On Thu, Mar 2, 2017 at 4:55 AM, Steven Allen <steven@stebalien.com> wrote:
>
>
> David Bremner <david@tethera.net> writes:
> > This would require some modifications of notmuch. Either modifying
> > lib/index.cc to add the terms at indexing (notmuch new/insert) time, or
> > providing some way of adding the terms later. The former actually sounds
> > simpler to me.
>
> To do this correctly, you'd want to be able to run an external text
> extraction tool (for PDFs, word documents, etc.) so I think the latter
> would be better in the long run (it would allow the user to index
> attachments in the hooks).

(As a non-dev...) I agree.  The ability to add (and delete!) content post-insert sounds more desirable.  I don't want to have to re-index all my email as the next version of <horrible-binary-object>-to-text gets released.  I'd like to be able to (search-for-attachment)-(delete)-(re-add).


I was thinking a really hacky solution would be fake up a new email with the same headers but body being the attachment text, doing a notmuch new/insert and then replacing the file on disk of the new email with a link to the original message (not sure if that will trigger notmuch new, I don't think so).  Doesn't feel robust, but...


What do ya reckon?