Skip to content

Offline, privacy-first grammar checker. Fast, open-source, Rust-powered

fix: improved perf with ref-counted pointers

Elijah Potterb73060845ed0
diff --git a/Cargo.lock b/Cargo.lock
index 3d5d775ab..f357a6d7a 100644
--- a/Cargo.lock
+++ b/Cargo.lock
@@ -530,7 +530,7 @@ dependencies = [
"divan",
"hashbrown",
"is-macro",
- "itertools 0.12.1",
+ "itertools 0.13.0",
"once_cell",
"paste",
"pulldown-cmark",
diff --git a/harper-core/Cargo.toml b/harper-core/Cargo.toml
index a511dff4c..5c8357b2e 100644
--- a/harper-core/Cargo.toml
+++ b/harper-core/Cargo.toml
@@ -11,7 +11,7 @@ repository = "https://github.com/elijah-potter/harper"
blanket = "0.4.0"
hashbrown = { version = "0.14.3", features = ["serde"] }
is-macro = "0.3.0"
-itertools = "0.12.1"
+itertools = "0.13.0"
once_cell = "1.19.0"
paste = "1.0.14"
pulldown-cmark = "0.11.0"
@@ -28,3 +28,7 @@ rand = "0.8.5"
[[bench]]
name = "parse_demo"
harness = false
+
+[features]
+default = []
+concurrent = []
diff --git a/harper-core/src/document.rs b/harper-core/src/document.rs
index 998f0bb68..ef732befd 100644
--- a/harper-core/src/document.rs
+++ b/harper-core/src/document.rs
@@ -1,3 +1,4 @@
+use crate::Lrc;
use std::cmp::Ordering;
use std::collections::VecDeque;
use std::fmt::Display;
@@ -13,9 +14,9 @@ use crate::token::NumberSuffix;
use crate::{FatToken, Token, TokenKind, TokenStringExt};
pub struct Document {
- source: Vec<char>,
+ source: Lrc<Vec<char>>,
tokens: Vec<Token>,
- parser: Box<dyn Parser>
+ parser: Box<dyn Parser>,
}
impl Default for Document {
@@ -30,14 +31,14 @@ impl Document {
pub fn new(text: &str, parser: Box<dyn Parser>) -> Self {
let source: Vec<_> = text.chars().collect();
- Self::new_from_vec(source, parser)
+ Self::new_from_vec(Lrc::new(source), parser)
}
- pub fn new_from_vec(source: Vec<char>, parser: Box<dyn Parser>) -> Self {
+ pub fn new_from_vec(source: Lrc<Vec<char>>, parser: Box<dyn Parser>) -> Self {
let mut doc = Self {
source,
tokens: Vec::new(),
- parser
+ parser,
};
doc.parse();
@@ -102,7 +103,7 @@ impl Document {
&old[indices
.last()
.map(|v| v + stretch_len)
- .unwrap_or(indices.len())..]
+ .unwrap_or(indices.len())..],
);
}
@@ -242,7 +243,7 @@ impl Document {
pub fn get_full_string(&self) -> String {
self.get_span_content_str(Span {
start: 0,
- end: self.source.len()
+ end: self.source.len(),
})
}
@@ -251,26 +252,28 @@ impl Document {
}
pub fn apply_suggestion(&mut self, suggestion: &Suggestion, span: Span) {
+ let source = Lrc::make_mut(&mut self.source);
+
match suggestion {
Suggestion::ReplaceWith(chars) => {
// Avoid allocation if possible
if chars.len() == span.len() {
for (index, c) in chars.iter().enumerate() {
- self.source[index + span.start] = *c
+ source[index + span.start] = *c
}
} else {
- let popped = self.source.split_off(span.start);
+ let popped = source.split_off(span.start);
- self.source.extend(chars);
- self.source.extend(popped.into_iter().skip(span.len()));
+ source.extend(chars);
+ source.extend(popped.into_iter().skip(span.len()));
}
}
Suggestion::Remove => {
- for i in span.end..self.source.len() {
- self.source[i - span.len()] = self.source[i];
+ for i in span.end..source.len() {
+ source[i - span.len()] = source[i];
}
- self.source.truncate(self.source.len() - span.len());
+ source.truncate(source.len() - span.len());
}
}
@@ -500,7 +503,7 @@ fn is_chunk_terminator(token: &TokenKind) -> bool {
