From mboxrd@z Thu Jan  1 00:00:00 1970
From: magnus@therning.org (Magnus Therning)
Date: Wed, 19 Sep 2007 23:02:00 +0100
Subject: [sup-talk] Minor maildir change
In-Reply-To: <1190216731-sup-3428@south>
References: <20070909224034.GA3540@die.therning.org>
	<1189740088-sup-6296@south>
	<20070916181936.GA5332@die.therning.org>
	<1190216731-sup-3428@south>
Message-ID: <20070919220200.GA3436@die.therning.org>

On Wed, Sep 19, 2007 at 08:59:58 -0700, William Morgan wrote:
>Excerpts from Magnus Therning's message of Sun Sep 16 11:19:36 -0700 2007:
>> It's also a little strange that after a message is read it doesn't
>> necessarily end up with the correct set of labels.  Instead one has to
>> go to the index to get the proper set.
>
>What do you mean? For a new, unseen message, the initial set of label
>is uniquely determined by the source's default labels, unioned with
>whatever state (read/unread, flagged, etc) we carry over from the
>source. This is what Source#each does.

Yes, but once it's known Source#each still get labels off the state and
default flags from the source.  The same goes for
PollManager#add_messages_from, where the messages will have different
labels compared to the entry.  Here's code that hopefully highlight
what I mean:

  #!/usr/bin/env ruby
  
  require "uri"
  require "sup"
  
  Redwood::start
  index = Redwood::Index.new
  index.load
  
  index.sources.each { |src|
      if "maildir" == URI(src.uri).scheme
          src.reset!
  
          Redwood::PollManager.add_messages_from(src) { |m, offset, entry|
              puts "#{offset} -"
              puts "\t#{m.labels.join(' ')}"
              puts "\t#{entry[:label]}"
              nil
          }
      end
  }

The output on a maildir source I use for some testing is:

  [Wed Sep 19 22:55:02 +0100 2007] loading index...
  [Wed Sep 19 22:55:02 +0100 2007] loaded index of 3 messages
  11755314090094973 -
          sup-test unread deleted inbox
          sup-test
  11765622040003515 -
          sup-test unread inbox
          sup-test deleted
  11776152050016455 -
          sup-test inbox unread
          sup-test unread

What's gotten from the source is useful for an unseen message, but why
not mark known messages with the labels from the index?

>> Another strange thing was that when using
>> PollManager::add_messages_from all my emails ended up being unread
>> (according to the labels in the message).  I tracked that down to it
>> using the value of a header ("Status") to see whether a mail was read
>> or not.  Is that something that's used for mboxes?  (I've attached a
>> patch that removes the code.)
>
>That is a hack to avoid having the mbox code read the header twice, one
>for the labels and again for the actual header contents.  Since I'm
>planning on reworking mbox anyways (to avoid keeping the file pointer
>always open, and to clean up the interface between loader and
>ssh-loader), so I'll try and think of a better way to do this. There
>might not be.
>
>In the mean time, I'll remove the second half of that statement, which
>should make everyone happy.

Excellent.

[..]
>>  - Make all sets of labels actual sets.
>
>Not sure what you mean by "actual sets", but I think this is pretty
>much the way Sup works now.

What I mean is the type set gotten from "require 'set'" :)  Currently
labels are lists, not sets.

/M

-- 
Magnus Therning                             (OpenPGP: 0xAB4DFBA4)
magnus?therning?org             Jabber: magnus?therning?gmail?com
http://therning.org/magnus
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://rubyforge.org/pipermail/sup-talk/attachments/20070919/8103ec01/attachment.bin