BUG: Path on Import selection

[pwalczysko]

During testing of Webtagging app, on Gretzky, found out that the selection of Path into the name of the imported image does not work on 5.0.0 at all. All the images have truncated names.
Using Insight, started Importer, importing to Gretzky.

• Select a .dv image from my Desktop (local machine) and add to the Queue
• Select the Options tab in Importer and choose "Full path".
• Import the image
• Check it in Insight and Web -> the name is not full path, just the name of the image itself is shown -> not expected
• The exactly same workflow works fine on 4.4 (import to HOwe)

Putting as a blocker, because this is a blocker for @drussell-x

[mtbcarroll]

If it is useful as an emergency stop-gap, the original full paths are actually still there in the model objects. From SQL,

omero=# select i.name, f.clientpath from image i, pixels p, originalfile o, filesetentry f where o.name = p.name and o.path = p.path || '/' and o.repo = p.repo and o.id = f.originalfile and i.id = p.image;
   name    |           clientpath            
-----------+---------------------------------
 brain.dcm | home/mtbc/Win7/images/brain.dcm
(1 row)

There is probably a nicer way to do this query via the fileset table though.

[drussell-x]

This is very serious for us/me as a lot of crucial information is encoded in the path which webtagging then uses. With big release of webtagging along with the other consortium applications it would be good to get this fixed as soon as possible so anyone trying it can actually use this feature with newly imported data.

[jamoore]

Dominik: do you have time to do an initial investigation?

[mtbcarroll]

I should mention that it was pointed out to me that the link from image to file suggested by my above query isn't as strong as I'd thought so you can ignore the pixels table, i.e.,

$ omero hql 'select i.name, fe.clientPath from FilesetEntry fe, Image i where fe.fileset = i.fileset'
Using session 6f5a6d85-765c-42b8-8ce8-27fdbb843655 (root@localhost:4064). Idle timeout: 10.0 min. Current group: system
 # | Col1      | Col2                            
---+-----------+---------------------------------
 0 | brain.dcm | home/mtbc/Win7/images/brain.dcm 
(1 row)

The fileset entries should be available through API queries as normal model objects and cannot be affected by users in the client.

[wmoore]

Douglas, for OMERO 5 images, you may be able to use the clientpath directly, when you load all your images, instead of using image Name.
This would also be useful even AFTER the import path bug is fixed, since it doesn't rely on the user remembering to use the whole path in the name. You can get the paths for an image like this

params = omero.sys.ParametersI()
params.addLong("iid", imageId)
query = "select fse from FilesetEntry fse join fse.fileset as fs "\
        "left outer join fs.images as image where image.id=:iid"

r = qs.findAllByQuery(query, params, conn.SERVICE_OPTS)

for i in r:
    print i.clientPath.val

This query can probably be improved to load multiple images with their associated paths, but I haven't had time to optimise it yet. Let me know if you need any more help here...

[dlindner]

Replying to jamoore:

Dominik: do you have time to do an initial investigation?

Ok, I'll see if I can find/fix the problem.

[dlindner]

Found the reason, problem was probably introduced with bugfix
​https://trac.openmicroscopy.org.uk/ome/ticket/10175
respectively ​https://trac.openmicroscopy.org.uk/ome/ticket/10953
Still working on a solution

[dlindner]

Found the reason, problem was probably introduced with bugfix
​https://trac.openmicroscopy.org.uk/ome/ticket/10175
respectively ​https://trac.openmicroscopy.org.uk/ome/ticket/10953
(see ImportCandidates?.java:427)
Still working on a solution

[mtbcarroll]

Looking at the description of #10953 that looks very likely to be the problem: needed awareness of that option on the import options tab that changes the image naming behavior.

[dlindner]

Fixed by: ​https://github.com/openmicroscopy/openmicroscopy/pull/2164