You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
But following complications I see immediately which would need to be considered/addressed to make it feasible to validate using current datalad-fuse based solution:
currently we do not have zarr subdatasets installed within dandisets/* dandisets (like 000108). They all just live in their original ../dandizarrs location. So we cannot just easily use dandisets-fuse/ location we have since those subdatasets/locations would be empty
That dandizarrs itself is not even a datalad dataset, so we cannot even use datalad-fuse on that location (Support plain directory (not super-dataset) with datasets under it datalad/datalad-fuse#35) . After that issue addressed I guess we could have implemented some adhoc jumping to another fused location upon getting into not installed submodule but it is "suboptimal" since assumes the correspondence of the git commit for submodule and repo under dandizars/ which most likely would be ok but not necessarily.
The text was updated successfully, but these errors were encountered:
ATM we have /mnt/backup/dandi/dandizarrs/tools/jsonschema-check-zattrs which produced http://www.oneukrainian.com/tmp/dandizarrs-jsonschema-checks.out . That script avoids using fuse since just needs to access a single file. But in the long run I hope that more thorough validation of zarrs would come about (ref: ome/ome-zarr-py#142 (comment) and there on).
But following complications I see immediately which would need to be considered/addressed to make it feasible to validate using current datalad-fuse based solution:
dandisets/*
dandisets (like 000108). They all just live in their original../dandizarrs
location. So we cannot just easily usedandisets-fuse/
location we have since those subdatasets/locations would be emptydandizarrs
itself is not even a datalad dataset, so we cannot even use datalad-fuse on that location (Support plain directory (not super-dataset) with datasets under it datalad/datalad-fuse#35) . After that issue addressed I guess we could have implemented some adhoc jumping to another fused location upon getting into not installed submodule but it is "suboptimal" since assumes the correspondence of the git commit for submodule and repo underdandizars/
which most likely would be ok but not necessarily.The text was updated successfully, but these errors were encountered: