Switch to finl_unicode for grapheme clustering

According to its benchmarks, it's almost 2x faster than
unicode_segmentation.  It doesn't appear to make a visible
difference to `time cat bigfile`, but I'll take anything
that gives more headroom for such little effort of switching.

Cargo.lock

@@ -1392,6 +1392,12 @@ dependencies = [
  "windows-sys",
 ]
 
+[[package]]
+name = "finl_unicode"
+version = "1.0.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "69152938fc3cf544c50065ed78d321d0c5bf3433344f227eb2323bcf78370d34"
+
 [[package]]
 name = "fixedbitset"
 version = "0.4.2"
@@ -2628,6 +2634,7 @@ dependencies = [
  "crossbeam",
  "downcast-rs",
  "filedescriptor",
+ "finl_unicode",
  "flume",
  "hostname",
  "k9",
@@ -2654,7 +2661,6 @@ dependencies = [
  "termwiz-funcs",
  "textwrap 0.15.0",
  "thiserror",
- "unicode-segmentation",
  "url",
  "wezterm-dynamic",
  "wezterm-ssh",
@@ -4650,6 +4656,7 @@ dependencies = [
  "criterion",
  "env_logger",
  "filedescriptor",
+ "finl_unicode",
  "fixedbitset",
  "fnv",
  "hex",
@@ -4691,11 +4698,11 @@ version = "0.1.0"
 dependencies = [
  "anyhow",
  "config",
+ "finl_unicode",
  "lazy_static",
  "luahelper",
  "terminfo",
  "termwiz",
- "unicode-segmentation",
  "wezterm-dynamic",
 ]
 
@@ -5488,6 +5495,7 @@ dependencies = [
  "enum-display-derive",
  "env_logger",
  "euclid",
+ "finl_unicode",
  "fontconfig",
  "freetype",
  "harfbuzz",
@@ -5500,7 +5508,6 @@ dependencies = [
  "rangeset",
  "termwiz",
  "thiserror",
- "unicode-segmentation",
  "walkdir",
  "wezterm-bidi",
  "wezterm-color-types",
@@ -5533,6 +5540,7 @@ dependencies = [
  "euclid",
  "fastrand",
  "filedescriptor",
+ "finl_unicode",
  "fnv",
  "frecency",
  "futures",
@@ -5701,6 +5709,7 @@ dependencies = [
  "bitflags",
  "csscolorparser",
  "env_logger",
+ "finl_unicode",
  "hex",
  "image",
  "k9",
@@ -5714,7 +5723,6 @@ dependencies = [
  "terminfo",
  "termwiz",
  "unicode-normalization",
- "unicode-segmentation",
  "url",
  "wezterm-bidi",
  "wezterm-dynamic",

lua-api-crates/termwiz-funcs/Cargo.toml

@@ -8,9 +8,9 @@ edition = "2021"
 [dependencies]
 anyhow = "1.0"
 config = { path = "../../config" }
+finl_unicode = "1.0.2"
 terminfo = "0.7"
 wezterm-dynamic = { path = "../../wezterm-dynamic" }
 luahelper = { path = "../../luahelper" }
 termwiz = { path = "../../termwiz", features=["use_serde"] }
-unicode-segmentation = "1.8"
 lazy_static = "1.4"

lua-api-crates/termwiz-funcs/src/lib.rs

@@ -1,5 +1,6 @@
 use config::lua::get_or_create_module;
 use config::lua::mlua::{self, Lua, ToLua};
+use finl_unicode::grapheme_clusters::Graphemes;
 use luahelper::impl_lua_conversion_dynamic;
 use std::str::FromStr;
 use termwiz::caps::{Capabilities, ColorLevel, ProbeHints};
@@ -8,7 +9,6 @@ use termwiz::color::{AnsiColor, ColorAttribute, ColorSpec, SrgbaTuple};
 use termwiz::input::Modifiers;
 use termwiz::render::terminfo::TerminfoRenderer;
 use termwiz::surface::change::Change;
-use unicode_segmentation::UnicodeSegmentation;
 use wezterm_dynamic::{FromDynamic, ToDynamic};
 
