[ENH] Automatic reference/stacks ordering

[pdedumast]

No description provided.

[sebastientourbier]

@pdedumast After thinking more about the way to reorder scans I think the best way would be to create a new interface which will be inserted right after the DataGrabber. It will takes as input an ordered list of stacks stack_order, a list of input_images (and inputs_masks if manual brain masks are used) and will generate the appropriate list of output_images and output_masks. These outputs could then be connected to the rest of the workflow.

[pdedumast]

@sebastientourbier after performing multiple tests (as you can partially see), I understaood that the different interfaces indeed mess up the order of the stacks between inputs and outputs. I will add a quick sorting of the inputs of each module, as it is done here.

Apart from that, we mentioned that all retrieved stacks by the DataGrabber are preprocessed even though they are not used in the reconstruction. They are even involved in the histogram normalization which does influence the reconstruction...
I can easily "clean" this to preprocess only useful stacks. However, I'm afraid this will mess up all the tests... Any idea to fix this easily?

[sebastientourbier]

@pdedumast Had time to review your PR and check CircleCI. Very nice work 👏
I just let a few comments for typos that should be corrected and then, I guess you are all good to merge 👍

[sebastientourbier]

@pdedumast I just merged PR #47 that integrates the script wrapping call to the Docker command with updated doc.

I also managed to solve:

1. Issue Issue with code coverage #45 for code coverage (seems it was due to running nipype with the MultiProc plugin) and
2. Issue Bad memory allocation issue #48 for bad memory allocation error, i.e the message corrupted size vs. prev_size: 0x000055d4457176f0 *** at the end of the BIDS App execution, by using tcmalloc instead of malloc.

Note
By fixing code coverage, I also disabled Nipype to generate provenance PROVN reports; it was causing issues with non-existing user/group id inside the container. But by doing that, it reduced the number of output files and so, I am sorry to tell you but you will need to update the test-XX_outputs.txt again. 😬

[pdedumast]

@sebastientourbier seems like our milestone for pymialsrtk v2.0.0 is ready! ;)

[sebastientourbier]

Seems all good to me!
Very nice work @pdedumast !
I like the refactoring and the way you output the json file for Super-Resolution. Indeed they might a space in BIDS to add this specification when it gonna get refined by practice.
You can merge 🥳

[sebastientourbier]

Then, I would take of making the release 👍