OSX - Colon replacement incorrectly reads closure curly braces as a struct

[jclausen]

With the following settings:

	"struct.separator": " : ",
	"struct.padding": true,
	"struct.empty_padding": false,
	"struct.multiline.leading_comma": false,
	"struct.multiline.leading_comma.padding": true,
	"struct.multiline.element_count": 2,
	"struct.multiline.min_length": 60,

The following block of code:

hyper
	.fake( {
		"https://mydomain.com/auth" : (
			newFakeResponse,
			req
		) => {
			var body = req.getBody();
			if ( body.User != "valid_user" || body.Secret != "valid_secret" ) {
				return newFakeResponse( 401, "Unauthorized" );
			}

			return newFakeResponse(
				200,
				"Created",
				serializeJSON( {
					"token"      : "valid_token",
					"expiration" : dateAdd( "d", 1, now() )
				} )
			);
		},
		...
	} );

ends up being formatted ( note the change of the var body = req.getBody(); to var body : req.getBody(); ):

hyper
	.fake(
		{
			"https://mydomain.com/auth" : (
				newFakeResponse,
				req
			) => {
			var body : req.getBody();
			if ( body.User ! : "valid_user" || body.Secret != "valid_secret" ) {
				return newFakeResponse( 401,
			"Unauthorized" );
		}

		returnnewFakeResponse(
			200,
			"Created",
			serializeJSON( {
				"token"      : "valid_token",
				"expiration" : dateAdd( "d", 1, now() )
			} )
		);
		},

		...
		});

This seems to happen on OSX but not on Linux. I have not tested on Windows. Once the replacement of = to the : is made, then all of the formatting afterwards is also messed up. This happens on all versions from 0.16.x and up.

[jcberquist]

In my testing, I can reproduce this on Linux. This is a limitation of the parser I am using, which is really more suited to syntax highlighting. It runs line by line, so it is not correctly identifying the arrow function and sees itself as being in a struct (and thus converts the valid = key-value separator to the default :). You can make cfformat ignore this section of the code by bracketing the whole hyper.fake() call in ignore comments (https://github.com/jcberquist/commandbox-cfformat?tab=readme-ov-file#ignoring-code-sections), but I know that's not very satisfying.

I do think it would be possible to improve this situation by identifying the => as the arrow function operator even when the arguments are "missed", and then it should be possible to know that a { following denotes the start of a statement block. But that still won't handle the arrow function parameters correctly, and the real solution is to use an actual parser to format from, and so I look forward to that being implemented via the AST one can get from BoxLang.