archify-vscode-ext-sync/crates/server/src/doc.rs

//! A space's Yjs document: one `Y.Text` per model file in the `files` map.

use std::collections::BTreeMap;

use archify_sync_protocol::FILES_MAP;
use archify_sync_store::StoredDoc;
use yrs::updates::decoder::Decode;
use yrs::{
    Doc, GetString, Map, OffsetKind, Options, Out, ReadTxn, StateVector, Text, TextPrelim,
    Transact, Update,
};

/// A document that counts text offsets in UTF-16, as Yjs in the extension does.
pub fn new_doc() -> Doc {
    Doc::with_options(Options {
        offset_kind: OffsetKind::Utf16,
        ..Options::default()
    })
}

pub fn load(stored: &StoredDoc) -> anyhow::Result<Doc> {
    let doc = new_doc();
    {
        let mut txn = doc.transact_mut();
        if let Some(state) = &stored.state {
            txn.apply_update(Update::decode_v1(state)?)?;
        }
        for update in &stored.updates {
            txn.apply_update(Update::decode_v1(update)?)?;
        }
    }
    Ok(doc)
}

pub fn from_state(state: &[u8]) -> anyhow::Result<Doc> {
    let doc = new_doc();
    doc.transact_mut().apply_update(Update::decode_v1(state)?)?;
    Ok(doc)
}

pub fn encode_state(doc: &Doc) -> Vec<u8> {
    doc.transact()
        .encode_state_as_update_v1(&StateVector::default())
}

/// Path → text of every model file.
pub fn files(doc: &Doc) -> BTreeMap<String, String> {
    let map = doc.get_or_insert_map(FILES_MAP);
    let txn = doc.transact();
    map.iter(&txn)
        .filter_map(|(path, value)| match value {
            Out::YText(text) => Some((path.to_owned(), text.get_string(&txn))),
            _ => None,
        })
        .collect()
}

/// Makes `doc` hold the files of `target` as ordinary edits, so every copy
/// takes them like any other change. Returns the update, empty when nothing changed.
pub fn replace_files(doc: &Doc, target: &BTreeMap<String, String>) -> Vec<u8> {
    let map = doc.get_or_insert_map(FILES_MAP);
    let mut txn = doc.transact_mut();
    let current: Vec<String> = map.keys(&txn).map(str::to_owned).collect();
    for path in current.iter().filter(|p| !target.contains_key(*p)) {
        map.remove(&mut txn, path);
    }
    for (path, text) in target {
        match map.get(&txn, path) {
            Some(Out::YText(existing)) => {
                let old = existing.get_string(&txn);
                if let Some((at, remove, insert)) = minimal_change(&old, text) {
                    if remove > 0 {
                        existing.remove_range(&mut txn, at, remove);
                    }
                    if !insert.is_empty() {
                        existing.insert(&mut txn, at, insert);
                    }
                }
            }
            _ => {
                map.insert(&mut txn, path.as_str(), TextPrelim::new(text.as_str()));
            }
        }
    }
    txn.encode_update_v1()
}

/// The one edit that turns `old` into `new`: UTF-16 offset, UTF-16 length to
/// remove, text to insert. Keeps the common start and end, as the extension's
/// `minimalChange` does, and never splits a character.
pub fn minimal_change<'a>(old: &str, new: &'a str) -> Option<(u32, u32, &'a str)> {
    if old == new {
        return None;
    }
    let prefix: usize = old
        .chars()
        .zip(new.chars())
        .take_while(|(a, b)| a == b)
        .map(|(a, _)| a.len_utf8())
        .sum();
    let (old_rest, new_rest) = (&old[prefix..], &new[prefix..]);
    let suffix: usize = old_rest
        .chars()
        .rev()
        .zip(new_rest.chars().rev())
        .take_while(|(a, b)| a == b)
        .map(|(a, _)| a.len_utf8())
        .sum();
    let removed = &old_rest[..old_rest.len() - suffix];
    let inserted = &new_rest[..new_rest.len() - suffix];
    let utf16 = |s: &str| s.encode_utf16().count() as u32;
    Some((utf16(&old[..prefix]), utf16(removed), inserted))
}

#[cfg(test)]
mod tests {
    use super::*;

    fn texts(pairs: &[(&str, &str)]) -> BTreeMap<String, String> {
        pairs
            .iter()
            .map(|(k, v)| (k.to_string(), v.to_string()))
            .collect()
    }

    #[test]
    fn minimal_change_keeps_common_ends() {
        assert_eq!(
            minimal_change("label: API", "label: Gateway API"),
            Some((7, 0, "Gateway "))
        );
        assert_eq!(minimal_change("abc", "abc"), None);
        assert_eq!(minimal_change("a😀b", "a😀cb"), Some((3, 0, "c")));
        assert_eq!(minimal_change("x😀y", "x😃y"), Some((1, 2, "😃")));
    }

    #[test]
    fn replace_files_reaches_target_and_syncs() {
        let doc = new_doc();
        replace_files(
            &doc,
            &texts(&[("main.archify", "model: a\n"), ("old.archify", "x")]),
        );
        let other = from_state(&encode_state(&doc)).unwrap();

        let target = texts(&[("main.archify", "model: b\n"), ("new.archify", "y")]);
        let update = replace_files(&doc, &target);
        assert_eq!(files(&doc), target);

        other
            .transact_mut()
            .apply_update(Update::decode_v1(&update).unwrap())
            .unwrap();
        assert_eq!(files(&other), target);
        assert!(
            replace_files(&doc, &target).len() <= 2,
            "no change gives an empty update"
        );
    }
}