OT: changeset: Implement compose and apply.

This commit is contained in:
Blaž Hrastnik 2020-05-26 18:11:11 +09:00
parent 44ff4d3c1f
commit 23109f1512
5 changed files with 330 additions and 22 deletions

44
Cargo.lock generated
View file

@ -172,6 +172,16 @@ version = "1.0.7"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "3f9eec918d3f24069decb9af1554cad7c880e2da24a9afd88aca000531ab82c1"
[[package]]
name = "futf"
version = "0.1.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "7c9c1ce3fa9336301af935ab852c437817d14cd33690446569392e65170aac3b"
dependencies = [
"mac",
"new_debug_unreachable",
]
[[package]]
name = "getrandom"
version = "0.1.14"
@ -199,6 +209,7 @@ dependencies = [
"anyhow",
"ropey",
"smallvec 1.4.0",
"tendril",
]
[[package]]
@ -232,6 +243,12 @@ dependencies = [
"cfg-if",
]
[[package]]
name = "mac"
version = "0.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "c41e0c4fef86961ac6d6f8a82609f55f31b05e4fce149ac5710e439df7619ba4"
[[package]]
name = "maybe-uninit"
version = "2.0.0"
@ -250,6 +267,12 @@ version = "0.1.1"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "a64a92489e2744ce060c349162be1c5f33c6969234104dbd99ddb5feb08b8c15"
[[package]]
name = "new_debug_unreachable"
version = "1.0.4"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "e4a24736216ec316047a1fc4252e27dabb04218aa4a3f37c6e7ddbf1f9782b54"
[[package]]
name = "nom"
version = "5.1.1"
@ -458,10 +481,9 @@ checksum = "7fe5bd57d1d7414c6b5ed48563a2c855d995ff777729dcd91c369ec7fea395ae"
[[package]]
name = "ropey"
version = "1.1.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "ba326a8508a4add47e7b260333aa2d896213a5f3572fde11ed6e9130241b7f71"
source = "git+https://github.com/cessen/ropey#083c34949274ef9800267e6bc64b76a45e401807"
dependencies = [
"smallvec 0.6.13",
"smallvec 1.4.0",
]
[[package]]
@ -554,6 +576,16 @@ dependencies = [
"unicode-xid 0.2.0",
]
[[package]]
name = "tendril"
version = "0.4.1"
source = "git+https://github.com/servo/tendril#08f7f292ab82c00e9cf491b5918a76e53af92c8c"
dependencies = [
"futf",
"mac",
"utf-8",
]
[[package]]
name = "terminfo"
version = "0.7.3"
@ -639,6 +671,12 @@ version = "0.2.0"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "826e7639553986605ec5979c7dd957c7895e93eabed50ab2ffa7f6128a75097c"
[[package]]
name = "utf-8"
version = "0.7.5"
source = "registry+https://github.com/rust-lang/crates.io-index"
checksum = "05e42f7c18b8f902290b009cde6d651262f956c98bc51bca4cd1d511c9cd85c7"
[[package]]
name = "utf8parse"
version = "0.1.1"

View file

@ -7,7 +7,9 @@ edition = "2018"
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
[dependencies]
ropey = "1.1.0"
# ropey = "1.1.0"
ropey = { git = "https://github.com/cessen/ropey" }
anyhow = "1.0.31"
smallvec = "1.4.0"
tendril = { git = "https://github.com/servo/tendril" }
# slab = "0.4.2"

View file

@ -10,4 +10,4 @@ pub use selection::Selection;
pub use state::State;
pub use transaction::{Change, Transaction};
pub use transaction::{Change, ChangeSet, Transaction};