 pub fn register(lua: &Lua) -> anyhow::Result<()> {
@@ -170,7 +170,8 @@ pub fn pad_left(mut result: String, width: usize) -> String {
 pub fn truncate_left(s: &str, max_width: usize) -> String {
     let mut result = vec![];
     let mut len = 0;
-    for g in s.graphemes(true).rev() {
+    let graphemes: Vec<_> = Graphemes::new(s).collect();
+    for &g in graphemes.iter().rev() {
         let g_len = grapheme_column_width(g, None);
         if g_len + len > max_width {
             break;
@@ -186,7 +187,7 @@ pub fn truncate_left(s: &str, max_width: usize) -> String {
 pub fn truncate_right(s: &str, max_width: usize) -> String {
     let mut result = String::new();
     let mut len = 0;
-    for g in s.graphemes(true) {
+    for g in Graphemes::new(s) {
         let g_len = grapheme_column_width(g, None);
         if g_len + len > max_width {
             break;

mux/Cargo.toml

@@ -17,6 +17,7 @@ config = { path = "../config" }
 crossbeam = "0.8"
 downcast-rs = "1.0"
 filedescriptor = { version="0.8", path = "../filedescriptor" }
+finl_unicode = "1.0.2"
 hostname = "0.3"
 lazy_static = "1.4"
 libc = "0.2"
@@ -40,7 +41,6 @@ termwiz = { path = "../termwiz" }
 termwiz-funcs = { path = "../lua-api-crates/termwiz-funcs" }
 textwrap = "0.15"
 thiserror = "1.0"
-unicode-segmentation = "1.8"
 url = "2"
 wezterm-ssh = { path = "../wezterm-ssh" }
 wezterm-dynamic = { path = "../wezterm-dynamic" }

mux/src/connui.rs

@@ -1,6 +1,7 @@
 use crate::termwiztermtab;
 use anyhow::{anyhow, bail, Context as _};
 use crossbeam::channel::{unbounded, Receiver, Sender};
+use finl_unicode::grapheme_clusters::Graphemes;
 use promise::spawn::block_on;
 use promise::Promise;
 use std::sync::Mutex;
@@ -9,7 +10,6 @@ use termwiz::cell::{unicode_column_width, CellAttributes};
 use termwiz::lineedit::*;
 use termwiz::surface::{Change, Position};
 use termwiz::terminal::*;
-use unicode_segmentation::UnicodeSegmentation;
 use wezterm_term::TerminalSize;
 
 #[derive(Default)]
@@ -146,7 +146,7 @@ impl ConnectionUIImpl {
             let mut reversed_string = String::new();
             let mut default_string = String::new();
             let mut col = 0;
-            for grapheme in message.graphemes(true) {
+            for grapheme in Graphemes::new(&message) {
                 // Once we've passed the elapsed column, full up the string
                 // that we'll render with default attributes instead.
                 if col > prog_width {

term/Cargo.toml

@@ -18,6 +18,7 @@ anyhow = "1.0"
 bitflags = "1.3"
 csscolorparser = "0.6"
 miniz_oxide = "0.4"
+finl_unicode = "1.0.2"
 hex = "0.4"
 image = "0.24"
 lazy_static = "1.4"
@@ -28,7 +29,6 @@ ordered-float = "3.0"
 serde = {version="1.0", features = ["rc"]}
 terminfo = "0.7"
 unicode-normalization = "0.1.21"
-unicode-segmentation = "1.8"
 url = "2"
 wezterm-bidi = { path = "../bidi" }
 wezterm-dynamic = { path = "../wezterm-dynamic" }
@@ -36,7 +36,6 @@ wezterm-dynamic = { path = "../wezterm-dynamic" }
 [dev-dependencies]
 env_logger = "0.9"
 k9 = "0.11.0"
-unicode-normalization = "0.1"
 
