Remove incorrect and bugged inspect_request

[LunarLanding]

With the previous code, shift+tab on the notebook runs code silently instead of showing the documentation. It's not the intended function of inspect_request and might create confusion with users.
(according to https://jupyter-client.readthedocs.io/en/latest/messaging.html#introspection )

Also, @backend.eval is called with the wrong number of arguments (running on ruby 2.3.0 )

[LunarLanding]

Hm, this is incomplete, since according to https://jupyter-client.readthedocs.io/en/latest/messaging.html#compatibility , the kernel should ignore messages for which there is no handler implemented. So more changes are necessary.

[mrkn]

@LunarLanding Can you continue this pull-request?

[v0dro]

@LunarLanding ping!

[LunarLanding]

I am continuing it now that I have some free time, sorry for the delay in replying.

[LunarLanding]

I have been reading pry's code, specifically the autocomplete section. If we could easily extract the scope of the line being inspected, we could then use almost the same code to obtain the object being inspected, and then it should be relatively easy to display information about it.
In order to extract the scope of potentially incomplete code, I looked into parsers. I started with ripper, as the source code suggested. However incomplete code such as:

class A
    some_method_to_inspect

Makes the parser error and not produce an AST. I looked into https://github.com/whitequark/parser and it seems to have the same problem, although it produces an error message with enough information to know which blocks / parentheses are open. If we dig into the internals we could probably "close" those contexts, get a valid AST, and then get the scope. I'm assuming people use iruby as they did the ipython notebook, i.e. running their definitions first in order to be able to inspect it / autocomplete it in a lower cell, and therefore I am trying to at least get the scope of the inspect request.

[v0dro]

@mrkn WDYT?

[mrkn]

It is great if we can support inspect_request message completely, but it isn't easy as @LunarLanding was investigated.

So I think that deactivating inspect_request message, which is the first aim of this pull-request, is good as the first step to fix it. And then we should make another new issue to realize the ideal inspect_request message handler.

To deactivate it, the simple solution may be making the empty inspect_reply response as below.

diff --git a/lib/iruby/kernel.rb b/lib/iruby/kernel.rb
index b9e9d6e..f7af276 100644
--- a/lib/iruby/kernel.rb
+++ b/lib/iruby/kernel.rb
@@ -142,14 +142,7 @@ module IRuby
     end

     def inspect_request(msg)
-      result = @backend.eval(msg[:content]['code'])
-      @session.send(:reply, :inspect_reply,
-                    status: :ok,
-                    data: Display.display(result),
-                    metadata: {})
-    rescue Exception => e
-      IRuby.logger.warn "Inspection error: #{e.message}\n#{e.backtrace.join("\n")}"
-      @session.send(:reply, :inspect_reply, status: :error)
+      @session.send(:reply, :inspect_reply, status: :ok, found: false, data: {}, metadata: {})
     end

     def comm_open(msg)

@LunarLanding How about this direction?

[v0dro]

@mrkn would you like to merge this now?

[kojix2]

I want to merge this pull request.
If the original inspect_request doesn't work, the blank code is better.