rinari-console and comint-previous-input

[Duncan Mak]

I started using rinari-mode and I noticed in the *ruby* buffer from rinari-console, comint-previous-input fails (M-p) with this error:

inf-ruby-get-old-input: Search failed: "^>> "

I see that the regexp is explicitly written in the code - is there a reason for that? 

Is there a fix?

Thank you.

Duncan.

[Aleksandar Simic]

Hello Duncan,

how are you using ruby interpreter ? Via RVM or just "native" (without RVM)?

I ask this since I've only seen this issue when using Ruby via RVM and
it has been annoying me, but I just didn't have time to poke it
further...

Aleksandar

[Duncan Mak]

Yeah, I'm using RVM also - I found some rvm emacs package and that's what I'm using.

[Steve Purcell]

I've committed what I think is a fix for this issue:

https://github.com/eschulte/rinari/commit/235f14b4c10312297cfaed29d45456ba361b2ddf

Please let me know if it doesn't work for you.

-Steve

[Aleksandar Simic]

Thank you Steve, I'll have a look at it later today

[Aleksandar Simic]

Hello Steve,

thanks again for providing this fix. However it doesn't work in some cases.

I've posted a gist here: https://gist.github.com/1033069 outlining a
solution for me and a slight tweak to the regexp.

It might be a good idea to get the inf-ruby/irb prompts for all the
versions of ruby, in order to have a catch all solution.

The only reason I posted the above as a gist and not in the comments
of your diff is because the comment text box at github was escaping
some of the regexp's escape characters...

Aleksandar

[Steve Purcell]

Aleksandar Simic <asi...@gmail.com> writes:
> Hello Steve,
>
> thanks again for providing this fix. However it doesn't work in some cases.
>
> I've posted a gist here: https://gist.github.com/1033069 outlining a
> solution for me and a slight tweak to the regexp.
>
> It might be a good idea to get the inf-ruby/irb prompts for all the
> versions of ruby, in order to have a catch all solution.

Thanks! I've modified the regexp again so that it matches all of your
examples, but I've purposely kept it a little more conservative than
your proposed alternative. I think it should handle all the likely cases
now.

-Steve

[Duncan Mak]

I'm trying to to run the latest checkout of rinari, and I keep on running into this error:

if: Wrong type argument: stringp, nil

Duncan.

[Aleksandar Simic]

Hi Duncan,

What does your irb prompt look like (outside rinari, on a console)?

Can you run inf-ruby mode without issues?

[Duncan Mak]

I think I'm crashing inside rinari-root.

My prompt looks like this: jruby-1.6.1 :001 > 

Duncan.

[Aleksandar Simic]

Hello,

Can you do the following:

1. describe your whole session, from the start until the crash

2. Separately, run "M-x inf-ruby". Do you experience any issues?

Thanks

[Steve Purcell]

Apologies if I've unwittingly broken things for you.

Please try "M-x toggle-debug-on-error" before triggering the error, then
post the traceback you get.

-Steve

[Duncan Mak]

Sorry it took me so long to reply:

Here's the stack trace that I get:

Debugger entered--Lisp error: (wrong-type-argument stringp nil)

  string-match("\\(^[[:alpha:]]:/$\\|^/[^/]+:/?$\\|^/$\\)" nil)

  (if (string-match "\\(^[[:alpha:]]:/$\\|^/[^/]+:/?$\\|^/$\\)" dir) nil (rinari-root new-dir))

  (unless (string-match "\\(^[[:alpha:]]:/$\\|^/[^/]+:/?$\\|^/$\\)" dir) (rinari-root new-dir))

  (let ((new-dir (expand-file-name (file-name-as-directory "..") dir))) (unless (string-match "\\(^[[:alpha:]]:/$\\|^/[^/]+:/?$\\|^/$\\)" dir) (rinari-root new-dir)))

  (if (file-exists-p (expand-file-name "environment.rb" (expand-file-name "config" dir))) dir (let ((new-dir (expand-file-name (file-name-as-directory "..") dir))) (unless (string-match "\\(^[[:alpha:]]:/$\\|^/[^/]+:/?$\\|^/$\\)" dir) (rinari-root new-dir))))

  rinari-root()

