Hi all,

I cannot reproduce this problem at all.  I do see the issue with TUs not really being 100% duplicate as most of the ones returned in the potential duplicate search are slightly different.

So I had a little play with a SQLite editor as I think the only way to do a smart deletion of the true duplicates with null field values would be to investigate this.  Dangerous I think, but once you had the query probably simple enough to run whenever you needed it.  Now, I don't know my way around SQL queries well enough to do this yet, but I think this is an interesting approach.  For example, I can run this script:

And it returns all the 100% source duplicates based on there being 2 or more of them like this:

Quite a few duplicates based on source segments only.  But I think if this script was extended so it did these things:

1. reported on duplicates in source AND target only

2. filtered these to only list those with empty field values

3. deleted them

Then you'd have what you need.  Perhaps someone in the community is already familiar with writing SQL scripts for these sorts of things and can help to develop it further.  Could be interesting to have a bunch of scripts for doing other maintenance operations as well where it would be possible through SQL to do all kinds of complex queries with no limitations imposed by software.

Maybe not a solution for everyone, and definitely back up your TM before trying this, but I like this idea.

Regards

Paul