Re: [Tracker] Initial email indexing support

[Laurent Aguerreche <laurent aguerreche free fr>]

Le mercredi 20 septembre 2006 Ã 14:11 +0100, Jamie McCracken a Ãcrit :
> Laurent Aguerreche wrote:
> > Hello,
> >
> > this is a patch for an initial email indexing support with Evolution. It
> > extracts content of sent and received emails: text + attachments.
>
> superb - my hero :)
>
> > Currently, tracker_db_save_email() tries to save text and header infos
> > but it is missing some DB stuff... (Jamie...)
>
> yes I will get on to this.
>
> I have almost finished finished sqlite support so I expect both this and 
> the email/db support to be ready this weekend
>
> > For email bodies in HTML, there is nothing done and I don't know exactly
> > how it will be done.  :-)
>
> Cant you get Gmime to convert it (iirc they have an html filter in there 
> - google or look at the docs for it)
>
> if not we already have an html->text converter using the htmless app in 
> filters/text/html which we could call as a last resort (it wont be nice 
> though cause we would need to store the body as a tmp file for htmless 
> to convert it to a text file)

I will look.

> > Attachments have their metadatas/content extracted but that's all, and
> > they have their uri in something like /tmp/pid/attachment...
> > Currently, parent directories of treated files won't be deleted! So,
> > check your /tmp directory regularly.
>
> thats okay so long as the actual attachement files are unlinked.
>
>
>
> > Now, I'm waiting for DB things in:
> > * tracker_db_get_last_mbox_offset
> > * tracker_db_update_mbox_offset
> > * tracker_db_save_email
> >
> > (Since tracker_db_get_last_mbox_offset() isn't implemented it always
> > returns 0 which it makes indexing to always start from the beginning of
> > the mbox to treat.)
> >
> >
> > I will take a look at KMail.
> > In thunderbird, there is a stupid file to parse to find profiles then
> > directories for mboxes...
> >
> >
> >
> > Note: this patch isn't intended for real use...
>
> thats okay I expect we should have an -evolution parameter to trackerd 
> to indicate whether to index evolution stuff (I will add this later)
>
> I also noticed no checks are done in tracker-mbox-evolution to determine 
> if an email is marked as junk or deleted  (X-Evolution entry has bits at 
> the end to determine status (flagged, replied, seen, deleted, junk)) - 
> you may have to experiment with deleteing/marking as junk to work out 
> the exact flags used.

Yes and I tested it. With Evolution 2.6.3 in Debian SID it seems
completely buggy... Some emails are marked as junk or already answered
just after I received them... While reading sources of
evolution-data-server I saw an error in camel source code and reported
it but I don't know how much it affects evolution.

However I will use it but we will probably have some surprises with!


> For thunderbird, I believe it uses MOZILLA-STATUS but with different 
> flags from EVO.

Ok.

> this stuff can wait I suppose - I will need a day or two to go through 
> your patch although it looks good at first glance.
>
> Thanks for your time on this.