match token {
TokenKind::Punctuation(punct) => [Punctuation::Comma].contains(punct),
- _ => false
+ _ => false,
}
}
@@ -509,11 +512,11 @@ fn is_sentence_terminator(token: &TokenKind) -> bool {
TokenKind::Punctuation(punct) => [
Punctuation::Period,
Punctuation::Bang,
- Punctuation::Question
+ Punctuation::Question,
]
.contains(punct),
TokenKind::Newline(count) => *count >= 2,
- _ => false
+ _ => false,
}
}
@@ -634,7 +637,7 @@ mod tests {
assert_token_count("This is the 3rd test", 9);
assert_token_count(
"It works even with weird capitalization like this: 600nD",
- 18
+ 18,
);
}
diff --git a/harper-core/src/lib.rs b/harper-core/src/lib.rs
index e646e8d73..61ed492c3 100644
--- a/harper-core/src/lib.rs
+++ b/harper-core/src/lib.rs
@@ -7,12 +7,14 @@ mod document;
mod language_detection;
mod lexing;
mod linting;
+mod sync;
pub mod parsers;
mod punctuation;
mod span;
mod spell;
mod token;
+pub use sync::Lrc;
pub use char_string::{CharString, CharStringExt};
pub use document::Document;
pub use linting::{Lint, LintGroup, LintGroupConfig, LintKind, Linter, Suggestion};
diff --git a/harper-core/src/linting/spell_check.rs b/harper-core/src/linting/spell_check.rs
index 4accfd0bf..f19d99bbc 100644
--- a/harper-core/src/linting/spell_check.rs
+++ b/harper-core/src/linting/spell_check.rs
@@ -1,31 +1,32 @@
use hashbrown::HashMap;
+use smallvec::ToSmallVec;
use super::lint::Suggestion;
use super::{Lint, LintKind, Linter};
use crate::document::Document;
use crate::spell::suggest_correct_spelling;
-use crate::{Dictionary, TokenStringExt};
+use crate::{CharString, Dictionary, TokenStringExt};
pub struct SpellCheck<T>
where
- T: Dictionary
+ T: Dictionary,
{
dictionary: T,
- word_cache: HashMap<Vec<char>, Vec<Vec<char>>>
+ word_cache: HashMap<CharString, Vec<CharString>>,
}
impl<T: Dictionary> SpellCheck<T> {
pub fn new(dictionary: T) -> Self {
Self {
dictionary,
- word_cache: HashMap::new()
+ word_cache: HashMap::new(),
}
}
}
impl<T: Dictionary> SpellCheck<T> {
- fn cached_suggest_correct_spelling(&mut self, word: &[char]) -> Vec<Vec<char>> {
- let word = word.to_vec();
+ fn cached_suggest_correct_spelling(&mut self, word: &[char]) -> Vec<CharString> {
+ let word = word.to_smallvec();
self.word_cache
.entry(word.clone())
@@ -37,7 +38,7 @@ impl<T: Dictionary> SpellCheck<T> {
while suggestions.is_empty() && dist < 5 {
suggestions = suggest_correct_spelling(&word, 100, dist, &self.dictionary)
.into_iter()
- .map(|v| v.to_vec())
+ .map(|v| v.to_smallvec())
.collect();
dist += 1;
@@ -86,7 +87,7 @@ impl<T: Dictionary> Linter for SpellCheck<T> {
"Did you mean to spell “{}” this way?",
document.get_span_content_str(word.span)
),
- priority: 63
+ priority: 63,
})
}
diff --git a/harper-core/src/sync.rs b/harper-core/src/sync.rs
new file mode 100644
index 000000000..30a24bc42
--- /dev/null
+++ b/harper-core/src/sync.rs
@@ -0,0 +1,5 @@
+#[cfg(feature = "concurrent")]
+pub use std::sync::Arc as Lrc;
+
+#[cfg(not(feature = "concurrent"))]
+pub use std::rc::Rc as Lrc;
diff --git a/harper-ls/Cargo.toml b/harper-ls/Cargo.toml
index c223d754b..b87d4f4d9 100644
--- a/harper-ls/Cargo.toml
+++ b/harper-ls/Cargo.toml
@@ -8,7 +8,7 @@ readme = "README.md"
repository = "https://github.com/elijah-potter/harper"
[dependencies]
-harper-core = { path = "../harper-core", version = "0.8.0" }
+harper-core = { path = "../harper-core", version = "0.8.0", features = ["concurrent"] }
