Re: [ANNOUNCE] mailaprop: modern popup-style autofill for email addresses

[Karl Fogel <kfogel@red-bean.com>]

"John Wiegley" <johnw@gnu.org> writes:
>KF> Hi. I wrote a new package, 'mailaprop', that provides popup-style autofill
>KF> for email addresses in recipient headers of message composition buffers.
>
>If you want a nice source of candidates, I wrote this unpublished package:
>
>https://github.com/jwiegley/dot-emacs/blob/master/lisp/gnus-harvest/gnus-harvest.el
>
>It watches who you write to, and who you read e-mail from (if you hook it into
>Gnus), and stores them all in an SQLite database with timestamps. It then lets
>you query for a candidate list. I use it via message-x.

Thanks, John.  That looks handy.

To be useable by mailaprop, it would need to remember the datestamp of the message the address was harvested from (or at least remember the date of the most recent message that address was harvested from), and keep separate counts for how often the address had been sent to and how often it had been received from.

This is because the order in which addresses are offered in autofill matters.  The most likely candidate should be at the top, the next-most-likely underneath it, and so on.  Right now, the inputs to the likelihood-calculating algorithm are the last-seen date, the total sent-to count, and total received-from count.  I could make things fancier later, but for now those are the inputs.

Would those pieces of information be useful for message-x and other integratees of gnus-harvest too?  Obviously, that information would be pretty easy to harvest.  If mailaprop and message-x could share inputs, that'd be nice.

By the way, mailaprop's current solution to the problem of gathering input data is a standalone Python script.  It crawls all your mbox files (other formats to be supported in the future, I hope) and builds the autofill database from that.  I've just added a paragraph to the README.md explaining that it could take input from other sources too, and linking to this discussion.

Best regards,
-Karl