Macro returns unexpected floating point number

[itirabasso]

Hi! I'm working on a buidler plugin for solpp to integrate it on the compilation pipeline, but I'm having a few problems with some macros that I found out on the readme's example.

I have this contract A.sol

pragma solidity >=0.4.21 <0.6.0;

contract A {
  // #def POW(a, b) a ** b
  constructor() public {
    uint n1 = $(POW(2, 3)) + $$(POW(16, 0.5));
  }
}

and then I run

npm init
npm install solpp@^0.9.1
npx solpp A.sol 

this retrieves the following output:

pragma solidity >=0.4.21<0.6.0;

contract A {
  constructor() public {
    uint n1 = 2 ** 3 + 3.99999999999999999999999999999999999999999999999999999999999999999999999999999999999999999999999999999999999999999999999;
  }
}

which makes the compilation fail.

[merklejerk]

Exponentiation is always done in floating point, so there is no guarantee you'll get back an integer with a perfect root.
I would recommend using round() to ensure you always end up with an integer, if that's what's needed. Something like:

contract A {
  // #def POW(a, b) a ** b
  constructor() public {
    uint n1 = $(POW(2, 3)) + $$(round(POW(16, 0.5)));
  }
}