harper-comments = { path = "../harper-comments", version = "0.8.0" }
tower-lsp = "0.20.0"
tokio = { version = "1.36.0", features = ["fs", "rt", "rt-multi-thread", "macros", "io-std", "io-util", "net"] }
diff --git a/harper-ls/src/backend.rs b/harper-ls/src/backend.rs
index bc16cf53d..cd57a2d1c 100644
--- a/harper-ls/src/backend.rs
+++ b/harper-ls/src/backend.rs
@@ -5,45 +5,20 @@ use std::sync::Arc;
use harper_comments::TreeSitterParser;
use harper_core::parsers::{Markdown, PlainEnglish};
use harper_core::{
- Dictionary,
- Document,
- FullDictionary,
- LintGroup,
- Linter,
- MergedDictionary,
- Token,
- TokenKind
+ Dictionary, Document, FullDictionary, LintGroup, Linter, MergedDictionary, Token, TokenKind,
};
use serde_json::Value;
use tokio::sync::{Mutex, RwLock};
use tower_lsp::jsonrpc::Result;
use tower_lsp::lsp_types::notification::PublishDiagnostics;
use tower_lsp::lsp_types::{
- CodeActionOrCommand,
- CodeActionParams,
- CodeActionProviderCapability,
- CodeActionResponse,
- Command,
- Diagnostic,
- DidChangeConfigurationParams,
- DidChangeTextDocumentParams,
- DidCloseTextDocumentParams,
- DidOpenTextDocumentParams,
- DidSaveTextDocumentParams,
- ExecuteCommandOptions,
- ExecuteCommandParams,
- InitializeParams,
- InitializeResult,
- InitializedParams,
- MessageType,
- PublishDiagnosticsParams,
- Range,
- ServerCapabilities,
- TextDocumentSyncCapability,
- TextDocumentSyncKind,
- TextDocumentSyncOptions,
- TextDocumentSyncSaveOptions,
- Url
+ CodeActionOrCommand, CodeActionParams, CodeActionProviderCapability, CodeActionResponse,
+ Command, Diagnostic, DidChangeConfigurationParams, DidChangeTextDocumentParams,
+ DidCloseTextDocumentParams, DidOpenTextDocumentParams, DidSaveTextDocumentParams,
+ ExecuteCommandOptions, ExecuteCommandParams, InitializeParams, InitializeResult,
+ InitializedParams, MessageType, PublishDiagnosticsParams, Range, ServerCapabilities,
+ TextDocumentSyncCapability, TextDocumentSyncKind, TextDocumentSyncOptions,
+ TextDocumentSyncSaveOptions, Url,
};
use tower_lsp::{Client, LanguageServer};
use tracing::{error, info, instrument};
@@ -59,7 +34,7 @@ struct DocumentState {
document: Document,
ident_dict: Arc<FullDictionary>,
linter: LintGroup<MergedDictionary<FullDictionary>>,
- language_id: Option<String>
+ language_id: Option<String>,
}
/// Deallocate
@@ -67,7 +42,7 @@ pub struct Backend {
client: Client,
static_dictionary: Arc<FullDictionary>,
config: RwLock<Config>,
- doc_state: Mutex<HashMap<Url, DocumentState>>
+ doc_state: Mutex<HashMap<Url, DocumentState>>,
}
impl Backend {
@@ -78,7 +53,7 @@ impl Backend {
client,
static_dictionary: dictionary.into(),
doc_state: Mutex::new(HashMap::new()),
- config: RwLock::new(config)
+ config: RwLock::new(config),
}
}
@@ -173,7 +148,7 @@ impl Backend {
#[instrument(skip(self))]
async fn generate_file_dictionary(
&self,
- url: &Url
+ url: &Url,
) -> anyhow::Result<MergedDictionary<FullDictionary>> {
let (global_dictionary, file_dictionary) = tokio::join!(
self.generate_global_dictionary(),
@@ -190,11 +165,11 @@ impl Backend {
async fn update_document_from_file(
&self,
url: &Url,
- language_id: Option<&str>
+ language_id: Option<&str>,
) -> anyhow::Result<()> {
let content = match tokio::fs::read_to_string(
url.to_file_path()
- .map_err(|_| anyhow::format_err!("Could not extract file path."))?
