Implement a debugger

[bbatsov]

A debugger is the biggest missing feature at this point and the biggest roadblock we have to conquer on the the road to CIDER 1.0.

I'm not sure how to best approach this task. I see two principle options:

• Port the Ritz debugger to CIDER. That should be fairly straightforward (at least in theory). The debugger code is here(Clojure) and here(Emacs Lisp).
• Implement a new solution from scratch. I have no idea if we can approach this problem from a different angle than @hugoduncan did for ritz. Feel free to suggest ideas.

I guess option one is the way to go, but we should still discuss alternatives before we start working on this. I'll probably take a break from working on features after 0.7, since I'd rewrite most of the nREPL client code (which is not particularly middleware-friendly right now), so I'm inviting the other CIDER hackers and users to step up and help out with this feature.

[turbopape]

Ritz seems to be too tied to java, or I'm not getting it ( using jdi and
offering features about spawned java processes etc. ..)? Wouldn't it be
more appropriate to rethink about the problem in terms of nrepl breakpoints
with commands for inspecting locals, and once we have that, adress the
problems of Java threads etc ? I mean addressing the debugger problem
incrementally ?
Le 3 août 2014 12:57, "Bozhidar Batsov" notifications@github.com a écrit :

A debugger is the biggest missing feature at this point and the biggest
roadblock we have to conquer on the the road to CIDER 1.0.

I'm not sure how to best approach this task. I see two principle options:

Port the Ritz debugger to CIDER. That should be fairly straightforward
(at least in theory). The debugger code is here
https://github.com/pallet/ritz/tree/develop/debugger(Clojure) and
here
https://github.com/pallet/ritz/blob/develop/nrepl/elisp/nrepl-ritz.el(Emacs
Lisp).

Implement a new solution from scratch. I have no idea if we can
approach this problem from a different angle than @hugoduncan
https://github.com/hugoduncan did for ritz. Feel free to suggest
ideas.

I guess option one is the way to go, but we should still discuss
alternatives before we start working on this. I'll probably take a break
from working on features after 0.7, since I'd rewrite most of the nREPL
client code (which is not particularly middleware-friendly right now), so
I'm inviting the other CIDER hackers and users to step up and help out with
this feature.

—
Reply to this email directly or view it on GitHub
#694.

[bbatsov]

Perhaps. Obviously anything will be a huge improvement over the current situation. I think some vim plugin relies on redl. We can probably do something similar for starters.

[turbopape]

Absolutely ! Maybe starting simple and focusing on clojure basic debugging
à la redl is the right thing to do ?
Le 3 août 2014 13:47, "Bozhidar Batsov" notifications@github.com a écrit :

Perhaps. Obviously anything will be a huge improvement over the current
situation. I think some vim plugin relies on redl
https://github.com/dgrnbrg/redl. We can probably do something similar
for starters.

—
Reply to this email directly or view it on GitHub
#694 (comment).

[bbatsov]

That's certainly fine by me.

[turbopape]

I'll try giving it a look then...
Le 3 août 2014 14:00, "Bozhidar Batsov" notifications@github.com a écrit :

That's certainly fine by me.

—
Reply to this email directly or view it on GitHub
#694 (comment).

[hugoduncan]

notifications@github.com writes:

Ritz seems to be too tied to java, or I'm not getting it ( using jdi and
offering features about spawned java processes etc. ..)?

The ritz implementation is very much Java specific, but the client
interaction was all via nrepl middleware ops.

[turbopape]

Yeah, I've seen it. I'll try to implement George jahad ' s inspired break
macros part in ritz for simple starters, that'll let me learn the basics
before I can delve into the java stuff...
Le 3 août 2014 18:27, "Hugo Duncan" notifications@github.com a écrit :

notifications@github.com writes:

Ritz seems to be too tied to java, or I'm not getting it ( using jdi and
offering features about spawned java processes etc. ..)?

The ritz implementation is very much Java specific, but the client
interaction was all via nrepl middleware ops.

—
Reply to this email directly or view it on GitHub
#694 (comment).

[turbopape]

I have one noob question though: how do I tell leiningen to use one local
copy of the cider - nrepl Middleware ? Excuse my laziness, but I wanted
you to point le in the right direction to begin developing asap ... thanks!
Le 3 août 2014 19:17, "Rafik NACCACHE" rafik.naccache@gmail.com a écrit :

Yeah, I've seen it. I'll try to implement George jahad ' s inspired break
macros part in ritz for simple starters, that'll let me learn the basics
before I can delve into the java stuff...
Le 3 août 2014 18:27, "Hugo Duncan" notifications@github.com a écrit :

notifications@github.com writes:

Ritz seems to be too tied to java, or I'm not getting it ( using jdi and
offering features about spawned java processes etc. ..)?

The ritz implementation is very much Java specific, but the client
interaction was all via nrepl middleware ops.

—
Reply to this email directly or view it on GitHub
#694 (comment)
.

[bbatsov]

Just run lein repl (or cider-jack-in) in the folder of cider-nrepl. This way you'll get to use the local copy of the middleware instead of the one you have installed via clojars.

[turbopape]

Ok thanks for the advice :)
Le 3 août 2014 19:37, "Bozhidar Batsov" notifications@github.com a écrit :

Just run lein repl (or cider-jack-in) in the folder of cider-nrepl. This
way you'll get to use the local copy of the middleware instead of the one
you'll have installed via clojars.

—
Reply to this email directly or view it on GitHub
#694 (comment).

[turbopape]

Hi @bbatsov ,

I investigated cider-nrepl middlewares structure and could feel more confident in designing a new basic one. Thank to your advice, I am movnig in the right direction in sketching in a very simplistic debugger (breakpoint + locals exploration). I more or less got the clojure part (cider-nrepl) of the problem.

To be able to test the whole process, though, I wanted -please - your help about the elisp part of the problem (cider) : from where do I send the ops ? cider-interaction.el ? cider-client.el ? Can you please point me to the proper file containing a sample call to an eval, factoring an nrpel message and processing the response ? Thank you very much. I promise this will be the last question, as I feel beginning getting it.

[turbopape]

I got it: it is by calling cider-send-op / cider-send-request-op. Now I am good to go !

[bbatsov]

Indeed. I'd suggest placing the debugger related code in a dedicated source file (say cider-debugger.el). You can use files like cider-macroexpand.el for reference.

[turbopape]

OK, got it. Actually it is by calling nrepl-send-request no ?