how do you deal with string interpolations?

[BrianHicks]

As I mentioned in #481, I'm making a toy formatter for Ruby just to learn how this all works. I have a question about how you'd preserve spaces in interpolated strings.

The Ruby grammar has strings with string_content and interpolation nodes that can live inside them. The string_content frequently has whitespace at either end. How do I preserve that? For example, here's an input string:

"1 plus 1 is #{2}, 1+1=#{2}"

If I have these rules:

;; Strings
; Leave spaces alone, except for removing spaces around interpolation delimiters
(string_content) @do_nothing
(interpolation . "#{" @append_antispace)
(interpolation "}" @prepend_antispace .)

Then it's formatted like so, when I would want it to be left exactly as it was:

"1 plus 1 is#{2}, 1+1=#{2}"

(Doing (string_content) @leaf does the same thing, by the way.)

Is there a way to turn off formatting within some node, but only for its immediate children? Or to declare that some child gets a pass and that space within the text content of that node should be included? (Examining the nodes with tree-grepper shows that this is the case; I've attached it below.) Maybe something with a name like @append_preserve_whitespace/@prepend_preserve_whitespace?

$ tree-grepper -q ruby '(string_content)' -f pretty-json ./topiary/tests/samples/input/ruby.rb
[
  {
    "file": "./topiary/tests/samples/input/ruby.rb",
    "file_type": "ruby",
    "matches": [
      {
        "kind": "string_content",
        "name": "query",
        "text": "1 plus 1 is ",
        "start": {
          "row": 7,
          "column": 2
        },
        "end": {
          "row": 7,
          "column": 14
        }
      },
      {
        "kind": "string_content",
        "name": "query",
        "text": ", 1+1=",
        "start": {
          "row": 7,
          "column": 20
        },
        "end": {
          "row": 7,
          "column": 26
        }
      }
    ]
  }
]

[ErinvanderVeen]

Hej, thanks for this issue!

The team is currently on holiday, but I will discuss with them when they return tomorrow!

[nbacquey]

Hi @BrianHicks, thanks for the issue indeed!
I'm haven't had a thorough look at your problem yet, but I'm not sure the problem lies with Topiary (it could be with the Ruby tree-sitter grammar instead).
Could you please share the output of:

echo '"1 plus 1 is #{2}, 1+1=#{2}"' | RUST_LOG=debug topiary --language rust -sf /dev/stdin

Also, as a side note about your query file, the rule (string_content) @do_nothing serves no purpose.
The do_nothing directive is used to "cancel" a matching pattern, usually if some condition is met. For instance:

(
  (string) @append_hardline
  .
  (comment)? @do_nothing
)

This query will match on any (string) node (the (comment) is optional), to which it will append a line break, except if the (string) node is followed by a (comment). In this case, the @do_nothing directive will trigger, and cancel all other directives of the pattern.

[BrianHicks]

I get an error when I run that: "error: invalid value 'rust' for '--language '". Same if I try with --language ruby, which I think you might have meant? I was running it with cargo run instead of topiary directly, though, although I don't think that should make a difference?

As I've shown above with another tree-sitter tool, the whitespace is included in the string_content node in the Ruby grammar. Is that the kind of issue you're thinking about?

[nbacquey]

@BrianHicks I did mean --language ruby indeed, thanks for noticing. The whitespace is indeed included in the string_content node up to the very end of the formatting process.

I managed to track down the source of your issue to #492, we'll work on that.

[BrianHicks]

How do you get that --language thing to work, though? I see it in the README but it always complains that I'm using an invalid value?

Thanks for finding #492!

[nbacquey]

How do you get that --language thing to work, though? I see it in the README but it always complains that I'm using an invalid value?

The --language flag is only active for "supported" languages (read "non-experimental"). To enable it, you have to edit the associated enum and function in supported.rs.