User Agent: Mozilla/5.0 (Windows NT 6.3; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/40.0.2214.94 Safari/537.36 Steps to reproduce: When composing email and filling To field. Actual results: Autocomplete list suggested addresses in random order (barely used e-maisl on beginning, frequently used somewhere down the list) Expected results: Suggested e-mails in order by usage frequency. It worked in older versions but stop working after some upgrade (unfotunately I'm not sure which one)
We first list the best result by matching the start of words. If there's a tie, frequency is used to determine order.
Well in my case it is not working lůike that. For example for search "olga" i got: koreckova.olga@... olga.mitasova@... olga.seda108@... o.zelinkova@... olga.mitasova@... <-- most used e-mail cieslarovaolga@... All of them match olga but order is wrong.
I think you listing that without the display names? You need to look at those too. For instance o.zelin... wouldn't match olga.
Yes, I listing in without names. With names it looks like this: Olga KOrečková <koreckova.olga@...> olga.mitasova@... olga.seda108@... Olga Zelinková <o.zelinkova@...> olga.mitasova@... <-- most used e-mail cieslarovaolga@... But it doesn't change fact that all of them match but order is wrong.
The first 5 seems would have the same score. It's not possible to see the PopularityIndex, but that's the order according to the code (implemented in bug 970456). Note that the popularityindex do not age, so it's it's an all-time-mailed-count kind of thing. Also note: number 2 and number 5 are the same in your listing (though I guess they may have different domain).
To address on first place I sent only 1 e-mail few years back. To e-mail on 5th place I sent 5 to 20 e-mail a day. Yes there is different domain for item 2 and item 5. It looks to me that popularity index is either not used or somehow damaged.
I'm starting to wonder if we could just expose the popularityIndex as an editable field in card properties. That way, users can both check and correct the actual popularity of their cards which they claim are unpopular but still get toplisted, most likely because they are popular but users aren't aware.