CiviCRM Community Forums (archive)

My duplicate rules are as follows

Last name = 7
Email = 10
City = 10

Threshhold = 15

The result it gives includes lots where none of these occur. For example
Last name = Blennerhassett
Email=eliot@blennerhassett.gen.nz
City = Christchurch

when I click select , the people it offers to merge with are

Last name = Black
Email= - blank -
City = Christchurch

and

Last name - Dann
Email = christine@horomaka.org   
City = christchurch


Have i lost the plot or are things going awry?

I created these three contacts in my 2.0 sandbox, altered the default rule to match your above and no duplicates were found. The three were found after lowering the threshold to 10, so it works for me properly…

Hm, maybe I’m missing something obvious. Can you replicate this issue on the demo (i.e., create the three contacts, setup the rules and let me know whether it works as it should or as in your install)?

If it works differently on demo and in your setup, I can investigate further if I have your database dump (if that’s an acceptable option for you). If so, please send it to shot@civicrm.org (you can encrypt the email with my GPG key of D128F14A if you wish).

So I just set Dupe Rules in Demo as above and ran it. gave me duplicate match for #128 and #129 which I can see have same Email address but their Last and City are different - ah ha, or are they, when I choose 'city' which 'city' does it use? primary? any?

Aaargh, right. It uses any city, as it matches the city against a contact id, not against a (contact id, location type) pair. I assume the latter would make more sense – i.e., consider this a valid match iff the location types match as well.

(That’s what you end up with when the default setting is just one location per contact, period; you miss such situations easily.)

I see that in the case above both show no home address though it is set as primary - is it matching 'city' because they are both blank because their 'city' details are in their 'billing' field?

It’s matching them because these two contact ids share the same city in the address table.

I’m thinking now whether it would make more sense (use-case-wise) that the requirement was that the location types match or that the primariness of the address matches… Thoughts?

Any chance you can make the fields that are used in the Duplicate Rule show as columns in the "Potentially Duplicate Contacts" table? That would certainly help in spotting which contacts to merge.

Hm, that’s an idea worth considering – and I like it. I’ll see how it flies with the default rules we’ll ship in CiviCRM 2.1.

It certainly makes it harder when they don't even appear in a-z order - what order are they showing in? Are they divided up depending on which bits of the rule caused the match to be identified?

No, I’m afraid they show up in the order the dedupe engine returns them, which you can assume is more or less random. I’ll make sure it’s fixed in CiviCRM 2.1.