Option to preserve trailing whitespace

[heiskane]

Currently working on something that requires me to keep track of a cursor in a message and wrap the message in a textbox with the cursor in the right place. This currently is not possible as wrapping trims the trailing whitespace making wrapped text different length from the original message misaligning any position i might track in the message.

Trimming trailing whitespace could be disabled in the options?

wanted behavior.

let text = "as     a hello";
let wrapped = textwrap::wrap(text, 5);
// ["as   ", "  a ", "hello"]

[mgeisler]

Hi @heiskane,

Currently working on something that requires me to keep track of a cursor in a message and wrap the message in a textbox with the cursor in the right place.

Interesting! Since you're doing something custom, have you considered using the Word struct directly? You might even want to implement Fragment for your own custom type.

All the wrapping logic operates on fragments: the "upper layer" simply splits the text into fragments using different options.

I think your use case could be handled by

• creating a Fragment for every space character and
• moving the whitespace to the "word" part of the Fragment instead of the "whitespace" part

Today, the string "as a hello" is turned into words and whitespace like this:

[("as", "     "), ("a", " "), ("hello". "")]

I'm suggesting instead turning it into

[("as", ""), (" ", ""), (" ", ""), (" ", ""), (" ", ""), (" ", ""), ("a", ""), (" ", ""), ("hello", "")]

That should do the trick. You should be able to do this with the WordSeparator::Custom setting.

[heiskane]

Thanks for the reply. I experimented a bit with what you suggested but i am still having some issues.

The issue with this is that WordSeparator::Custom takes a closure that returns iterator over Words and Word::from() strips whitespace so you can't really do something like Word::from(" "). I can't construct Word manually either because it needs width which is a private field. Not sure how i can make this happen with these restrictions.

If WordSeparator::Custom took a closure that returns an iterator over any type that implements Fragment i feel like this would be simpler to solve. Although i do feel like it would be nice if the library offered a NoTrim option directly.

Update:
I implemented a simple method in textwrap to create Word without trimming the whitespace and that did allow me to do what i described in the issue.

the method:

impl<'a> Word<'a> {
    // snip....
    pub fn no_trim(word: &str) -> Word<'_> {
        Word {
            word,
            width: display_width(word),
            whitespace: "",
            penalty: "",
        }
    }
}

my (quick and dirty) solution to the issue:

use textwrap::{core::Word, Options, WordSeparator};

fn main() {
    let text = "as     a hello";

    let opts = Options::new(5).word_separator(WordSeparator::Custom(|line| {
        let mut start = 0;
        let mut prev_char = ' ';
        let mut char_indices = line.char_indices();

        println!("line: {:?}", line);
        Box::new(std::iter::from_fn(move || {
            for (idx, ch) in char_indices.by_ref() {
                if prev_char == ' ' && start != 0 {
                    let word = Word::no_trim(" ");
                    println!("word0: {:?}", word);
                    prev_char = ch;
                    start = idx;
                    return Some(word);
                }
                if prev_char != ' ' && ch == ' ' {
                    let word = Word::from(&line[start..idx]);
                    println!("word1: {:?}", word);
                    start = idx;
                    prev_char = ch;
                    return Some(word);
                }
                prev_char = ch;
            }

            if start < line.len() {
                let word = Word::no_trim(&line[start..]);
                println!("word2: {:?}", word);
                start = line.len();
                return Some(word);
            }

            None
        }))
    }));
    let wrapped = textwrap::wrap(text, opts);
    println!("{wrapped:?}");
}

prints:

line: "as     a hello"
word1: Word { word: "as", whitespace: "", penalty: "", width: 2 }
word0: Word { word: " ", whitespace: "", penalty: "", width: 1 }
word0: Word { word: " ", whitespace: "", penalty: "", width: 1 }
word0: Word { word: " ", whitespace: "", penalty: "", width: 1 }
word0: Word { word: " ", whitespace: "", penalty: "", width: 1 }
word0: Word { word: " ", whitespace: "", penalty: "", width: 1 }
word1: Word { word: "a", whitespace: "", penalty: "", width: 1 }
word0: Word { word: " ", whitespace: "", penalty: "", width: 1 }
word2: Word { word: "hello", whitespace: "", penalty: "", width: 5 }
["as  ", "   a ", "hello"]

Could this method (or something similar) be added to the crate so that wrapping without trimming would be possible? I can submit a PR if the method i created is fine (and maybe create a new WordSeparator type for this?). Let me know how you want to move forward with this 👍

[nstolpe]

@mgeisler would you be open to a PR that makes Word.width pub instead of pub(crate)? I have a similar use case to @heiskane and that's the only change necessary to use a custom WordSeperator.

[JesseCSlater]

I am also trying to track a cursor position, so I would also appreciate if wrapping without trimming was possible. A built in NoTrim option would be ideal, but just making it possible by making Word.width pub would also be excellent. Currently I have just rolled my own text wrapping, but it's Ascii only and greedy, so I'd love to be able to use textwrap.

[mgeisler]

Cc @HidenoriKobayashi who told me he needs the same functionality for a dprint plugin to format PO files (for https://github.com/google/comprehensive-rust).

[besquared]

+1

Also ran into this trying to implement cursor management in a multi-line tui input box. Love the diff linked from @HidenoriKobayashi in the other issue (assuming tests pass etc.)

[JesseCSlater]

We now have the option to preserve trailing space characters, but not newline characters. I would like to be able to preserve these as well. My use case is a typing practice app.

I know this can be achieved by using wrap_optimal_fit or wrap_first_fit directly, but I wish it was the default behavior of wrap with preserve_trailing_space.

If this was the default but a user wanted to preserve spaces but not newlines, they could easily do it by splitting the string on newlines and then calling wrap on each piece separately, or just by removing newlines after wrapping.

@mgeisler, if you are open to this change, I would be happy to implement it.