 [dependencies.termwiz]
 version = "0.17"

term/src/terminalstate/performer.rs

@@ -3,6 +3,7 @@ use crate::terminalstate::{
     default_color_map, CharSet, MouseEncoding, TabStop, UnicodeVersionStackEntry,
 };
 use crate::{ClipboardSelection, Position, TerminalState, VisibleRowIndex, DCS, ST};
+use finl_unicode::grapheme_clusters::Graphemes;
 use log::{debug, error};
 use num_traits::FromPrimitive;
 use ordered_float::NotNan;
@@ -126,7 +127,7 @@ impl<'a> Performer<'a> {
             p.as_str()
         };
 
-        for g in unicode_segmentation::UnicodeSegmentation::graphemes(text, true) {
+        for g in Graphemes::new(text) {
             let g = self.remap_grapheme(g);
 
             let print_width = grapheme_column_width(g, Some(self.unicode_version));

term/src/test/mod.rs

@@ -1177,8 +1177,8 @@ fn test_1573() {
     let recomposed: String = sequence.nfc().collect();
     assert_eq!(recomposed, "\u{d55c}");
 
-    use unicode_segmentation::UnicodeSegmentation;
-    let graphemes: Vec<_> = sequence.graphemes(true).collect();
+    use finl_unicode::grapheme_clusters::Graphemes;
+    let graphemes: Vec<_> = Graphemes::new(sequence).collect();
     assert_eq!(graphemes, vec![sequence]);
 }
 

termwiz/Cargo.toml

@@ -18,6 +18,7 @@ cassowary = {version="0.3", optional=true}
 cfg-if = "1.0"
 anyhow = "1.0"
 filedescriptor = { version="0.8", path = "../filedescriptor" }
+finl_unicode = "1.0.2"
 fixedbitset = "0.4"
 fnv = {version="1.0", optional=true}
 hex = "0.4"

termwiz/src/cell.rs

@@ -5,6 +5,7 @@ use crate::emoji_variation::WCWIDTH_TABLE;
 pub use crate::escape::osc::Hyperlink;
 use crate::image::ImageCell;
 use crate::widechar_width::WcWidth;
+use finl_unicode::grapheme_clusters::Graphemes;
 #[cfg(feature = "use_serde")]
 use serde::{Deserialize, Deserializer, Serialize, Serializer};
 use std::hash::{Hash, Hasher};
@@ -922,8 +923,7 @@ pub const LATEST_UNICODE_VERSION: UnicodeVersion = UnicodeVersion {
 /// Calls through to `grapheme_column_width` for each grapheme
 /// and sums up the length.
 pub fn unicode_column_width(s: &str, version: Option<UnicodeVersion>) -> usize {
-    use unicode_segmentation::UnicodeSegmentation;
-    s.graphemes(true)
+    Graphemes::new(s)
         .map(|g| grapheme_column_width(g, version))
         .sum()
 }
@@ -1023,7 +1023,6 @@ pub enum AttributeChange {
 #[cfg(test)]
 mod test {
     use super::*;
-    use unicode_segmentation::UnicodeSegmentation;
 
     #[test]
     fn teeny_string() {
@@ -1135,7 +1134,7 @@ mod test {
         let x_ideographic_space_x = "x\u{3000}x";
         assert_eq!(unicode_column_width(x_ideographic_space_x, None), 4);
         assert_eq!(
-            x_ideographic_space_x.graphemes(true).collect::<Vec<_>>(),
+            Graphemes::new(x_ideographic_space_x).collect::<Vec<_>>(),
             vec!["x".to_string(), "\u{3000}".to_string(), "x".to_string()],
         );
 
