fix #10175: set FilesetEntry client path to String <256 chars

[mtbc]

Fixes http://trac.openmicroscopy.org.uk/ome/ticket/10175. To test, try importing images from absurdly long-named/deep paths. Cc: @manics.

--no-rebase as it's FS-specific.

[mtbc]

If some of the used files have a shorter full path than the longest ones, this code takes the opportunity to squeeze an extra bit of depth into them. If it is important to start all the used files from the same parent, this PR could be changed accordingly.

[manics]

Should this be defined as a constant/property somewhere? Or is there no point since everything else is hard-coded?

[mtbc]

I was hoping to figure out how to extract it from whatever's setting the VARCHAR(255) on that column in the first place, but I couldn't find it.

[manics]

Fair enough.

[manics]

Is it worth warning the user that the path was truncated?

[mtbc]

I don't think so: they probably won't care about the higher-up path components anyway and we already bother them with quite enough things. One could slip a ".../" at the front if you think it warranted?

[joshmoore]

I think clientPath must be inviolate since we've sold it as the way to identify filesets despite our munging of server-side paths, even if that means we have to go to using text. /cc @chris-allan, @jburel

[joshmoore]

NB: Another alternative would be to log all these values in a property file with the fileset:

2013-11-04-fs059
2013-11-04-fs059.log
2013-11-04-fs059.client_paths

in which `client_paths includes:

2013-11-04-fs059/foo/bar.txt=c:\Users\Moore\foo\bar.txt
...

[jburel]

The main issue is the image name which is limited to 256.
The current strategy is a major change between pre-fs and FS. the image name == client path. This point was raised during Beta1 but we did not have a chance to look at that.

[joshmoore]

If image.name == entry.clientPath is the problem, then I certainly don't have any objections to changing the image.name strategy, but the modification here is truncation of entry.clientPath, which I see as problematic.

[jburel]

I agree that we cannot truncate the clientPath, I did not save my last comment.

[mtbc]

Unacceptable approach.

[mtbc]

Superseded by #1710.