Empty cloned dataset from OSF

[zhao-cy]

Hi there,

I hope to push a datalad dataset onto OSF so that I can clone it elsewhere on another computer. I first refered the Quick Start as it says we can clone the dataset later. There was no warning when creating OSF sibling or pushing to OSF. However the dataset I cloned from OSF is empty. Meanwhile, I tried a more complicated command (with more arguments) and it succeeded. In details:

When creating OSF sibling + pushing - commands that led to failure:

datalad create-sibling-osf --title data4babs_rawBIDS_multises -s osf
# # printed messages:
# create-sibling-osf(ok): https://osf.io/my5b7/
# [INFO   ] Configure additional publication dependency on "osf-storage" 
# configure-sibling(ok): . (sibling)

datalad push --to osf
# all okay, no error, no warning

Then on the OSF webpage, I change the OSF project to public. When cloning on another computer:

datalad clone osf://my5b7
# # printed messages:
# [WARNING] You appear to have cloned an empty repository.                                     
# [WARNING] Cloned /path/to/my/local/folder/my5b7 but could not find a branch with commits 
install(ok): /path/to/my/local/folder/my5b7 (dataset)

ls -a my5b7/
# printed messages:
# .    ..   .git

Basically the cloned dataset from OSF is empty, without files or folders.

Very interestingly, I tried using a more complicated datalad command (ref: here), and it succeeded:

datalad create-sibling-osf --title data4babs_testout -s osf2 --category data --tag reproducibility --public
# # printed messages:
# create-sibling-osf(ok): https://osf.io/fhm8b/
# [INFO   ] Configure additional publication dependency on "osf2-storage" 
# configure-sibling(ok): . (sibling)

I'm wondering if there is anything missing in the simplest version of the datalad create-sibling-osf command I first tried?

Versions:

• datalad 0.17.2
• datalad_osf 0.2.3.1

Thank you so much in advance!
Best,
Chenying

[mih]

The prominent difference I notice is that you have --public in the call that leads to a working sibling. I wonder if setting the project to public via the web UI is not 100% equivalent. I need to investigate that.

[yarikoptic]

FWIW I have tried to clone another fresh (not mine) but random OSF dataset and got similar result

❯ git clone osf://n6tma/
Cloning into 'n6tma'...
warning: You appear to have cloned an empty repository.
❯ git -C n6tma status
On branch master

No commits yet

nothing to commit (create/copy files and use "git add" to track)

[adswa]

But that particular dataset is an annex-only storage sibling, right, @yarikoptic? If so, there can't be any commits on there. I would assume that that particular repo has a sibling on GitHub somewhere. But its an issue that the autogenerated project description on OSF gives instructions for cloning - it makes it sound like that would work

EDIT: I actually don't know what mode that dataset is in - we have the annex key-value pairs, but there is not Git repository representation. From my explorations in #195, I'm not aware that an existing mode causes this, so my assumption is that the XLDRA refs may have been removed by hand. Their lack explains the "empty repository" result, that ultimately boils down to datalad/datalad-next#636

[yarikoptic]

if there is no git presentation, then I would have expected git clone to fail as it does whenever I try to clone something which has no git there, e.g.

❯ git clone https://github.com/datalad/datalad-osf/issues/160
Cloning into '160'...
remote: Not Found
fatal: repository 'https://github.com/datalad/datalad-osf/issues/160/' not found