@@ -1155,21 +1154,17 @@ mod test {
         assert_eq!(unicode_column_width(victory_hand, None), 1);
 
         assert_eq!(
-            victory_hand_text_presentation
-                .graphemes(true)
-                .collect::<Vec<_>>(),
+            Graphemes::new(victory_hand_text_presentation).collect::<Vec<_>>(),
             vec![victory_hand_text_presentation.to_string()]
         );
         assert_eq!(
-            victory_hand.graphemes(true).collect::<Vec<_>>(),
+            Graphemes::new(victory_hand).collect::<Vec<_>>(),
             vec![victory_hand.to_string()]
         );
 
         let copyright_emoji_presentation = "\u{00A9}\u{FE0F}";
         assert_eq!(
-            copyright_emoji_presentation
-                .graphemes(true)
-                .collect::<Vec<_>>(),
+            Graphemes::new(copyright_emoji_presentation).collect::<Vec<_>>(),
             vec![copyright_emoji_presentation.to_string()]
         );
         assert_eq!(unicode_column_width(copyright_emoji_presentation, None), 2);
@@ -1180,9 +1175,7 @@ mod test {
 
         let copyright_text_presentation = "\u{00A9}";
         assert_eq!(
-            copyright_text_presentation
-                .graphemes(true)
-                .collect::<Vec<_>>(),
+            Graphemes::new(copyright_text_presentation).collect::<Vec<_>>(),
             vec![copyright_text_presentation.to_string()]
         );
         assert_eq!(unicode_column_width(copyright_text_presentation, None), 1);
@@ -1202,11 +1195,11 @@ mod test {
         assert_eq!(unicode_column_width(raised_fist_text, None), 2);
 
         assert_eq!(
-            raised_fist_text.graphemes(true).collect::<Vec<_>>(),
+            Graphemes::new(raised_fist_text).collect::<Vec<_>>(),
             vec![raised_fist_text.to_string()]
         );
         assert_eq!(
-            raised_fist.graphemes(true).collect::<Vec<_>>(),
+            Graphemes::new(raised_fist).collect::<Vec<_>>(),
             vec![raised_fist.to_string()]
         );
     }

termwiz/src/surface/change.rs

@@ -2,10 +2,10 @@ use crate::cell::{unicode_column_width, AttributeChange, CellAttributes};
 use crate::color::ColorAttribute;
 pub use crate::image::{ImageData, TextureCoordinate};
 use crate::surface::{CursorShape, CursorVisibility, Position};
+use finl_unicode::grapheme_clusters::Graphemes;
 #[cfg(feature = "use_serde")]
 use serde::{Deserialize, Serialize};
 use std::sync::Arc;
-use unicode_segmentation::UnicodeSegmentation;
 
 /// `Change` describes an update operation to be applied to a `Surface`.
 /// Changes to the active attributes (color, style), moving the cursor
@@ -190,7 +190,7 @@ impl ChangeSequence {
             | Change::Title(_)
             | Change::ClearToEndOfScreen(_) => {}
             Change::Text(t) => {
-                for g in t.as_str().graphemes(true) {
+                for g in Graphemes::new(t.as_str()) {
                     if self.cursor_x == self.screen_cols {
                         self.cursor_y += 1;
                         self.cursor_x = 0;

termwiz/src/surface/line/clusterline.rs

@@ -1,11 +1,11 @@
 use crate::cell::{Cell, CellAttributes};
 use crate::surface::line::CellRef;
+use finl_unicode::grapheme_clusters::Graphemes;
 use fixedbitset::FixedBitSet;
 #[cfg(feature = "use_serde")]
 use serde::{Deserialize, Deserializer, Serialize, Serializer};
 use std::convert::TryInto;
 use std::num::NonZeroU8;
-use unicode_segmentation::UnicodeSegmentation;
 
