#3144: Parent and child match pattern modifier
--------------------------+----------------------
Reporter: jlh | Owner: mutt-dev
Type: enhancement | Status: new
Priority: minor | Milestone:
Component: mutt | Version:
Resolution: | Keywords:
--------------------------+----------------------
Comment (by vinc17):
Replying to [comment:11 kevin8t8]:
> The behavior of "not" (!) is unclear and even contradictory in the
patch. Should it apply to the pattern or to the parent/child operator?
Should there be a difference between
>
> {{{
> !>~f foo.com
> !>(~f foo.com)
> !(>~f foo.com)
> >!~f foo.com
> >(!~f foo.com)
> }}}
IMHO, the first three should be equivalent (just as they are equivalent
without {{{>}}}): this should be all the messages for which {{{>~f
foo.com}}} is false. And indeed, the current behavior is incorrect. The
last two should be equivalent: all the messages that have at least a child
for which {{{~f foo.com}}} is false.
And currently, {{{>(!~P)}}} and {{{>(!~P !~P)}}} return different results!
> Other notes:
>
> * Does it make sense for a parent/child operator to be applied to ~()
operator? Right now the code is, but this seems a bad idea.
It may be strange to use this, but I don't see any reason to forbid this
unless this makes the code too complex. {{{>~(PATTERN)}}} would mean: all
the messages that have at least a child for which {{{~(PATTERN)}}} is
true. So, if I understand correctly, compared to {{{~(PATTERN)}}}, this
would eliminate leaf messages (messages with no children).
> * Vincent: an important correction to your version of the patch. Note
that the pattern_cache_t applies to a single header: it doesn't deal with
threads. So inside the mutt_pattern_exec(), the first two calls to
pattern_exec() should pass NULL for the cache because they are looking at
parent/children. Only the third call should actually pass the cache
through.
Thanks!
--
Ticket URL: <https://dev.mutt.org/trac/ticket/3144#comment:12>
Mutt <http://www.mutt.org/>
The Mutt mail user agent
