On Mon, Jun 6, 2016 at 1:29 PM, David Bremner <david@tethera.net> wrote:

Sebastian Fischmeister <sfischme@uwaterloo.ca> writes:

>
> I ran into this problem before as well. Storage is cheap. Notmuch could
> index all emails with reversed text to get around some of this
> problem. It doesn't solve the problem of *analysis*, but it's still an
> improvement.

It would probably be more useful to have brute force regexp searches on
headers. Austin did some experiments that sounded promising, where you
basically postprocess the result of a xapian query with a regexp. OTOH,
I don't know what kept him from proposing this for mainline. If it was
just parser issues, those are probably more or less solved now, at least
for people using xapian 1.3+

The experiment was specifically for regexp matching subject, but it should work for any header we store a literal copy of in the database. The code is here, though in its current form it builds on my custom query parser: https://github.com/aclements/notmuch/commit/ce41b29aba4d9b84e2f1eb6ed8df67065196c960. Based on my understanding of Xapian 1.3+ field processors, these days it should be quite easy to hook the PostingSource in that commit into the Xapian QueryProcessor.