 #[cfg_attr(feature = "use_serde", derive(Serialize, Deserialize))]
 #[derive(Debug, Clone, PartialEq)]
@@ -166,7 +166,7 @@ impl ClusteredLine {
         let mut clusters = self.clusters.iter();
         let cluster = clusters.next();
         ClusterLineCellIter {
-            graphemes: self.text.graphemes(true),
+            graphemes: Graphemes::new(&self.text),
             clusters,
             cluster,
             idx: 0,
@@ -307,7 +307,7 @@ impl ClusteredLine {
 }
 
 pub(crate) struct ClusterLineCellIter<'a> {
-    graphemes: unicode_segmentation::Graphemes<'a>,
+    graphemes: Graphemes<'a>,
     clusters: std::slice::Iter<'a, Cluster>,
     cluster: Option<&'a Cluster>,
     idx: usize,

termwiz/src/surface/line/line.rs

@@ -7,6 +7,7 @@ use crate::surface::line::linebits::LineBits;
 use crate::surface::line::storage::{CellStorage, VisibleCellIter};
 use crate::surface::line::vecstorage::{VecStorage, VecStorageIter};
 use crate::surface::{Change, SequenceNo, SEQ_ZERO};
+use finl_unicode::grapheme_clusters::Graphemes;
 #[cfg(feature = "use_serde")]
 use serde::{Deserialize, Serialize};
 use siphasher::sip128::{Hasher128, SipHasher};
@@ -15,7 +16,6 @@ use std::borrow::Cow;
 use std::hash::Hash;
 use std::ops::Range;
 use std::sync::{Arc, Mutex, Weak};
-use unicode_segmentation::UnicodeSegmentation;
 use wezterm_bidi::{Direction, ParagraphDirectionHint};
 
 #[cfg_attr(feature = "use_serde", derive(Serialize, Deserialize))]
@@ -137,7 +137,7 @@ impl Line {
     ) -> Line {
         let mut cells = Vec::new();
 
-        for sub in s.graphemes(true) {
+        for sub in Graphemes::new(s) {
             let cell = Cell::new_grapheme(sub, attrs.clone(), unicode_version);
             let width = cell.width();
             cells.push(cell);
@@ -850,7 +850,7 @@ impl Line {
         attr: CellAttributes,
         seqno: SequenceNo,
     ) {
-        for (i, c) in text.graphemes(true).enumerate() {
+        for (i, c) in Graphemes::new(text).enumerate() {
             let cell = Cell::new_grapheme(c, attr.clone(), None);
             let width = cell.width();
             self.set_cell(i + start_idx, cell, seqno);

termwiz/src/surface/mod.rs

@@ -2,12 +2,12 @@ use crate::cell::{AttributeChange, Cell, CellAttributes};
 use crate::color::ColorAttribute;
 use crate::image::ImageCell;
 use crate::surface::line::CellRef;
+use finl_unicode::grapheme_clusters::Graphemes;
 use ordered_float::NotNan;
 #[cfg(feature = "use_serde")]
 use serde::{Deserialize, Serialize};
 use std::borrow::Cow;
 use std::cmp::min;
-use unicode_segmentation::UnicodeSegmentation;
 use wezterm_dynamic::{FromDynamic, ToDynamic};
 
 pub mod change;
@@ -407,7 +407,7 @@ impl Surface {
     }
 
     fn print_text(&mut self, text: &str) {
-        for g in UnicodeSegmentation::graphemes(text, true) {
+        for g in Graphemes::new(text) {
             if g == "\r\n" {
                 self.xpos = 0;
                 let new_y = self.ypos + 1;

wezterm-font/Cargo.toml

@@ -19,6 +19,7 @@ config = { path = "../config" }
 encoding_rs = "0.8"
 enum-display-derive = "0.1"
 euclid = "0.22"
+finl_unicode = "1.0.2"
 freetype = { path = "../deps/freetype" }
 harfbuzz = { path = "../deps/harfbuzz" }
 lazy_static = "1.4"
@@ -29,7 +30,6 @@ ordered-float = "3.0"
 rangeset = { path = "../rangeset" }
 termwiz = { path = "../termwiz" }
 thiserror = "1.0"
-unicode-segmentation = "1.8"
 walkdir = "2"
 wezterm-color-types = { path = "../color-types" }
 wezterm-input-types = { path = "../wezterm-input-types" }

