query by example (e.g. when creating a new ticket, check for similar tickets)

[Christian Boos]

There could be some ways to reduce the amount of duplicate tickets getting created. The simplest way would be to advertise doing a search:

E.g (any better english is welcome, of course):

Create New Ticket

Have you checked that the issue you want to report was not already raised before? If not, please Search for existing tickets describing your problem before creating a new one.

Your email or username: …

---

A more sophisticated way would be to do a search (in the tickets) based on the content of the Short summary field and the Keywords field, when the reporter Previews his new ticket. The result of the search would also appear on the ticket preview page. From that point, the reporter could choose either to go on with the creation of his new ticket (maybe clarifying what distinguishes it from other, similar tickets), or to contribute to existing similar tickets if some are found and relevant for him.

[Christopher Lenz]

I agree, automatically searching for dupes would be cool.

[Matthew Good]

#1595 has been marked as a duplicate of this ticket

[Gunnar Wagenknecht <gunnar@…>]

The feature would be nice but should be optional. I didn't like the feature when working with Scarab. In 80% of the case the found duplicates were absolutly unrelated to the entered ticket. Looks like the search algorithm needs a lot attention for such a feature.

[Christian Boos]

Adds a Search Similar… button in the new ticket page, that redirects to an adequate custom query

[Christian Boos]

What about using the newticket form as an alternative way to prepare a custom ticket query?

See attachment:search_for_similar_tickets_r3013.diff​

The above patch also adds the possibility to search for the Description field in the query page.

[Alex Buccino <alexander.buccino@…>]

The utility of this feature really hinges on your definition of "similar". I've done some work in document similarity analysis and I would love to see something like that in Trac. However, to have real utility, I think it would need to be built on a probablistic search engine, like Xapian. Roundup is using this search library, but I don't think it has a "find similar" function.

[Christian Boos]

It seems that Alec has started something like this in the WorkFlow branch (see r3111). My approach was more a kind of query by example…

[Alec Thomas]

The similarity detection is now exposed as a plugin interface, so it can be extended. The default one is basic.

[Christian Boos]

#2775 was a similar request for existing tickets. The patch could be easily adapted to allow for that. I'll look into making an updated version of the patch.

[Christian Boos]

Instead of "Search Similar" which has the connotation of requiring advanced AI (as in alect proposal), this could be renamed e.g. "Build Query from Ticket"…

[ThurnerRupert]

see

• ​NewTicket duplicates discussion
• ​ticket cuplicate discussion
• TracDev/Proposals/NewTicketDuplicateCheck

[Christian Boos]

This could be done as a plugin (similar to the clone ticket​ one + the _do_search in attachment:search_for_similar_tickets_r3013.diff​).

[Christian Boos]

#8096 was closed as duplicate.

[Christian Boos]

Milestone 2.0 deleted

[figaro]

A plugin that supports the original request is provided in ​DuplicateTicketSearchPlugin. This plugin searches within the summary field only.

Another plugin that may be helpful, depending on local processes, is ​TracDupPlugin.