---- Reported by adam@yorba.org 2012-08-10 09:38:00 -0700 ----

Original Redmine bug id: 5644
Original URL: http://redmine.yorba.org/issues/5644
Searchable id: yorba-bug-5644
Original author: Adam Dingle
Original description:

I think address autocompletions should be arranged with the people I email
most often at the top. (The pending autocompetion patch does not appear to
work this way.)

Related issues:
blocked by geary - Feature #4284: autocomplete addresses in To, Cc, Bcc fields (Fixed)



---- Additional Comments From geary-maint@gnome.bugs 2012-08-10 12:50:00 -0700 ----

### History

####

#1

Updated by Matthew Pirocchi over 1 year ago

In the current patch, addresses are sorted as follows:

  1. First order by importance. People you have emailed appear at the top, then people you have CC's, etc. See Geary.ContactImportance for the full importance ranking.
  2. Then order by lexicographically by real name (considering no name as first in lexicographical order).
  3. Then order by email address.

For considering email frequency, I think it would be best to put it second. In
other words:

  1. First order by importance. People you have emailed appear at the top, then people you have CC's, etc. See Geary.ContactImportance for the full importance ranking.
  2. Then order by email frequency (details on this below).
  3. Then order by lexicographically by real name (considering no name as first in lexicographical order).
  4. Then order by email address.

The being said, "email frequency" is somewhat complicated to define. Recall
that "I emailed them" is not the only reason that a contact might appear in
autocompletion. Here are some potential definitions of email frequency:

  1. The number of messages that the contact has appeared in.
  2. The number of messages that the contact has appeared in at the highest importance level. In other words, if the "highest importance" that the contact was seen at was "they emailed me", the email frequency would be "the number of times that they emailed me".
  3. A weighted combination of the two. Example below.

I don't think #1 is a good idea. You could imagine that someone on the same
mailing list as you (from a naive sender who didn't use BCC) would occur far
more often than anyone else.

#2 is _ok_, but what if they only appear once in FROM_TO, but appear many
times in FROM_CC? The FROM_CC occurences would be ignored.

Regarding #3, these are our current importance levels ("FROM_TO" means "My
address appeared in 'from'/'sender', their address appeared in 'to'"):

    
    
    public enum Geary.ContactImportance {
        FROM_FROM = 100,
        FROM_TO = 90,
        FROM_CC = 80,
        TO_FROM = 70,
        TO_TO = 60,
        TO_CC = 50,
        CC_FROM = 40,
        CC_TO = 30,
        CC_CC = 20,
    
        VISIBILITY_THRESHOLD = TO_TO;
    }
    

Email frequency could be calculated as:

    
    
    CC_CC * the number of times that the contact appeared at the CC_CC importance level +
    CC_TO * the number of times that the contact appeared at the CC_TO importance level +
    ...
    

While this approach is slightly more complicated, I think some variation of it
would give the best results.

Any thoughts?

####

#2

Updated by Matthew Pirocchi over 1 year ago

Just to clarify: We currently only keep track of the **highest** importance
level that a contact has appeared in.

####

#3

Updated by Adam Dingle over 1 year ago

  * **Target version** deleted (<strike>_0.2_</strike>)



--- Bug imported by chaz@yorba.org 2013-11-21 20:26 UTC  ---

This bug was previously known as _bug_ 5644 at http://redmine.yorba.org/show_bug.cgi?id=5644

Unknown Component 
   Using default product and component set in Parameters 
Unknown version " in product geary. 
   Setting version to "!unspecified".
Unknown milestone "unknown in product geary. 
   Setting to default milestone for this product, "---".
Setting qa contact to the default for this product.
   This bug either had no qa contact or an invalid one.
Resolution set on an open status.
   Dropping resolution