[jim-easterbrook]

The change log says what I intended to do, the software (and my test modules) do the opposite! I've no idea how I got that wrong.

Reading back over my changes I think I intended to make the Python interface more like the C API, where you have to pass a CameraFile, but this is less convenient for the Python programmer. It's not "Pythonic" to create an empty object and have a function fill it in.

[mgineer85]

Ha, ok. Great! I love the way it should be according to the changelog way more :)

So I use it without CameraFile again and just ignore the warnings assuming they vanish in a future release :)

Thank you for the quick clarification, sending good weekend vibes 🏝️

[jim-easterbrook]

I've created a bug (#192) from this discussion and committed a fix.

I think my reasoning was that it's ever so slightly more efficient to reuse an existing CameraFile object than creating a new one each time. But I've decided that the "Pythonic" way of doing things wins from now on.

[jim-easterbrook]

An aside to the above: file_get will fail on very large files as it needs too much memory to fetch the file in one chunk. (My camera will record video files up to 4 GByte.) I now use camera.file_read to download from the camera, with a bytearray buffer of up to 32 MByte.

                buf = bytearray(min(info['size'], 32 * 1024 * 1024))
                try:
                    with open(info['dest_path'], 'wb') as of:
                        offset = 0
                        while offset < info['size']:
                            count = camera.file_read(
                                info['folder'], info['name'],
                                gp.GP_FILE_TYPE_NORMAL, offset, buf)
                            of.write(buf[:count])
                            offset += count
                except Exception:
                    os.unlink(info['dest_path'])
                    raise

[mgineer85]

Oh, good to know. I would like to get videos at one point from the camera, but not yet there…