wezterm-font/src/shaper/harfbuzz.rs

@@ -4,13 +4,13 @@ use crate::units::*;
 use crate::{ftwrap, hbwrap as harfbuzz};
 use anyhow::{anyhow, Context};
 use config::ConfigHandle;
+use finl_unicode::grapheme_clusters::Graphemes;
 use log::error;
 use ordered_float::NotNan;
 use std::cell::{RefCell, RefMut};
 use std::collections::HashMap;
 use std::ops::Range;
 use termwiz::cell::{unicode_column_width, Presentation};
-use unicode_segmentation::UnicodeSegmentation;
 use wezterm_bidi::Direction;
 
 // Changing these will switch to using harfbuzz's opentype functions.
@@ -80,7 +80,7 @@ pub struct HarfbuzzShaper {
 /// original string.  That isn't perfect, but it should
 /// be good enough to indicate that something isn't right.
 fn make_question_string(s: &str) -> String {
-    let len = s.graphemes(true).count();
+    let len = Graphemes::new(s).count();
     let mut result = String::new();
     let c = if !is_question_string(s) {
         std::char::REPLACEMENT_CHARACTER

wezterm-gui/Cargo.toml

@@ -48,6 +48,7 @@ env-bootstrap = { path = "../env-bootstrap" }
 euclid = "0.22"
 fastrand = "1.6"
 filedescriptor = { version="0.8", path = "../filedescriptor" }
+finl_unicode = "1.0.2"
 fnv = "1.0"
 frecency = { path = "../frecency" }
 futures = "0.3"

wezterm-gui/src/tabbar.rs

@@ -1,5 +1,6 @@
 use crate::termwindow::{PaneInformation, TabInformation, UIItem, UIItemType};
 use config::{ConfigHandle, TabBarColors};
+use finl_unicode::grapheme_clusters::Graphemes;
 use mlua::FromLua;
 use termwiz::cell::{unicode_column_width, Cell, CellAttributes};
 use termwiz::color::ColorSpec;
@@ -401,7 +402,7 @@ pub fn parse_status_text(text: &str, default_cell: CellAttributes) -> Line {
     let mut print_buffer = String::new();
 
     fn flush_print(buf: &mut String, cells: &mut Vec<Cell>, pen: &CellAttributes) {
-        for g in unicode_segmentation::UnicodeSegmentation::graphemes(buf.as_str(), true) {
+        for g in Graphemes::new(buf.as_str()) {
             let cell = Cell::new_grapheme(g, pen.clone(), None);
             let width = cell.width();
             cells.push(cell);

wezterm-gui/src/termwindow/box_model.rs

@@ -11,11 +11,11 @@ use crate::utilsprites::RenderMetrics;
 use ::window::{RectF, WindowOps};
 use anyhow::anyhow;
 use config::{Dimension, DimensionContext};
+use finl_unicode::grapheme_clusters::Graphemes;
 use std::cell::RefCell;
 use std::rc::Rc;
 use termwiz::cell::{grapheme_column_width, Presentation};
 use termwiz::surface::Line;
-use unicode_segmentation::UnicodeSegmentation;
 use wezterm_font::units::PixelUnit;
 use wezterm_font::LoadedFont;
 use wezterm_term::color::{ColorAttribute, ColorPalette};
@@ -583,7 +583,7 @@ impl super::TermWindow {
 
                 for info in infos {
                     let cell_start = &s[info.cluster as usize..];
-                    let mut iter = cell_start.graphemes(true).peekable();
+                    let mut iter = Graphemes::new(cell_start).peekable();
                     let grapheme = iter
                         .next()
                         .ok_or_else(|| anyhow!("info.cluster didn't map into string"))?;