address completion issues in notmuch-emacs

address completion issues in notmuch-emacs

[Rollins, Jameson]

Hi.  I'm using address completion in notmuch-emacs, but I keep having
problems.  No matter what "Notmuch Address Internal Completion"
customization configuration I use ("sent" or "received") there are tons
of missing addresses that I need.  It seems, though, that the two
configurations might be compliments of each other, and if I could use
both sent *and* received then maybe I would get all the addresses I
need.  Is there some technical reason why the completion doesn't just
use both?

I don't understand the parenthetical comments around these options
either:

( ) sent (more accurate)
(*) received (faster)

Why would they have such performance differences?  And does it really
matter at all?  My understanding is that the addresses are collected
once, so a performance difference at collection seem irrelevant to me,
since I only care about performance when I'm trying to do the actual
completion.

jamie.

Re: address completion issues in notmuch-emacs

[David Bremner]

"Rollins, Jameson" <jrollins@caltech.edu> writes:

> Hi.  I'm using address completion in notmuch-emacs, but I keep having
> problems.  No matter what "Notmuch Address Internal Completion"
> customization configuration I use ("sent" or "received") there are
> tons of missing addresses that I need.  It seems, though, that the two
> configurations might be compliments of each other, and if I could use
> both sent *and* received then maybe I would get all the addresses I
> need.  Is there some technical reason why the completion doesn't just
> use both?

I don't think so (at least not as an option). It would need an update to
the CLI, or calling "notmuch address" twice.

> I don't understand the parenthetical comments around these options
> either:
>
> ( ) sent (more accurate)
> (*) received (faster)
>
> Why would they have such performance differences?

This as has to do with what is stored in the database (explained in notmuch-address(1)).

> And does it really matter at all?  My understanding is that the
> addresses are collected once, so a performance difference at
> collection seem irrelevant to me, since I only care about performance
> when I'm trying to do the actual completion.

That probably depends

1) whether you use the caching. This is off by default due to
   (mild) privacy concerns. It makes another copy of your addresses,
   e.g. on your laptop for people using remote notmuch.

2) if the answer to (1) is no, how often you restart emacs.

But I'd entertain the idea of making "both" default, if backed by some
timing experiments.

Before getting too far into coding/design, it would be nice to know if
the union of those two options contained all the addresses you were
looking for.

Note that there is some other discussion about improving
notmuch-address:

        https://github.com/aperezdc/notmuch-addrlookup-c/issues/23

I _think_ that is mainly about better sorting, but I'm not 100%
sure.