CSCI 241 Systems Programming Labs

There are many ways to implement this. Here’s one way.

Since redirections consist of an optional digit, a redirection operator, <, >, or >>, and a path, it would be nice to reuse the code you’ve already written to parse Words to parse the paths. You can do this by maintaining additional state about what type of redirection is currently being parsed.

#![allow(unused)]
fn main() {
enum RedirectionState {
    None,
    Input(u32),
    Output(u32),
    Append(u32),
}
}

The None variant is when a normal word is being parsed. The u32 keeps track of which file descriptor will be redirected.

In addition, you will need to be able to distinguish input 5>path from 5x. Since the state machine looks at the input one character at a time, you’ll need to modify State to have an additional variant, Digit which represents the case where a digit was read in the Blank state.

#![allow(unused)]
fn main() {
enum State {
    Blank,
    Unquoted,
    SingleQuoted,
    DoubleQuoted,
    Digit(u32),
}
}

Here’s an updated state machine.

What isn’t represented is how the RedirectionState changes.

The intended behavior is in the Blank state, if the next input character is a digit, transition to state Digit(num) where num is the number.

In the Digit(num) state, if the next input character is < or >, set the redirection state to RedirectionState::Input(num) or RedirectionState::Output(num) and change to the Unquoted state.

In the Digit(num) state, if the next input character is a quotation mark, write num into the current word and transition to the appropriate quoted state. This is handling input like 3"foo"

Otherwise, in the Digit(num) state, if the next input character is anything else, write num into the current word and transition to the Unquoted state. This handles input like 4bar.

The transitions from the Blank and Unquoted states need to be updated to handle the redirection state.

In the Blank state, if the next input character is < or >, the redirection state should be set to RedirectionState::Input(0) or RedirectionState::Output(1) and transition to the Unquoted state.

In the Unquoted state, if the next input character is a |, then you either need to insert a Word token or one of the Redirect* token into the results vector. And then you need to insert a Pipe token. Here’s some sample code for this.

    (State::Unquoted, '|') => {
        let token = match redirection_state {
            RedirectionState::None => InputToken::Word(current_word),
            _ if current_word.is_empty() => return Err("Empty redirection target".into()),
            RedirectionState::Input(num) => InputToken::RedirectInput(num, current_word),
            RedirectionState::Output(num) => InputToken::RedirectOutput(num, current_word),
            RedirectionState::Append(num) => InputToken::RedirectAppend(num, current_word),
        };
        result.push(token);
        result.push(InputToken::Pipe);
        current_word = String::new();
        state = State::Blank;
        redirection_state = RedirectionState::None;
    }

Note that if the current word is empty, and a redirection is being parsed, this is an error. I.e., the input 2>| is an error.

In the Unquoted state, if the next input character is whitespace, then this behaves similar to the | case except you don’t push a Pipe token.

In the Unquoted state, if the next input character is a < or a >, then based on the redirection state, you need to do one of the following:

1. insert a new Word token
2. insert a new Redirect* token
3. change the redirection state from RedirectionState::Output(num) to RedirectionState::Append(num)
4. throw an Error

The third option occurs when the redirection state is RedirectionState::Output(num) and the current word is empty. For example, when parsing input 3>>foo, the first > changes the redirection state to Output(3) and the state to Unquoted. The second > changes the redirection state to Append(3).

For example, the input >foo>>bar<baz should be parsed as three separate redirection tokens.

Finally, after all of the input has been consumed, based on the state and the redirection state, you may need to insert either a Word or a Redirect* token. You will also need to update your code to handle the Digit state.