On 4/23/06, Alec Thomas <[EMAIL PROTECTED]> wrote: > On Sat, Apr 22, 2006 at 03:51:51PM -0400, Sergey Lipnevich wrote: > > ... A better UI would > > somehow limit choices in such a list to a manageable number, say 10-20 > > most suitable tickets. A "short list" may be constrained by a search > > word (harder to do nicely without AJAX stuff) or by similarity. For > > example, picking 10 tickets that have the same words as this ticket in > > their summary would be a first approximation. So, if I want to find a > > parent ticket for something called "Workflow incorrectly closes > > tickets," Trac behind the scenes would do a query for these words, > > find tickets that have words "workflow" or "close" in their summary, > > and present ten most suitable ones. > > This is actually already implemented in WorkFlow [1]. It was primarily > designed for detecting duplicate tickets, but the code could easily do > double duty. > > That being said however, it is quite a difficulty problem to solve, as > its rare that users actually enter tickets with similarities, even if > the tickets are for the same issue. > > The current code uses words common between tickets, weighted by their > length, to detect similarity. > > [1] http://projects.edgewall.com/trac/changeset/3111#file21
i'm keen to test that on real world examples :) finding duplicates and depending tickets would be a time saver especially for bigger projects. -solo. _______________________________________________ Trac-dev mailing list [email protected] http://lists.edgewall.com/mailman/listinfo/trac-dev