View file

@ -83,7 +83,7 @@ impl Range {
/// A selection consists of one or more selection ranges.
pub struct Selection {
// TODO: decide how many ranges to inline SmallVec<[Range; 1]>
ranges: Vec<Range>,
ranges: SmallVec<[Range; 1]>,
primary_index: usize,
}
@ -100,7 +100,7 @@ impl Selection {
self
} else {
Self {
ranges: vec![self.ranges[self.primary_index]],
ranges: smallvec![self.ranges[self.primary_index]],
primary_index: 0,
}
}
@ -112,18 +112,18 @@ impl Selection {
/// Constructs a selection holding a single range.
pub fn single(anchor: usize, head: usize) -> Self {
Self {
ranges: vec![Range { anchor, head }],
ranges: smallvec![Range { anchor, head }],
primary_index: 0,
}
}
pub fn new(ranges: Vec<Range>, primary_index: usize) -> Self {
fn normalize(mut ranges: Vec<Range>, primary_index: usize) -> Selection {
pub fn new(ranges: SmallVec<[Range; 1]>, primary_index: usize) -> Self {
fn normalize(mut ranges: SmallVec<[Range; 1]>, primary_index: usize) -> Selection {
let primary = ranges[primary_index];
ranges.sort_unstable_by_key(|range| range.from());
let mut primary_index = ranges.iter().position(|&range| range == primary).unwrap();
let mut result: Vec<Range> = Vec::new();
let mut result: SmallVec<[Range; 1]> = SmallVec::new();
// TODO: we could do with one vec by removing elements as we mutate
@ -174,7 +174,7 @@ mod test {
#[test]
fn test_create_normalizes_and_merges() {
let sel = Selection::new(
vec![
smallvec![
Range::new(10, 12),
Range::new(6, 7),
Range::new(4, 5),
@ -200,7 +200,7 @@ mod test {
#[test]
fn test_create_merges_adjacent_points() {
let sel = Selection::new(
vec![
smallvec![
Range::new(10, 12),
Range::new(12, 12),
Range::new(12, 12),

