Inconsistent indentation for macros

[rrudakov]

Expected behavior

1. Macros indentation is consistent with cljfmt
2. Indentation is should be consistent when M-x cider-format-buffer is called.

Actual behavior

Indentation is inconsistent (see screenshots)

Steps to reproduce the problem

1. Create a simple project:

• deps.edn content:

{:paths ["src"]
 :deps  {tortue/spy {:mvn/version "2.13.0"}}}

• src/com/example/app.clj content:

(ns com.example.app
  (:require
   [clojure.test :refer [deftest]]
   [spy.assert :as assert]
   [spy.core :as spy]
   [spy.test]))

(defn- foo
  [bar]
  (println bar))

(deftest dummy-test
  (with-redefs [foo (spy/spy foo)]
    (foo 123)
    (assert/called-once-with? foo
                              123)))

2. Call M-x cider-jack-in
3. Place cursor on the last line (before 123) and press TAB (indent-for-tab-command). 123 will be indented as "body" (with 2 spaces indentation relative to the previous line)
4. Execute M-x cider-format-buffer. 123 will be aligned vertically with foo (as function arguments)

If I place foo to the new line the indentation is even more weird:

I assume it's related to indentation inference? In stable CIDER 1.8 the indentation is correct:

Environment & Version information

CIDER version information

;; CIDER 1.8.0-snapshot (package: 20230929.455), nREPL 1.0.0
;; Clojure 1.11.1, Java 17.0.8.1

Lein / Clojure CLI version

Clojure CLI 1.11.1

Emacs version

GNU Emacs 29.1.50 (build 1, aarch64-apple-darwin23.0.0, NS appkit-2487.00 Version 14.0 (Build 23A344)) of 2023-09-28

Operating system

MacOS 14

JDK distribution

openjdk version "17.0.8.1" 2023-08-24
OpenJDK Runtime Environment Homebrew (build 17.0.8.1+0)
OpenJDK 64-Bit Server VM Homebrew (build 17.0.8.1+0, mixed mode, sharing)

[vemv]

Thanks for the detailed report!

Checking out https://github.com/alexanderjamesking/spy/blob/e27887b3a0a9c9e5078e773b92b84c7571ce59a5/src/clj/spy/assert.clj, I'd say that the issue is that (defmacro called-once-with? [f & args] is expected to be indented like a function, despite being implemented as a macro.

There was a similar case asked about here: https://clojurians.slack.com/archives/C0617A8PQ/p1695695401220199?thread_ts=1695694483.271689&cid=C0617A8PQ

The solution is to add :style/indent metadata. It actually always has been best practice to add it, instead of letting an arbitrary behavior set in.

So I'd recommend a PR against that repo.

You can iteratively try out indentation styles using (alter-meta! #'spy.assert/called-once-with? assoc :style/indent ,,,). Such modifications are good to leave in e.g. user.clj while the PR is open.

Cheers - V

[rrudakov]

Thank you for quick response @vemv!

Maybe it would be a good idea to introduce some defcustom to disable automatic indentation reference and prefer the old behaviour (at least compatibility with cljmft still seems broken in the current state)?

[dakra]

I had a similar problems a few times already (e.g. promesa threading macros).
I think a defcustom that you can set in Emacs to overwrite or add :style/indent for a specific macro would be a good idea.
Obviously adding it upstream to the lib is preferred but sometimes you can't or don't want to.

[vemv]

Without inference, a lot of other macros will look non-idiomatic.

CIDER has always fostered :style/indent. Throughout the years I've had to manually add it many times for things to look as envisioned in clojure-mode, the clojure style guide, etc.

I believe inference forces users to be more aware of https://docs.cider.mx/cider/indent_spec.html and to add it from time to time. That's a good thing.

CIDER's job is not only to be compatible with cljfmt "no matter how", but for your Clojure code to look actually more accurate/informative.

The best we could do, probably, is to add a whiltelist for macros that shouldn't be infered. But then again, if you are configuring that, you could as well be configuring :style/indent as a PR or locally under user.clj (or any file that will be loaded, really).

[vemv]

Another approach that could alleviate this as a "side-effect" is supporting .cljfmt.edn files (clojure-emacs/cider-nrepl#754)

I intend to get back at that PR after we release CIDER 1.8.0.

[bbatsov]

@vemv Perhaps short-term you can just expand the docs to feature something in "Troubleshooting" or the indentation section, so it's easier for to find the information they are looking for.

[dakra]

I guess this is a different issue but reading https://docs.cider.mx/cider/indent_spec.html#indentation-inference
I think promesa.core/-> should be treated after inference as clojure.core/-> (same name, same args).

I can create a new issue for it if you think this is a bug. Otherwise I'll make a PR to promesa adding :style/indent.

[vemv]

Thanks! I'll check it out today

[vemv]

I've fixed that (if you are curious: clojure-emacs/orchard@13c3ea7) and will let you know when cider-nrepl is bumped - should happen during this weekend.

[dakra]

🎉 Awesome, thanks.

[vemv]

cider-nrepl 0.39.0 has been released with those changes.

It's injected by CIDER latest (Git master, or the MELPA snapshot that will be visible within a couple hours)

[vemv]

I've released cider-nrepl 0.40.0 / cider 1.18.2 with a further fix.

It's not directly related to indentation inference, however it fixes certain indentation cases for clojurescript.

[DerGuteMoritz]

I've released cider-nrepl 0.40.0 / cider 1.18.2 with a further fix.

I am (still?) seeing unexpected indentation with Promesa's threading macros here (i.e. the first form is treated as a "head") using ClojureScript. You mention CIDER 1.18.2 which I assume is a typo - judging from the date you posted the comment, you probably meant 1.8.2. In any case, there appears to be a regression...

[vemv]

Please create a separate issue in detail

Thanks - V