[ome-users] Deleting an object together with its directory in ManagedRepository

[Mark Carroll]

On 11/16/2017 12:28 PM, Andrii Iudin wrote:

> Thank you for explaining the current situation. I will check with my
> supervisor if we can allocate resources to implementing this, however,
> please point to the relevant code.

When a file is deleted by normal means then the row from the
originalfile table in the database is removed by the execute() in,

https://github.com/openmicroscopy/openmicroscopy/blob/v5.4.0/components/blitz/src/omero/cmd/graphs/Delete2I.java#L170

This execute method applies the deleteInstances in,

https://github.com/openmicroscopy/openmicroscopy/blob/v5.4.0/components/blitz/src/omero/cmd/graphs/BaseGraphTraversalProcessor.java#L54

Later the underlying files are deleted by buildResponse in,

https://github.com/openmicroscopy/openmicroscopy/blob/v5.4.0/components/blitz/src/omero/cmd/graphs/Delete2I.java#L199

These are deleted from the bottom of the tree up. There is a check at,

https://github.com/openmicroscopy/openmicroscopy/blob/v5.4.0/components/dsl/resources/ome/dsl/psql-footer.vm#L849

that a directory may not be deleted unless it is empty so back at,

https://github.com/openmicroscopy/openmicroscopy/blob/v5.4.0/components/blitz/src/omero/cmd/graphs/Delete2I.java#L155

we rearranged the list of things to delete so that the "leaves" get
deleted first. It is the value returned from here that was fed to
buildResponse above.

Separately you may find it helpful to see how we implement
http://downloads.openmicroscopy.org/omero/5.4.0/api/slice2html/omero/grid/Repository.html#deletePaths
(when recursive==true) in,

https://github.com/openmicroscopy/openmicroscopy/blob/v5.4.0/components/blitz/src/ome/services/blitz/repo/PublicRepositoryI.java#L291

Depending on your needs, you may find it easiest to simply make
client-side adjustments -- whether to OMERO.cli or OMERO.web or your own
scripts or whatever -- that follows something like this path:

1. Do the Delete2 cmd with "dryRun" set to find which OriginalFile
instances are to be deleted.
2. Query the database for the path of those instances.
3. Do the delete from (1) for real, without dryRun.
4. Query the database for any OriginalFiles that have the same path as
anything from (2).
5. For anything from (2) for which (4) returns no results (i.e. that
directory is now empty), shunt the last component of that path over to
the "name" property so you can query for the OriginalFile for that
parent directory.
6. For each of (5)'s parent directories where canDelete() == true add it
to a new to-delete list.
7. Start from (1) with your new to-delete list.

It's neither fast nor pretty, but it's at least easier than hacking the
server's Delete2I directly.

-- Mark

The University of Dundee is a registered Scottish Charity, No: SC015096