address completion when composing

address completion when composing

[Florian Friesdorf]

[-- Attachment #1: Type: text/plain, Size: 1009 bytes --]


According to the docs [1], there are two programs that can be used with
notmuch-address.el: notmuch-addresses (python) and addrlookup (vala/c).

Further it is possible to import all addresses into bbdb [2], supposedly
faster than notmuch-address.el.

Is it also faster than notmuch-address.el + addrlookup? 

What about shipping some / all of these with notmuch? It feels that
newbies would benefit from this inclusion and more extensive
documentation about what to do to get address completion.

Probably it would be sane to decide either for notmuch-addresses or
addrlookup, as they seem to do the same.

In addition the import script for bbdb could be nice.

[1] http://notmuchmail.org/emacstips/#address_completion
[2] http://notmuch.198994.n3.nabble.com/bbdb-import-notmuch-addresses-td520249.html
-- 
Florian Friesdorf <flo@chaoflow.net>
  GPG FPR: 7A13 5EEE 1421 9FC2 108D  BAAF 38F8 99A3 0C45 F083
Jabber/XMPP: flo@chaoflow.net
IRC: chaoflow on freenode,ircnet,blafasel,OFTC

[-- Attachment #2: Type: application/pgp-signature, Size: 835 bytes --]

Re: address completion when composing

[Florian Friesdorf]

[-- Attachment #1: Type: text/plain, Size: 1382 bytes --]

On Wed, 11 May 2011 14:18:01 +0200, Florian Friesdorf <flo@chaoflow.net> wrote:
> 
> According to the docs [1], there are two programs that can be used with
> notmuch-address.el: notmuch-addresses (python) and addrlookup (vala/c).
> 
> Further it is possible to import all addresses into bbdb [2], supposedly
> faster than notmuch-address.el.
> 
> Is it also faster than notmuch-address.el + addrlookup? 
> 
> What about shipping some / all of these with notmuch? It feels that
> newbies would benefit from this inclusion and more extensive
> documentation about what to do to get address completion.
> 
> Probably it would be sane to decide either for notmuch-addresses or
> addrlookup, as they seem to do the same.
> 
> In addition the import script for bbdb could be nice.
> 
> [1] http://notmuchmail.org/emacstips/#address_completion
> [2] http://notmuch.198994.n3.nabble.com/bbdb-import-notmuch-addresses-td520249.html

In addition, about a year ago, Sebastian indicated, that notmuch itself
could offer returning email addresses [3]. Are there arguments against
or just nobody found the time to do it?

[3] http://article.gmane.org/gmane.mail.notmuch.general/1911
-- 
Florian Friesdorf <flo@chaoflow.net>
  GPG FPR: 7A13 5EEE 1421 9FC2 108D  BAAF 38F8 99A3 0C45 F083
Jabber/XMPP: flo@chaoflow.net
IRC: chaoflow on freenode,ircnet,blafasel,OFTC

[-- Attachment #2: Type: application/pgp-signature, Size: 835 bytes --]

Re: address completion when composing

[Sebastien Binet]

[-- Attachment #1: Type: text/plain, Size: 1118 bytes --]

On Wed, 11 May 2011 14:18:01 +0200, Florian Friesdorf <flo@chaoflow.net> wrote:
> 
> According to the docs [1], there are two programs that can be used with
> notmuch-address.el: notmuch-addresses (python) and addrlookup
> (vala/c).

for completeness...
there is also notmuch-addrlookup in go:
http://git.notmuchmail.org/git/notmuch/tree/HEAD:/bindings/go/cmds

(which is broken with the latest go release r57.1 because one of the 3rd
party package (to parse config files) is broken...)

> 
> Further it is possible to import all addresses into bbdb [2], supposedly
> faster than notmuch-address.el.
> 
> Is it also faster than notmuch-address.el + addrlookup? 
presumably: crawling thru the db is I/O bound...

> What about shipping some / all of these with notmuch? It feels that
> newbies would benefit from this inclusion and more extensive
> documentation about what to do to get address completion.
> 
> Probably it would be sane to decide either for notmuch-addresses or
> addrlookup, as they seem to do the same.
> 
> In addition the import script for bbdb could be nice.
+1

-s

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: address completion when composing

[Jesse Rosenthal]

On Wed, 11 May 2011 14:18:01 +0200, Florian Friesdorf <flo@chaoflow.net> wrote:
> 
> According to the docs [1], there are two programs that can be used with
> notmuch-address.el: notmuch-addresses (python) and addrlookup (vala/c).
> 
> What about shipping some / all of these with notmuch? It feels that
> newbies would benefit from this inclusion and more extensive
> documentation about what to do to get address completion.

I certainly agree with this. I've done a pretty craptacular job of
maintaining the python version (just now changed the URL on the wiki,
though, to be fair to myself, I didn't know it was on the wiki in the
first place).

I'd personally argue, not surprisingly, for unifying on the python
version, if unifying we must, just because the python bindings are much
more used than vala bindings (and way more people hack on python than
vala) so it stands a better chance of keeping up with the binary. I
don't see speed as being too much of an issue (either between the two,
or in relation too BBDB), since both are well under a second on my
underpowered netbook and my 6-year-old celeron.

Sebastian might have other opinions, though.

Ideally, though, it would be best if it were part of the bin from the
get-go. Or, even better, if the command-line options made duplicating
the functionality possible.

Best,
Jesse

Re: address completion when composing

[Jameson Graef Rollins]

[-- Attachment #1: Type: text/plain, Size: 427 bytes --]

On Wed, 11 May 2011 14:31:36 +0200, Florian Friesdorf <flo@chaoflow.net> wrote:
> In addition, about a year ago, Sebastian indicated, that notmuch itself
> could offer returning email addresses [3]. Are there arguments against
> or just nobody found the time to do it?