+ .map_err(|_| anyhow::format_err!("Could not extract file path."))?,
)
.await
{
@@ -213,7 +188,7 @@ impl Backend {
&self,
url: &Url,
text: &str,
- language_id: Option<&str>
+ language_id: Option<&str>,
) -> anyhow::Result<()> {
let mut doc_lock = self.doc_state.lock().await;
let config_lock = self.config.read().await;
@@ -225,7 +200,7 @@ impl Backend {
let mut doc_state = DocumentState {
linter: LintGroup::new(
config_lock.lint_config,
- self.generate_file_dictionary(url).await?
+ self.generate_file_dictionary(url).await?,
),
language_id: language_id
.map(|v| v.to_string())
@@ -241,6 +216,7 @@ impl Backend {
doc_state.document =
if let Some(ts_parser) = TreeSitterParser::new_from_language_id(language_id) {
let source: Vec<char> = text.chars().collect();
+ let source = Arc::new(source);
if let Some(new_dict) = ts_parser.create_ident_dict(source.as_slice()) {
let new_dict = Arc::new(new_dict);
@@ -275,7 +251,7 @@ impl Backend {
async fn generate_code_actions(
&self,
url: &Url,
- range: Range
+ range: Range,
) -> Result<Vec<CodeActionOrCommand>> {
let (config, mut doc_states) = tokio::join!(self.config.read(), self.doc_state.lock());
let Some(doc_state) = doc_states.get_mut(url) else {
@@ -307,7 +283,7 @@ impl Backend {
actions.push(CodeActionOrCommand::Command(Command::new(
"Open URL".to_string(),
"HarperOpen".to_string(),
- Some(vec![doc_state.document.get_span_content_str(span).into()])
+ Some(vec![doc_state.document.get_span_content_str(span).into()]),
)))
}
@@ -327,7 +303,7 @@ impl Backend {
lints_to_diagnostics(
doc_state.document.get_full_content(),
&lints,
- config.diagnostic_severity
+ config.diagnostic_severity,
)
}
@@ -338,7 +314,7 @@ impl Backend {
let result = PublishDiagnosticsParams {
uri: url.clone(),
diagnostics,
- version: None
+ version: None,
};
self.client
@@ -368,12 +344,12 @@ impl LanguageServer for Backend {
change: Some(TextDocumentSyncKind::FULL),
will_save: None,
will_save_wait_until: None,
- save: Some(TextDocumentSyncSaveOptions::Supported(true))
- }
+ save: Some(TextDocumentSyncSaveOptions::Supported(true)),
+ },
)),
..Default::default()
- }
+ },
})
}
@@ -404,7 +380,7 @@ impl LanguageServer for Backend {
.update_document(
&params.text_document.uri,
&params.text_document.text,
- Some(&params.text_document.language_id)
+ Some(&params.text_document.language_id),
)
.await;
@@ -493,7 +469,7 @@ impl LanguageServer for Backend {
error!("Unable to open URL: {}", err);
}
},
- _ => ()
+ _ => (),
}
Ok(None)
@@ -524,7 +500,7 @@ impl LanguageServer for Backend {
// Update documents with new config
futures::future::join_all(
keys.iter()
- .map(|key| self.update_document_from_file(key, None))
+ .map(|key| self.update_document_from_file(key, None)),
)
.await;
}
diff --git a/harper-wasm/Cargo.toml b/harper-wasm/Cargo.toml
index ae2f04133..eca966b51 100644
--- a/harper-wasm/Cargo.toml
+++ b/harper-wasm/Cargo.toml
@@ -14,5 +14,5 @@ console_error_panic_hook = "0.1.7"
tracing = "0.1.40"
tracing-wasm = "0.2.1"
wasm-bindgen = "0.2.91"
