image.convert function is called incorrectly, causing error during registration

[joshuacwnewton]

A user on the forum is reporting an error: https://forum.spinalcordmri.org/t/register-template-to-mt1/654

The bug is from this line of code:

https://github.com/neuropoly/spinalcordtoolbox/blob/0b7be07b7e64a57a303824e85930fbceb8b890f0/spinalcordtoolbox/registration/register.py#L373

The usage of this function doesn't match its parameters:

https://github.com/neuropoly/spinalcordtoolbox/blob/0b7be07b7e64a57a303824e85930fbceb8b890f0/spinalcordtoolbox/image.py#L1341-L1348

I believe image.convert came about in #2871/#2812. I'll dig into that and find out what the correct usage should be.

[joshuacwnewton]

Ah, I see what happened. There are multiple functions named convert in SCT:

• image.convert
• sct_convert.convert

When sct_image was refactored in #2812, the function sct_convert.convert was mistaken for image.convert.

[joshuacwnewton]

The two functions are basically identical. sct_convert.convert simply adds file I/O to image.convert, which doesn't really justify a function.

https://github.com/neuropoly/spinalcordtoolbox/blob/0b7be07b7e64a57a303824e85930fbceb8b890f0/spinalcordtoolbox/scripts/sct_convert.py#L70-L79

Because scripts should be importing from API (image.convert) and not the CLI (sct_convert.convert), I'm going to remove sct_convert.convert and replace all calls with image.convert. That way, there will be no confusion in the future.

[Drulex]

Woops this would be my fault.

The two functions are basically identical. sct_convert.convert simply adds file I/O to image.convert, which doesn't really justify a function.

The initial motivation to keep sct_convert.convert() was to maintain backwards compatibility while using the refactored image.convert() API internally until we are ready to break the interface. (some scripts were importing sct_convert.py from scripts at that point, but this was fixed in #2966)