Incorrect query matches on a condition including "typeid"

[dnns92]

Dear Community,

I receive an incorrect query-capture using the treesitter api on a rare if-statement edge case. I want to grab the condition inside an if-statement. In this particular case, there is an if-statement, that contains the keyword "typeid". This leads to an incorrect grab, it doesnt find the end of the if-statement correctly and parses until the next if-statement finishes.

My Setup:

Win10
py3.8
tree-sitter-python with compiled tree-sitter-cpp language

I am using this query in order to grab the condition inside the if-statement:
DEFAULT_QUERY = "(if_statement(condition_clause)@if_statement)"

this works really well every case I tried out except one (admittedly rare) case I encountered so far. Consider this valid c++:

bool including_namespace_identifier_and_primitive_type(int x, int y) {
     if(std::strcmp(typeid(unsigned int).name(), something_else)))
        return false;
     // this is a comment that will be included IN THE result of the query!
     if(the.catched.condition.will.include.this_statement)
        return false;
     doSomething();
     return true;
}

using the query above and code[capture.start_byte : capture.stop_byte] we get this captured continous statement:

(std::strcmp(typeid(unsigned int).name(), something_else)))

        return false;

     // this is a comment that will be included IN THE result of the query!

     if(the.catched.condition.will.include.this_statement)

Instead of getting two seperate statements. I am not into generating valid grammar files so far, I hope its an easy fix for someone.

Example where this kind of c++ is used: https://github.com/apple/turicreate/blob/master/src/external/boost/boost_1_68_0/libs/interprocess/test/segment_manager_test.cpp#L337

---------------------------------------------- Some more Notes ----------------------------

If I delete the typeid it works as intended, see:

bool including_namespace_identifier_and_primitive_type(int x, int y) {
     if(std::strcmp((unsigned int).name()))
        return false;
     // this is a comment that will be included IN THE result of the query!
     if(the.catched.condition.will.include.this_statement)
        return false;
     doSomething();
     return true;
}

This finds two seperate statments:

(std::strcmp((unsigned int).name())) 
(the.catched.condition.will.include.this_statement)

[maxbrunsfeld]

Can you report the output of tree-sitter parse <the-file> for the code that you're matching? It sounds like there is just a syntax error happening, so that we have a tree that's different from expected.

[dnns92]

I only have the python bindungs installed.
I think I can provide you the corresponding info like this:

exanples/weird.stuff.cpp

bool including_namespace_identifier_and_primitive_type(int x, int y) {
     if(std::strcmp(typeid(unsigned int).name(), something_else)))
        return false;
     // this is a comment that will be included IN THE result of the query!
     if(the.catched.condition.will.include.this_statement)
        return false;
     doSomething();
     return true;
}

python-script

from tree_sitter import Language, Parser
import pickle

relative_path_to_parser = 'build/my-languages.so'
file = "examples/weird_stuff.cpp"
LANGUAGE = Language(relative_path_to_parser, "cpp")
query_statement = "(if_statement(condition_clause)@if_statement)"


#%% read file
with open(file, "rb") as f:
    data = f.read()
f.close()

#%% parse
parser = Parser()
parser.set_language(LANGUAGE)
tree = parser.parse(data)
query = LANGUAGE.query(query_statement)
for finding in query.captures(tree.root_node):
    print(finding[0].sexp())

This returns:
(condition_clause initializer: (declaration type: (scoped_type_identifier namespace: (namespace_identifier) name: (type_identifier)) declarator: (parenthesized_declarator (ERROR (function_declarator declarator: (identifier) parameters: (parameter_list (parameter_declaration type: (sized_type_specifier type: (primitive_type))))) (function_declarator declarator: (identifier) parameters: (parameter_list))) (identifier)) (ERROR (false))) (comment) value: (call_expression function: (identifier) arguments: (argument_list (field_expression argument: (field_expression argument: (field_expression argument: (field_expression argument: (field_expression argument: (identifier) field: (field_identifier)) field: (field_identifier)) field: (field_identifier)) field: (field_identifier)) field: (field_identifier)))) (MISSING ")"))

Is this was you expected? I'm not very familiar with the API yet. Thank you very much in advance!
Cheers!