inconsistent use of backslash in regexps

* On 2007.05.07, in <038.29cad208b03a923a88ee8ab75959843e@mutt.org>,
*	"Mutt" <fleas@mutt.org> wrote: 
> #2886: inconsistent use of backslash in regexps
> 
>  If I want to match the + character from the index, I need to write, e.g.
>    ~Bc\\+\\+
>  (with two backslashes before the +). But from the pager, I need to search
>  for:
>    c\+\+
>  with a single backslash this time.

This isn't inconsistent, exactly, because the contexts aren't parallel.
When you enter a search expression at the index, the expression is
parsed by the command parser.  At the pager, it's simply a regexp -- no
parsing required.

Annoying, perhaps, but the only solution is to make the pager require a
parse as well, and that would make it inconsistent with external pagers
and general expectation.

Why is it parsed by the command parser in the index? Couldn't the same parser as in the pager be used?

The regexp isn't parsed individually, it's the entire pattern that's
parsed.  So you can do things like:

  ~s 'a subject with a space in it' ~B match.*expression

This should be made clear in the manual. It currently suggests that there is the same parsing in the pager and in the index:

3.2. The Pager
[...]
   /        search for a regular expression (pattern)
                                             ^^^^^^^
[...]
   \        toggle search pattern coloring
                          ^^^^^^^

I also wonder if, when invoking the search command in the pager, "Search" should be replaced by "Search (regexp)" or "Search for regexp" (as usually a search/limit/... command expects a pattern).

(In [4e37699a7750]) Document difference between index and pager search. Closes #2886.