-harper-core = { path = "../harper-core" }
+harper-core = { path = "../harper-core", version = "0.8.0", features = ["concurrent"] }
once_cell = "1.19.0"
diff --git a/harper-wasm/src/lib.rs b/harper-wasm/src/lib.rs
index 68ea5f71c..3991c86eb 100644
--- a/harper-wasm/src/lib.rs
+++ b/harper-wasm/src/lib.rs
@@ -1,17 +1,16 @@
#![doc = include_str!("../README.md")]
-use std::rc::Rc;
use std::sync::Mutex;
use harper_core::parsers::Markdown;
-use harper_core::{remove_overlaps, Document, FullDictionary, LintGroup, Linter};
+use harper_core::{remove_overlaps, Document, FullDictionary, LintGroup, Linter, Lrc};
use once_cell::sync::Lazy;
use wasm_bindgen::prelude::wasm_bindgen;
static LINTER: Lazy<Mutex<LintGroup<FullDictionary>>> = Lazy::new(|| {
Mutex::new(LintGroup::new(
Default::default(),
- FullDictionary::create_from_curated()
+ FullDictionary::create_from_curated(),
))
});
@@ -38,6 +37,7 @@ pub fn use_spell_check(set: bool) {
#[wasm_bindgen]
pub fn lint(text: String) -> Vec<Lint> {
let source: Vec<_> = text.chars().collect();
+ let source = Lrc::new(source);
let document = Document::new_from_vec(source.clone(), Box::new(Markdown));
@@ -45,11 +45,9 @@ pub fn lint(text: String) -> Vec<Lint> {
remove_overlaps(&mut lints);
- let text = Rc::new(source);
-
lints
.into_iter()
- .map(|l| Lint::new(l, text.clone()))
+ .map(|l| Lint::new(l, source.clone()))
.collect()
}
@@ -57,7 +55,7 @@ pub fn lint(text: String) -> Vec<Lint> {
pub fn apply_suggestion(
text: String,
span: Span,
- suggestion: &Suggestion
+ suggestion: &Suggestion,
) -> Result<String, String> {
let mut document = Document::new_markdown(&text);
document.apply_suggestion(&suggestion.inner, span.into());
@@ -67,13 +65,13 @@ pub fn apply_suggestion(
#[wasm_bindgen]
pub struct Suggestion {
- inner: harper_core::Suggestion
+ inner: harper_core::Suggestion,
}
#[wasm_bindgen]
pub enum SuggestionKind {
Replace,
- Remove
+ Remove,
}
#[wasm_bindgen]
@@ -88,14 +86,14 @@ impl Suggestion {
pub fn get_replacement_text(&self) -> String {
match &self.inner {
harper_core::Suggestion::Remove => "".to_string(),
- harper_core::Suggestion::ReplaceWith(chars) => chars.iter().collect()
+ harper_core::Suggestion::ReplaceWith(chars) => chars.iter().collect(),
}
}
pub fn kind(&self) -> SuggestionKind {
match &self.inner {
harper_core::Suggestion::Remove => SuggestionKind::Remove,
- harper_core::Suggestion::ReplaceWith(_) => SuggestionKind::Replace
+ harper_core::Suggestion::ReplaceWith(_) => SuggestionKind::Replace,
}
}
}
@@ -103,12 +101,12 @@ impl Suggestion {
#[wasm_bindgen]
pub struct Lint {
inner: harper_core::Lint,
- source: Rc<Vec<char>>
+ source: Lrc<Vec<char>>,
}
#[wasm_bindgen]
impl Lint {
- pub(crate) fn new(inner: harper_core::Lint, source: Rc<Vec<char>>) -> Self {
+ pub(crate) fn new(inner: harper_core::Lint, source: Lrc<Vec<char>>) -> Self {
Self { inner, source }
}
@@ -146,7 +144,7 @@ impl Lint {
#[wasm_bindgen]
pub struct Span {
pub start: usize,
- pub end: usize
+ pub end: usize,
}
#[wasm_bindgen]