Parsing issue: Back ticks are preserved in identifier.

[traceyyoshima]

import java.lang.Integer.MAX_VALUE
val m = `MAX_VALUE`

[kunli2]

from PSI perspective,

`MAX_VALUE`

is indeed an identifier, so I think it's fine.

    \---(35,54) | PROPERTY | KtProperty | Text: "val m = `MAX_VALUE`"
        |---(35,38) | val | LeafPsiElement | Text: "val"
        |---(38,39) | WHITE_SPACE | PsiWhiteSpaceImpl | Text: " "
        |---(39,40) | IDENTIFIER | LeafPsiElement | Text: "m"
        |---(40,41) | WHITE_SPACE | PsiWhiteSpaceImpl | Text: " "
        |---(41,42) | EQ | LeafPsiElement | Text: "="
        |---(42,43) | WHITE_SPACE | PsiWhiteSpaceImpl | Text: " "
        \---(43,54) | REFERENCE_EXPRESSION | KtNameReferenceExpression | Text: "`MAX_VALUE`"
            \---(43,54) | IDENTIFIER | LeafPsiElement | Text: "`MAX_VALUE`"

[traceyyoshima]

from PSI perspective,

MAX_VALUE is indeed an identifier, so I think it's fine.

Unfortunately, the PSI is not a reliable source to make decisions because it's generated by a high-level lexer later interpreted by the compiler.

Here is a link to the documentation: https://kotlinlang.org/docs/java-interop.html#escaping-for-java-identifiers-that-are-keywords-in-kotlin.

Consider what is happening in the code. The identifier

`MAX_VALUE`

is from Integer.MAX_VALUE. The MAX_VALUE field on Integer does not have the escapes, but the Kotlin code happens to be escaped. This is important because recipes like ChangeType will update the name if the identifier and type match.

Using ChangeType from Map to List. I'd expected a change:

`Map`

to

`List`

Other applicable cases, like escaped method names, would fail if we used the raw PSI text. The example is to show the issue, but identifiers should generally contain the correct text.

fun `escapedMethodName`(){}

[knutwannheden]

I think we should add a Quoted marker for this purpose. Many languages have that concept.

[traceyyoshima]

I think we should add a Quoted marker for this purpose. Many languages have that concept.

That sounds good -- I know this happens in rewrite-javascript as well, though I don't recall the rules for other languages off the top of my head.

That'll allow changes to occur on the identifier while preserving the escapes. And we could technically detect unnecessary escapes and remove them in a clean-up recipe.