yari.el provides an Emacs frontend to Ruby's `ri' documentation
tool. It offers lookup and completion.

This version will load all completion targets the first time it's
invoked. This can be a significant startup time, but it will not
have to look up anything after that point.

This library tries to by compatible with any version of `rdoc' gem.
Self-testing covers all versions from 1.0.1 to 2.5.8 (current).

The main function you should use as interface to ri is M-x yari. I
recommend to bind it on some key local when you are ruby-mode. Here
is the example:

(defun ri-bind-key ()
  (local-set-key [f1] 'yari))

 or

(defun ri-bind-key ()
  (local-set-key [f1] 'yari-anything))

(add-hook 'ruby-mode-hook 'ri-bind-key)

You can use C-u M-x yari to reload all completion targets.

I added 'buttons' to the original code.  I am abandoning ri-emacs for yari because
it uses the ri command directly so there is far less compatibility problems.

There is no nice way I can see to communicate to the users so I'll add
my thoughts in here.

# I'd like to figure out a nicer way for the completions to work.  I
# can't do yari and type in create_type.  Instead I have to type in
# ActiveRecord::ConnectionAdapters::SchemaStatements#create_table which
# is rather painful.  It woudl be nice if I could type in "foo" and have
# the list of options be the various classes (or modules) that contain a
# method named "foo".

The above "wish" was accomplished by turning on ido-mode.  I'm
comfortable having ido-mode on all the time.  However I can see where
some might not want it on all the time.  It would be trivial to
introduce a customizable yari option so that only yari uses ido mode
keeping ido mode off for the rest of the emacs interfaces.  If I get a
request for this, I'll implement it.

I experimented with rvm and rvm.el.  If you are in a Rails app
directory (top level) and do M-x rvm-activate-corresponding-ruby, it
will load the local .rvmrc file and know which ruby and gemset to
use.  This is one step towards my goal.

The next step will be to create some new rake tasks to create the ri
documentation for the Rails application -- similar to ``rake doc:app''
but cause it to create the ri version instead of the rdoc version.  I
need to decide where to put it.  It is unfortunate that the Rails guys
decided to put the applications rdoc files in doc/app.  Perhaps put
them in doc/app/rdoc and doc/app/ri.  Then the api would be in
doc/api/rdoc and doc/api/ri.

The next step will be to invoke ri (from yari) with an
--doc-dir=doc/app/ri so that it will pick up the application's
documentation.  At that point, yari will know all about the
application, the particular rails it is using, and the particular set
of gems it is using.

Finally, I can not get the C-u M-x yari to reload the completion
targets.  On my emacs (GNU 23.1.1), I just get an error.  That is less
of an issue right now.