  (expand-file-name "script" (rinari-root))

  (file-name-as-directory (expand-file-name "script" (rinari-root)))

  (concat (file-name-as-directory (expand-file-name "script" (rinari-root))))

  rinari-script-path()

  (let* ((default-directory (rinari-root)) (script (rinari-script-path)) (command (expand-file-name (if (file-exists-p (expand-file-name "console" script)) "console" "rails console") script))) (if rinari-rails-env (setq command (concat command " " rinari-rails-env))) (setq command (if edit-cmd-args (read-string "Run Ruby: " (concat command " ")) command)) (run-ruby command) (save-excursion (set-buffer "*ruby*") (set (make-local-variable (quote inf-ruby-prompt-pattern)) "^\\(j?ruby[^> ]+\\|J?RUBY[^> ]+\\|irb([^> ]+\\)?\\( ?:[0-9]+\\)* ?>>? ") (set (make-local-variable (quote inf-ruby-first-prompt-pattern)) inf-ruby-prompt-pattern) (rinari-launch)))

  rinari-console(nil)

  call-interactively(rinari-console t nil)

  execute-extended-command(nil)

  call-interactively(execute-extended-command nil nil)

[Steve Purcell]

Are you perhaps executing rinari-console while in a buffer that has no
associated file, e.g. *scratch*? It looks like rinari-root is seeing a nil
value for 'default-directory.

Do things work if you open a file inside your rails project, and execute
rinari-console while your cursor is in the buffer for that file?

-Steve

[Duncan Mak]

I was running it in the dired buffer of the directory that contains my project. I also tried running rinari-console when the buffer is open on a Rakefile in the project.

I'm running Emacs 24 now, but when I first encountered this problem, I was running Emacs23. 

From describe-variable:

default-directory is a variable defined in `C source code'.
Its value is "/home/duncan/git/Scraper-Node/"
Local in buffer Rakefile; global value is nil
  Automatically becomes buffer-local when set in any fashion.
  This variable is safe as a file local variable if its value
  satisfies the predicate `stringp'.
Documentation:
Name of default directory of current buffer.  Should end with slash.
To interactively change the default directory, use command `cd'.

Duncan. 

[Steve Purcell]

I've filed this as an issue here so I don't forget, but I'm a bit snowed
under, so may not be able to address it immediately:

https://github.com/eschulte/rinari/issues/8

-Steve

[Duncan Mak]

I'm really surprised that I'm the only look experiencing this problem.

I must have something configured wrong?

Duncan.

[Steve Purcell]

I've been coming to the same conclusion. If you could show how to
reproduce the error from a clean Emacs, I'd have somewhere to start.

-Steve

[Duncan Mak]

So what's the best way to run rinari-console?

Normally, I open up a dired buffer on the folder with "script/console" and then I run M-x rinari-console from there.

Is that not the right way to start the console?

Duncan.

[Steve Purcell]

Aha! No, "M-x rinari-console" will actually run "script/console" (or "rails console") for you.

-Steve

[Duncan Mak]

So in which buffer do you run rinari-console?

[Steve Purcell]

Duncan Mak <dunc...@gmail.com> writes:
> So in which buffer do you run rinari-console?

Any buffer in which 'rinari-launch has been called, I think. So normally you'd
have something like this:

(add-hook 'ruby-mode-hook 'rinari-launch)

and then in any ruby buffer inside a Rails source tree, rinari-console
should do the right things.

-Steve

[Duncan Mak]

Aha - the project that I'm working is not actually a rails project, but it does have a script/console file.

Once I made a dummy config/environment.rb file, rinari-console started to work.

New problem: 

LoadError: no such file to load -- bundler

       require at org/jruby/RubyKernel.java:1038

       require at /home/duncan/.rvm/rubies/jruby-1.6.3/lib/ruby/site_ruby/1.8/rubygems/custom_require.rb:29

Duncan.

[Duncan Mak]

Woohoo - I ran M-x use-rvm and I'm cooking with gas!

It'd be nice if that can be run automatically, but that's a problem for a different Google Group.

Thanks for all the help!

Duncan.

[Steve Purcell]

Happy you got things working!

-Steve