Rule 3 indentation should only apply to lists

[yuhan0]

Rule 3 is all about maintaining semantic indentation on function / macro calls, it shouldn't apply in other collections like maps and vectors which don't have such semantics.

Given the input:

[:div {}
      "hello"]

{:a 1
    :b 2}

#{1 2 3
    4 5 6}

Expected

[:div {}
 "hello"]

{:a 1
 :b 2}

#{1 2 3
  4 5 6}

Actual

(no change)

[yuhan0]

Somewhat related: Rule 3 seems to be triggered in multiple positions when prefixes are involved, I don't think this is intended behavior?

;; All of the following are left unchanged
(foo ~'@(bar)
     baz)
(foo ~'@(bar)
      baz)
(foo ~'@(bar)
       baz)
(foo ~'@(bar)
        baz)

(I would consider only the first one to be correct)

Less obvious is how to treat metadata and discarded forms – should it trigger in any of the following situations? Note that all but the last 2 cases are currently accepted.

(defonce ^:private
         foo)
(defonce ^:private
          foo)
(foo #_bar
     baz)
(foo #_bar
       baz)

(foo #_bar baz
           qux)
(#_foo bar baz
           qux)

[oakmac]

Thank you for creating this issue. I am leaning toward supporting Rule 3 for any nested form (ie: maps, vectors, and sets).

Curious if @shaunlebron or @tonsky has any opinion on this?

[oakmac]

Somewhat related: Rule 3 seems to be triggered in multiple positions when prefixes are involved, I don't think this is intended behavior?

Some of these seem like a bug to me. I have added some test cases on PR-195 for what I think the behavior should be.

[shaunlebron]

My initial feeling is that Rule 3 shouldn’t apply to maps and sets, maybe because Rule 3 is contingent on there being a need for variable indentation, and I’m not aware of any such conventions for maps or sets.

I don’t know about vectors though. I can see wanting to have it for hiccups, especially in Reagent where [foo ...] acts exactly like a function call for building components.

[yuhan0]

But the issue is that there's no precedence or tooling support (as far as I'm aware) for applying this indentation style outside of lists, making a deliberate choice to introduce a new idiom feels arbitrary and counter to the project's goals.
Note that even under Fixed formatting rules, lists are already treated differently from non-list collections (2-spaces after delimiter va 1-space) so it makes sense to single them out.

[shaunlebron]

yeah I agree now, if not just for the simple reason that it would be way more offensive to see a let vector aligned with rule 3, than to see hiccups 1-space indented. And all the hiccup examples I just looked up on github use 1-space. I vote rule 3 should ignore vectors

[oakmac]

There are some good points being made in this thread. This is a breaking change to how Standard Clojure Style currently works, so I want to be careful about introducing it. Beyond some bug fixes, the main formatting algorithm has been stable for several months now.

Does anyone have any counter examples here? Some code that might not look well without Rule 3 support inside of a Vector, Map, Set?

Off the top of my head, I thought of something like this where you want to show structure inside of a vector:

;; before formatting
(def my-vec
  [:aaa :bbb :ccc
        :ddd :eee
             :fff])

But with the current implementation, this is already changed to:

;; after formatting using v0.24.0
(def my-vec
  [:aaa :bbb :ccc
        :ddd :eee
        :fff])

This is what it would look like if we do not support Rule 3 for vectors:

;; this is what we are proposing in this thread
(def my-vec
  [:aaa :bbb :ccc
   :ddd :eee
   :fff])

And even this example feels like a toy to me; not "code that I might see for real".

Ideally I am looking for a counter example that is similar to this one from @yuhan0 on Issue #60.

[yuhan0]

Commas are handy for this sort of thing, although I've never seen / used it for leading indentation before:

(def my-vec
  [:aaa :bbb :ccc
   ,,,, :ddd :eee
   ,,,, ,,,, :fff])


;; Slightly more realistic: visually indicating the semantics of a clunky 6-parameter constructor
(symmetric-3x3-matrix
  +2 -1  0
  ,, +2 -1
  ,, ,, -2)

Here's a counter-counter-example relating to @shaunlebron's comment about 'offensive' let bindings:

(let [a :aaaa b :bbb
      c :cc   d :dd]
  (do-something a b d))

Say you want to remove the unused binding by placing the cursor on line 2 and deleting 5 characters c :cc, placing the document in the following undesired state:

(let [a :aaaa b :bbb
        d :dd]
  (do-something a b d))

[shaunlebron]

Does anyone have any counter examples here? Some code that might not look well without Rule 3 support inside of a Vector, Map, Set?

I can’t think of any. I once formatted tetris piece coordinates spatially (below), but Rule 3 wouldn’t preserve them. Commas could be inserted to fix it, as mentioned.

(def tetris-pieces
  {:S [
                 [ 0 -1] [ 1 -1]
         [-1  0] [ 0  0]                 ]
   :Z [  
         [-1 -1] [ 0 -1]
                 [ 0  0] [ 1  0]         ]
   :O [
                 [ 0 -1] [ 1 -1]
                 [ 0  0] [ 1  0]         ]
   :T [
                 [ 0 -1]
         [-1  0] [ 0  0] [ 1  0]         ]
   :I [
         [-1  0] [ 0  0] [ 1  0] [ 2  0] ]
   :L [
                         [ 1 -1]
         [-1  0] [ 0  0] [ 1  0]         ]
   :J [
         [-1 -1]
         [-1  0] [ 0  0] [ 1  0]         ]})

[oakmac]

I once formatted tetris piece coordinates spatially (below), but Rule 3 wouldn’t preserve them. Commas could be inserted to fix it, as mentioned.

This is also a good use case for #_:standard-clj/ignore:

#_:standard-clj/ignore
(def tetris-pieces
  ...
  )