On Thu, 2014-05-29 at 12:04 +0100, Giles Coochey wrote: > On 29/05/2014 11:43, Axb wrote: > > On 05/29/2014 12:22 PM, Arthur Dent wrote: > > > >> So... > >> > >> Will this work for sa-learn? > >> > >> ============================8<============================================================ > >> > >> > >> # Proposed sa-learn maildir script > >> #!/bin/bash > >> > >> sa-learn --ham ~/Maildir/.Hobby/{cur,new} > >> sa-learn --ham ~/Maildir/.{Misc,Personal,etc}.*/{cur,new} > >> sa-learn --spam ~/Maildir/.Malware.*/{cur,new} > >> ============================8<============================================================ > >> > >> > > > > "new" means unread - you really want to run sa-learn on stuff you > > haven't looked at? (as in "learning false negatives as ham?") > > > If it was his Inbox then perhaps it would be best to avoid "new", but a > lot of the time stuff that is "unread" in other folders generally means > that it has been looked at - perhaps not totally read, or perhaps put > aside for later inspection. > For me, I use "unread / read" as a marker to whether I have actioned a > particular email and keep messages "unread" until such time that they > are dealt with.
Yes, quite right. All the stuff in each of those mbox files has been either put there manually, or by a well-tested procmail recipe from known contacts etc. I do sometimes file an unread message for later reading, so I think I will need "new". The (very few) FNs that slip through get hand-filed be me into the Malware/Spam mbox and get re-learned on the next run of the script (every night). So - is the syntax correct for a maildir format? In particular will it work with the current structure (i.e. will the line "sa-learn --ham ~/Maildir/.{Misc,Personal,etc}.*/{cur,new}" correctly catch: .Misc.Clubs.cur .Misc.Clubs.new .Misc.Car.cur .Misc.Car.new etc...? Should I use --no-sync? Many thanks for the help so far... Mark