On Wed, 10 Feb 2016, David Edmondson <dme@dme.org> wrote: > On Wed, Feb 10 2016, Mark Walters wrote: >> This basically looks fine to me and all tests pass. The code movement >> and cleanup all looks fine. > > Thanks. > >> Two minor things, one tiny nit below; and I wonder whether just having >> the buffer say "No search results" (or something similar) and leave >> the user to kill it would be nicer than dinging (and more in line with >> the way search and tree behave). >> >> [In some sense I think this way is right and search and tree are wrong, >> but that is probably difficult to get round as search and tree run >> asynchronously.] > > What if we did "notmuch count $query" first in the search and tree case, > and did the "(ding) (message ...)" thing if the count returned 0? (Just > wondering about whether having everything behave that way would be > possible and acceptable.) > > The original impetus for this change was someone who hits an id: button > that is either a false match (i.e it wasn't ever intended to be a > notmuch reference) or for a message that they don't have. In both of > those cases popping up a buffer that says only "No match." would be > annoying. If we were considering the case where people are using "M-x > notmuch-show", it seems less clear on the right thing to do, but overall > I prefer this approach to the useless buffer that I have to kill/quit. > I am quite happy to leave it as you have it with the ding; currently we have an obscure error message and this is obviously better so (with the comment change below) +1 from me. Best wishes Mark >>> + ;; Cache the original tags for each message so that we can display >>> + ;; changes. >> >> ^^ I think "Store the original tags for each message" would be better, >> particularly as this is nothing to do with the tag cache as used by say >> notmuch-tag-clear-cache. > > Agreed - fixed.