sql - Limit results to x groups -

I am developing a system using Trek, and I want to limit the number of "changelog" entries . The point is that Trac connects these entries with several tables using a union , and later adds them into single 'changes' based on their timestamp. I want to limit the results to the latest I 3 changes, but I need to recover several rows until it is needed unless I have 3 unique timestamps. Solution needs to work for SQLite / Postgres.

Current SQL Results

  NEWVALUE Permanent OLDVALUE Time User Field ========== ================================================== =========== 1371806593507544 1 1371806593507544 A comment 2 lipsum bc One owner 1 1371806593507544 Description foo times 1 1371806593324529 b Comment Hello world 1 1371806593125677 c Priority minor minor 1 1371806592492812 d comment xy 1   
  SQL results (like 1 timestamp to be limited) with intent   
  time nEWVALUE lasting == OLDVALUE User field ======= ==== ================================================= ======== 1371806593507544 1 1371806593507544 One Comment 2 lipsum 1 1371806593507544 Description foo bar 1    
 
 
 As you've already said on your own, this result can not be solved in SQL due to an indefinite number of results. And I think that this is also not necessary.  
 You can use a slightly modified template to get what you want. Change  
  & lt; Div id = "changelog" & gt; & Lt; Py: Each = "Changes in Changes" & gt;   
 In  
  & lt; Div id = "changelog" & gt; & Lt; Py: Each = "Changes in Changes [-3:]" & gt;   
 and the file should be  and lt; Env & gt; Restart your web-server in / template /  but look for a change of ticket, whenever you try to upgrade your track's upgrade. Each time you do this, you may need to re-apply this change to the current template of the related edition. But the IMHO track core code is still much faster and cleaner than the patch.