When (and when not) to use the new keyword in Solidity - solidity

Question

I have two contracts in Solidity like so (based upon a tutorial):
pragma solidity >=0.4.16 <0.9.0;
contract ERC20Token {
string public name;
mapping(address => uint) public balances;
function mint() public {
balances[tx.origin]++;
}
}
contract MyContract {
address payable wallet;
address public token;
constructor(address payable _wallet, address _token) {
wallet = _wallet;
token = _token;
}
function buyToken() public payable {
ERC20Token _token = ERC20Token(address(token));
_token.mint();
wallet.transfer(msg.value);
}
}
My question is why they use ERC20Token _token = ERC20Token(address(token)); in order to access the class and call he .mint() function.
My base reflex would be to do the following instead:
ERC20Token _token = new ERC20Token();
_token.mint();
This is because the ERC20Token contract does not have a constructor, so I'm unsure where the RC20Token(address(token)) constructor parameters come from.

Answer 1

The new ERC20Token() expression deploys a new ERC20Token contract and returns its (newly deployed) address.
Since your ERC20Token doesn't have any constructor (and constructor params), you don't need to pass any params. But if it did have constructor params, you'd need to pass them. Example:
contract ERC20Token {
constructor (string memory _name, string memory _symbol) {
}
}
address deployedTo = new ERC20Token("MyToken", "MyT");
The ERC20Token _token = ERC20Token(<address>) creates a helper object allowing the _token to be treated as a contract implementing the ERC20Token interface, so you can execute and call its public and external functions.
It's the address of the external contract that you're passing - not the constructor params.
contract ERC20Token {
function mint() public {
}
}
ERC20Token _token = ERC20Token(address(token));
_token.mint();