sp-forward-slurp-sexp should know about Clojure set literal, lambda

[lvh]

Example, | being point

(abc|) #{1 2 3}

sp-forward-slurp-sexp results in:

(abc| #){1 2 3}

Doing it a second time results in:

(abc| # {1 2 3})

(having introduced a space between # and {).

Instead, I would have liked it to:

(abc| #{1 2 3})

The same thing is true for Clojure lambda expressions, e.g. #(inc %); where the # is captured separately.

I think smartparens currently gives me the configurability to fix this for my setup, but I wanted to report it upstream since this seems generally useful.

[Fuco1]

It should be enough to do what we did for racket, that is adding the # as a prefix symbol.

Thanks for reporting, it's a nice feature to add.

[lvh]

This also affects conditional read statements for cljc files, which have #?(...) and #?@(...) which should be interpreted as one token.

[Fuco1]

Where could I find all the possible prefixes clojure supports?

[lvh]

That's a good question! Here's an attempt at an exhaustive list:

Reader macros (http://www.clojure.org/reference/reader):

• #{...} (sets)
• ^{:a 1}` (metadata maps , appear to be handled correctly already)
• #"..." (regex pattern)
• #'xyzzy (varquote)
• #(xyzzy 1 2 %) (anonymous fn)
• #_ (comment reader macro). This comments the next form. This appears to be an edge case, since it correctly captures the #_ entirely, but it doesn't capture the thing it comments out... Perhaps that's OK; the user can just slurp twice :) Same thing with tagged literals.
• Reader conditionals, e.g.:

#?(:clj     Double/NaN
   :cljs    js/NaN
   :default nil)

and

#?@(:clj [3 4] :cljs [5 6])

[Fuco1]

The issue is that some of those characters are correctly set as prefix class by the clojure mode but others are not. Since characters can only have one class, in SP we do the ugly thing and just recognize prefixes by a regexp... that's why I asked :) I would also need all the valid combinations (for example in elisp you can have a prefix ',@ and theoretically it can be infinitely nested (like '''''',@''').

So the best thing, and I should've asked right away, would probably be if you could write a regexp to recognize the prefixes, at least the most common variants.

[vspinu]

if you could write a regexp to recognize the prefixes

Where can I hook this regexp for generic use?

This is what I have so far:

(let ((clj-prefix "\\([@`'#~,_?^]* *\\)"))
  (sp-local-pair 'clojure-mode "(" ")" :prefix clj-prefix)
  (sp-local-pair 'clojure-mode "[" "]" :prefix clj-prefix)
  (sp-local-pair 'clojure-mode "{" "}" :prefix clj-prefix))

This works correctly on the following patterns:

#{...}
^{:a 1}
#(xyzzy 1 2 %)
#_ (comment reader macro)
#?(:clj     Double/NaN
   :cljs    js/NaN
   :default nil)
~unquote-me
`~'unquote-me
~@(splice-me)
`(quote-me)
```quote-me
@(deref-me)
@deref-me
#?@(:clj [3 4] :cljs [5 6])

but fails in the following cases

#"..."
#'xyzzy

[Fuco1]

sp-sexp-prefix applies to the mode, then the :prefix setting on each pair can override it specifically for that pair.

[vspinu]

Aha. Thanks.

The following works except for the same two troublesome cases from above.

(add-to-list 'sp-sexp-prefix '(clojure-mode regexp "\\([@`'#~,_?^]* *\\)"))