View file

@ -1,25 +1,293 @@
pub struct Change {
from: usize,
to: usize,
insert: Option<String>,
// pub struct Change {
// from: usize,
// to: usize,
// insert: Option<String>,
// }
// 40 bytes (8 + 24 + 8) -> strings are really big 24 as String, 16 as &str
// pub struct Change {
// /// old extent
// old_extent: usize,
// /// inserted text, new extent equal to insert length
// insert: Option<String>,
// /// distance from the previous change
// distance: usize,
// }
use crate::{Buffer, Selection};
use ropey::Rope;
use tendril::StrTendril as Tendril;
// TODO: divided into three different operations, I sort of like having just
// Splice { extent, Option<text>, distance } better.
// insert: Splice { extent: 0, text: Some("a"), distance: 2 }
// delete: Splice { extent: 2, text: None, distance: 2 }
// replace: Splice { extent: 2, text: Some("abc"), distance: 2 }
// unchanged?: Splice { extent: 0, text: None, distance: 2 }
// harder to compose though.
#[derive(Debug, Clone, PartialEq, Eq)]
pub enum Change {
/// Move cursor by n characters.
Retain(usize),
/// Delete n characters.
Delete(usize),
/// Insert text at position.
Insert(Tendril),
}
impl Change {
pub fn new(from: usize, to: usize, insert: Option<String>) {
pub fn new(from: usize, to: usize, insert: Option<Tendril>) {
// old_extent, new_extent, insert
}
}
pub struct Transaction {}
// ChangeSpec = Change | ChangeSet | Vec<Change>
// ChangeDesc as a ChangeSet without text: can't be applied, cheaper to store.
// ChangeSet = ChangeDesc with Text
#[derive(Debug)]
pub struct ChangeSet {
// basically Vec<ChangeDesc> where ChangeDesc = (current len, replacement len?)
// (0, n>0) for insertion, (n>0, 0) for deletion, (>0, >0) for replacement
sections: Vec<(usize, isize)>,
// sections: Vec<(usize, isize)>,
changes: Vec<Change>,
/// The required document length. Will refuse to apply changes unless it matches.
len: usize,
}
//
impl ChangeSet {
pub fn new(buf: &Buffer) -> Self {
let len = buf.contents.len_chars();
Self {
changes: vec![Change::Retain(len)],
len,
}
}
// TODO: from iter
/// Combine two changesets together.
/// In other words, If `this` goes `docA` → `docB` and `other` represents `docB` → `docC`, the
/// returned value will represent the change `docA` → `docC`.
pub fn compose(self, other: ChangeSet) -> Result<Self, ()> {
if self.len != other.len {
// length mismatch
return Err(());
}
let len = self.changes.len();
let mut changes_a = self.changes.into_iter();
let mut changes_b = other.changes.into_iter();
let mut head_a = changes_a.next();
let mut head_b = changes_b.next();
let mut changes: Vec<Change> = Vec::with_capacity(len); // TODO: max(a, b), shrink_to_fit() afterwards
loop {
use std::cmp::Ordering;
use Change::*;
match (head_a, head_b) {
// we are done
(None, None) => {
break;
}
// deletion in A
(Some(change @ Delete(..)), b) => {
changes.push(change);
head_a = changes_a.next();
head_b = b;
}
// insertion in B
(a, Some(change @ Insert(..))) => {
changes.push(change);
head_a = a;
head_b = changes_b.next();
}
(None, _) => return Err(()),
(_, None) => return Err(()),
(Some(Retain(i)), Some(Retain(j))) => match i.cmp(&j) {
Ordering::Less => {
changes.push(Retain(i));
head_a = changes_a.next();
head_b = Some(Retain(j - i));
}
Ordering::Equal => {
changes.push(Retain(i));
head_a = changes_a.next();
head_b = changes_b.next();
}
Ordering::Greater => {
changes.push(Retain(j));
head_a = Some(Retain(i - j));
head_b = changes_b.next();
}
},
(Some(Insert(mut s)), Some(Delete(j))) => {
let len = s.chars().count();
match len.cmp(&j) {
Ordering::Less => {
head_a = changes_a.next();
head_b = Some(Delete(j - len));
}
Ordering::Equal => {
head_a = changes_a.next();
head_b = changes_b.next();
}
Ordering::Greater => {
// figure out the byte index of the truncated string end
let (pos, _) = s.char_indices().nth(len - j).unwrap();
// calculate the difference
let to_drop = s.len() - pos;
s.pop_back(to_drop as u32);
head_a = Some(Insert(s));
head_b = changes_b.next();
}
}
}
(Some(Insert(mut s)), Some(Retain(j))) => {
let len = s.chars().count();
match len.cmp(&j) {
Ordering::Less => {
changes.push(Insert(s));
head_a = changes_a.next();
head_b = Some(Retain(j - len));
}
Ordering::Equal => {
changes.push(Insert(s));
head_a = changes_a.next();
head_b = changes_b.next();
}
Ordering::Greater => {
// figure out the byte index of the truncated string end
let (pos, _) = s.char_indices().nth(j).unwrap();
// calculate the difference
let to_drop = s.len() - pos;
s.pop_back(to_drop as u32);
head_a = Some(Insert(s));
head_b = changes_b.next();
}
}
}
(Some(Retain(i)), Some(Delete(j))) => match i.cmp(&j) {
Ordering::Less => {
changes.push(Delete(i));
head_a = changes_a.next();
head_b = Some(Delete(j - i));
}
Ordering::Equal => {
changes.push(Delete(j));
head_a = changes_a.next();
head_b = changes_b.next();
}
Ordering::Greater => {
changes.push(Delete(j));
head_a = Some(Retain(i - j));
head_b = changes_b.next();
}
},
};
}
Ok(Self {
len: self.len,
changes,
})
}
/// Given another change set starting in the same document, maps this
/// change set over the other, producing a new change set that can be
/// applied to the document produced by applying `other`. When
/// `before` is `true`, order changes as if `this` comes before
/// `other`, otherwise (the default) treat `other` as coming first.
///
/// Given two changes `A` and `B`, `A.compose(B.map(A))` and
/// `B.compose(A.map(B, true))` will produce the same document. This
/// provides a basic form of [operational
/// transformation](https://en.wikipedia.org/wiki/Operational_transformation),
/// and can be used for collaborative editing.
pub fn map(self, other: Self) -> Self {
unimplemented!()
}
/// Returns a new changeset that reverts this one. Useful for `undo` implementation.
pub fn invert(self) -> Self {
unimplemented!()
}
pub fn apply(&self, text: &mut Rope) {
// TODO: validate text.chars() == self.len
let mut pos = 0;
for change in self.changes.iter() {
use Change::*;
match change {
Retain(n) => {
pos += n;
}
Delete(n) => {
text.remove(pos..pos + *n);
// pos += n;
}
Insert(s) => {
text.insert(pos, s);
pos += s.len();
}
}
}
}
// iter over changes
}
// trait Transaction
// trait StrictTransaction
/// Transaction represents a single undoable unit of changes. Several changes can be grouped into
/// a single transaction.
pub struct Transaction {
/// Changes made to the buffer.
changes: ChangeSet,
/// When set, explicitly updates the selection.
selection: Option<Selection>,
// effects, annotations
// scroll_into_view
}
#[cfg(test)]
mod test {
use super::*;
#[test]
fn composition() {
use Change::*;
let a = ChangeSet {
changes: vec![
Retain(5),
Insert("!".into()),
Retain(1),
Delete(2),
Insert("ab".into()),
],
len: 7,
};
let b = ChangeSet {
changes: vec![Delete(5), Insert("world".into()), Retain(4)],
len: 7,
};
let mut text = Rope::from("hello xz");
// should probably return cloned text
a.compose(b).unwrap().apply(&mut text);
unimplemented!("{:?}", text);
}
#[test]
fn map() {}
}