macrostep like macroexpand

[colonelpanic8]

Is there any reason that cider opted to use a separate buffer for outputting the result of expanding a macro? I quite like the way https://github.com/joddie/macrostep expands macros inline in emacs lisp. Would the maintainers consider an optional change that maybe even reuses some of that code?

[bbatsov]

Legacy. We just did it the way it was done in SLIME. Including inline expansion of macros shouldn't be hard and we discussed it in the past. If someone is willing to implement this we'd be happy to accept it.

[xiongtx]

Inline expansion isn't hard, but "stepping" is a bit trickier. The macroexpansion functions in CIDER do not currently do that.

It'd be necessary to ask the middleware whether the form we're looking at is a macro. See slime-macrostep, for example.

[bbatsov]

You can already get this info from the var-info middleware.

On Monday, 26 September 2016, Tianxiang Xiong [email protected]
wrote:

Inline expansion isn't hard, but "stepping" is a bit trickier. The
macroexpansion functions in CIDER do not currently do that.

It'd be necessary to ask the middleware whether the form we're looking at
is a macro. See slime-macrostep
https://github.com/slime/slime/blob/master/contrib/slime-macrostep.el#L117,
for example.

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub
#1850 (comment),
or mute the thread
https://github.com/notifications/unsubscribe-auth/AAGVykDYL5hFSybrIzYs2cOwlZHgeSLMks5qtuftgaJpZM4KBaVU
.

Best Regards,
Bozhidar Batsov

http://www.batsov.com

[xiongtx]

Here's a proof of concept:

First define some Clojure functions:

(defn macro? [form]
  (not (= (macroexpand-1 form) form)))

(defn expand-once [form]
  (let [expanded? (atom false)]
    (walk/prewalk (fn [x]
                    (if (and (not @expanded?)
                             (macro? x))
                      (do (reset! expanded? true)
                          (macroexpand-1 x))
                      x))
                  form)))

Then set the relevant values for macrostep (these can be improved!):

(setq-default macrostep-sexp-bounds-function
              (lambda ()
                (let ((pos (cider-sexp-at-point t)))
                  (cons (car pos) (cadr pos)))))

(setq-default macrostep-sexp-at-point-function
              (lambda (&rest _ignore)
                (cider-sexp-at-point)))

(setq-default macrostep-expand-1-function
              (lambda (form env)
                (nrepl-dbind-response (cider-nrepl-sync-request:eval (format "(expand-once '%s)" form))
                    (value)
                  value)))

(setq-default macrostep-print-function
              (lambda (sexp env)
                (insert sexp)))

Obviously we should be setting these in a hook, but this is a POC...

Then call macrostep-expand on something like:

(defn foo [x]
  (defn bar [y]
    (defn baz [z]
      (+ 1 2))))

[bbatsov]

@xiongtx I think you forgot to include something. :-)

[xiongtx]

Yes, accidentally hit RET too quickly.

Is there a way to pretty-print a Clojure form in an Emacs buffer? Not clojure.pprint, but like macrostep-pp or Emacs pp, i.e. aligning the text of the form in a nice way.

I don't see anything in the cider- namespace, and results from various CIDER commands don't do this (they just output the result in one long line), so my guess is "no".

[bbatsov]

CIDER's macroexpansion pretty-prints the macro forms (they are processed in the middleware).

I'm not sure it's a good idea to plug something into macrostep - I was thinking more in terms of just implementing the existing macroexpansion we have as on overlay. Seems to me this is pretty much what macrostep does. Integrations with 3rd party packages are notoriosly brittle and users generally don't want to install a ton of add-ons anyways.

[bbatsov]

P.S. I'm fine with developing a macrostep integration if macrostep brings a lot of added functionality to the table.

[xiongtx]

CIDER's macroexpansion pretty-prints the macro forms (they are processed in the middleware).

It does Clojure pprint, not Emacs pp. pp will take a one-liner like

(defn foo [x] (defn bar [y] (defn baz [z] (+ 1 2))))

and output:

(defn foo
  [x]
  (defn bar
    [y]
    (defn baz
      [z]
      (+ 1 2))))

which isn't quite right, but close. Clojure's pprint does not do this.

if macrostep brings a lot of added functionality to the table.

I'd think that the most desirable feature of macrostep is the "stepping" (which you can experiment with using the POC above), not just expanding inline. But I see that expanding inline is what the OP has asked for...does no one care about the stepping then?

[colonelpanic8]

@xiongtx
You are correct that I was mostly asking for inline expansion, but I think that stepping is also a valuable feature that would certainly be nice to have.

[bbatsov]

@xiongtx I'm guessing clj-fmt or https://github.com/kkinnear/zprint should be able to format the code in this manner

[xiongtx]

Would that need to be pulled into nREPL or something? How do we deal with Clojure dependencies or extra code in CIDER?

[bbatsov]

Would that need to be pulled into nREPL or something? How do we deal with Clojure dependencies or extra code in CIDER?

We simply add them as deps in cider-nrepl. Ideally they should be exposed via some nREPL op instead of using direct evaluation - this gives us the flexibility to replace some dep, but the preserve the nREPL interface. You can see the clj-format code in cider-nrepl as an example.

[xiongtx]

It seems that cljfmt is already a dependency. Is that exposed through Emacs Lisp as an op or something?

[bbatsov]

Sure. See

cider/cider-interaction.el

Line 1561 in c698df1

(defun cider--format-buffer (formatter)

[xiongtx]

Macro-stepping is now part of cider-nrepl.

Do we want cider-macrostep to be part of CIDER or a separate repo? slime-macrostep is a contrib package for SLIME, though SLIME packages all contrib packages within SLIME itself.

[bbatsov]

I guess that depends on whether cider-macrostep depends on macrostep or not.

[bbatsov]

@xiongtx Ping :-)

[bbatsov]

@xiongtx Ping 2 ;-)

[xiongtx]

Trying to work on it, but I'm not satisfied with the current way of hooking into macrostep:

• Formatting of code is not preserved
• macrostep-next-macro and macrostep-previous-macro are not working; need to mark macros w/ text properties

Needs more investigation...

[colonelpanic8]

@xiongtx status?

[colonelpanic8]

@xiongtx @bbatsov
Stepping would be nice, but not essential to a v1. Since that seems to be the hold up, why not just get inline expansion working for now?

[vspinu]

What is the added value of this? And does it really work? Macros could expand to something completely unrelated to the original code, so my gut feeling is that this functionality is bound to be broken outside of very basic macros.

Plus, there are benefits of having it in separate buffer. You see the whole expansion and can continue working on your macro at the same time. I personally find the "read-only" mode of elisp debugger quite limiting.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contribution and understanding!

[stale]

This issues been automatically closed due to lack of activity. Feel free to re-open it if you ever come back to it.

[nbfalcon]

I have just pushed cider support in macrostep-geiser.

Alternatively, I can offer an adapted version of that package that is cider-specific and could theoretically be merged into cider (at the cost of lots of code/maintanence duplication): gist. Of course such a functionality would (at least at runtime) depend on macrostep.

[bbatsov]

@nbfalcon That's for letting me know. The package is relatively small, so I'm not particularly worried about code duplication. That being said - I'm fine with this being a separate package, although I think with a name with macrostep-geiser the likelihood any CIDER user coming across it would be more or less 0. I can mention it in the docs, but the problem is that no one reads the docs. :D

It'd be nice if there was a package named cider-macrostep/macrostep-cider that people can organically discover.