As with pretty much everything in notmuch, manpower time is probably the
issue.  I think that if you wanted to get this working that would be
great.

jamie.

[-- Attachment #2: Type: application/pgp-signature, Size: 835 bytes --]

Re: address completion when composing

[Sebastian Spaeth]

[-- Attachment #1: Type: text/plain, Size: 1555 bytes --]

On Wed, 11 May 2011 10:49:50 -0400, Jesse Rosenthal <jrosenthal@jhu.edu> wrote:
> > What about shipping some / all of these with notmuch? It feels that
> > newbies would benefit from this inclusion and more extensive
> > documentation about what to do to get address completion.
> 
> I certainly agree with this. I've done a pretty craptacular job of
> maintaining the python version (just now changed the URL on the wiki,
> though, to be fair to myself, I didn't know it was on the wiki in the
> first place).

So have I with the vala version. I did it to scratch my itch and it has
been working for me ever since. However, I am not very interested in
maintaining or improving this piece of code. So, if the vala version
goes into the repository, I would appreciate if I would not be the
maintainer.
 
> I'd personally argue, not surprisingly, for unifying on the python
> version, if unifying we must, just because the python bindings are much
> more used than vala bindings (and way more people hack on python than
> vala) so it stands a better chance of keeping up with the binary. [...]
> Sebastian might have other opinions, though.

You know that I like python ;). And the python version is certainly more
hackable for me too. The vala version is cool because it is translatable
into pure C and it has some logic for minimizing the number of
lookups. (I think that is now also in the python version). Once the
from/to/cc addresses are stored in xapian itself, we are much less IO
bound than now, I'd think.

Sebastian

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: address completion when composing

[Florian Friesdorf]

[-- Attachment #1: Type: text/plain, Size: 1748 bytes --]

On Wed, 11 May 2011 10:18:43 -0700, Jameson Graef Rollins <jrollins@finestructure.net> wrote:
> On Wed, 11 May 2011 14:31:36 +0200, Florian Friesdorf <flo@chaoflow.net> wrote:
> > In addition, about a year ago, Sebastian indicated, that notmuch itself
> > could offer returning email addresses [3]. Are there arguments against
> > or just nobody found the time to do it?
> 
> As with pretty much everything in notmuch, manpower time is probably the
> issue.  I think that if you wanted to get this working that would be
> great.

ok, if there is anybody eager to do this - please go on! otherwise, I
would give it a try, but considering my current workload this might take
a bit. But, I promise I will not start using any other address book
solution to increase the pressure.

It would be great if people more knowledgeable about the notmuch code
base could provide some pointers and outline the solution.

What I heard here so far:
1. store to/cc/bcc in xapian
2. provide a mechanism to query those fields from the cmdline

Questions:

- Do we want a concept of
  one-person-several-emails(-for-different-purposes)?

- Do we want to filter only on the email/name data itself or also things
  like:
  
  * give me all email addresses of people who sent a mail between 1st
    and 5th May

  * ... who wrote mails mentioning xyz

  * ... who wrote mails tagged X

  * ... who wrote to foo@bar.com

- Do we want to tag persons/email addresses to create groups. 


A list of "you should read this"-links would be great. 

-- 
Florian Friesdorf <flo@chaoflow.net>
  GPG FPR: 7A13 5EEE 1421 9FC2 108D  BAAF 38F8 99A3 0C45 F083
Jabber/XMPP: flo@chaoflow.net
IRC: chaoflow on freenode,ircnet,blafasel,OFTC

[-- Attachment #2: Type: application/pgp-signature, Size: 835 bytes --]

Re: address completion when composing

[Jameson Graef Rollins]

[-- Attachment #1: Type: text/plain, Size: 529 bytes --]

On Thu, 12 May 2011 02:04:41 +0200, Florian Friesdorf <flo@chaoflow.net> wrote:
> 1. store to/cc/bcc in xapian

Hey, Florian.  I think all of these fields are actually already stored
in the database.

> 2. provide a mechanism to query those fields from the cmdline

I think this is the idea.  You would also want to match against subsets
of the fields, and output all matching fields.

I think just starting with this, without worrying about any other
options, would be a great way to start, and useful in and of itself.

jamie.

[-- Attachment #2: Type: application/pgp-signature, Size: 835 bytes --]

Re: address completion when composing

[Sebastian Spaeth]

[-- Attachment #1: Type: text/plain, Size: 396 bytes --]

On Thu, 12 May 2011 00:38:20 -0700, Jameson Graef Rollins wrote:
> On Thu, 12 May 2011 02:04:41 +0200, Florian Friesdorf <flo@chaoflow.net> wrote:
> > 1. store to/cc/bcc in xapian
> 
> Hey, Florian.  I think all of these fields are actually already stored
> in the database.

No, just the terms are, not the full string. (although for search, that
could actually be enough)

Sebastian

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]

Re: address completion when composing

[Carl Worth]

[-- Attachment #1: Type: text/plain, Size: 411 bytes --]

On Mon, 16 May 2011 22:04:42 +0200, Sebastian Spaeth <Sebastian@SSpaeth.de> wrote:
> No, just the terms are, not the full string. (although for search, that
> could actually be enough)

Right. It's easy enough to search on the "To:" field. But for address
completion, I think we realyl will want to fix the database to store
complete names+addresses that can be recovered for messages matching a
search.

-Carl

[-- Attachment #2: Type: application/pgp-signature, Size: 197 bytes --]