Usually -2^2 is considered to be -4, because: the minus is interpreted as a unary operator with lower precedence, than ^ (power)... E.g.: http://www.isthe.com/chongo/tech/comp/calc/
_but_: I would like to have a parser, [1] that binds the sign of a number stronger than a ^ (power), and [2] that binds the unary inversion-operator of an expression weaker than a ^ (power). My parse works, but the bison manual says, that i shalt fix the conflict, because it is somehow worrisome. https://www.gnu.org/software/bison/manual/html_node/Reduce_002fReduce.html example: [1] -2^2=(-2)^2=4 [2] a=2 ; -a^2=-(a^2)=-(2^2)=-4 In this yy-file is a reduce/reduce conflict: %left "+" "-"; %right "^"; exp: "-" "num" { $$ = -*new Float($2); std::cout << "NUMinv" << $$ << std::endl; } | "num" { $$ = new Float($1); std::cout << "num" << $$ << std::endl; } | "+" exp { $$ = $2; std::cout << "noninv" << $$ << std::endl; } | "-" exp { $$ = -*$2; std::cout << "inv" << $$ << std::endl; } | exp "+" exp { $$ = *$1 + $3; std::cout << "add" << $$ << std::endl; } | exp "-" exp { $$ = *$1 - $3; std::cout << "sub" << $$ << std::endl; } | exp "^" exp { $$ = *$1 ^ $3; std::cout << "pow" << $$ << std::endl; }; bison -Wcounterexamples says: Example: "-" "num" • First reduce derivation exp ↳ 4: "-" "num" • Second reduce derivation exp ↳ 7: "-" exp ↳ 5: "num" • How can i solve that without a preprocessor, that wraps negative numbers (e. g.: "-4*(-2+-3)" --> "(-4)*((-2)+(-3))") with brackets? -Arne