Problem

I observed a strange phenomenon in Emacs and can reproduce it in C. Given the following rust source file:

fn foo() -> i32 {
    4
}

fn main() {
    println!("{:?}", foo());
}

We create a rust parser and set its ranges so it only contains ("{:?}", foo()), ie, the part after the println! macro but before the semicolon.

At this point, querying the parser works fine.

Now, insert a whitespace at the end of the line, ie, after the semicolon. Now the code looks like this (<space> represents a whitespace):

fn foo() -> i32 {
    4
}

fn main() {
    println!("{:?}", foo());<space>
}

Now the parser doesn't return any matches for any query. Note that we didn't change the ranges for the parser. All that we did is to append a whitespace at the end of the line, and re-parse.

Steps to reproduce

I created a reproduce repo here: https://github.com/casouri/tree-sitter-rust-bug-repo

As always, running ./run.sh should download latest tree-sitter and tree-sitter-rust, and compile the reproduce recipe program. If everything works right, the console output should look like this:

First parse root node:
(source_file (expression_statement (tuple_expression (string_literal (string_content)) (call_expression function: (identifier) arguments: (arguments))) (MISSING ";")))
First parse query: has match
Second parse root node:
(source_file (expression_statement (tuple_expression (string_literal (string_content)) (call_expression function: (identifier) arguments: (arguments))) (MISSING ";")))
Second parse query: no match

Note that the node string from two parses are the same. After the first parse, we can find matches for the query; after the second parse, we no long get matches for the same query.

The query is a simple query: (call_expression) @cap

Expected behavior

The console should print Second parse query: has match instead of Second parse query: no match

Tree-sitter version (tree-sitter --version)

tree-sitter master (325b320)

Operating system/version

macOS 14.6.1 (23G93)