release: v4.0.0

LoricAndre · LoricAndre · commit 85c6964da5fb · 2026-03-10T16:25:28.000+01:00
diff --git a/CHANGELOG.md b/CHANGELOG.md
@@ -5,6 +5,79 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [4.0.0] - 2026-03-10
+
+### Changes
+
+`sk` is now beating `fzf` in interactive matching for time, peak memory usage and CPU usage ! See the benchmarks below for details, including a brief explanation of the benchmark itself.
+
+This release brings multiple breaking changes, please read the following if you have doubts about the update.
+
+#### Default to the Arinae matcher
+
+The biggest change of them all is that the default algorithm is now `Arinae`, skim's latest and most performant algorithm, featuring typo-resistance. If you want to keep using `SkimV2`, pass `--algo skim_v2` and please take the time to open an issue explaining why if possible.
+This also makes the `--scheme` option available to the default matcher, allowing for more refined matching scenarii.
+
+#### Default to non-typo resistant
+
+Typo-resistant behavior is now disable by default, add `--typos` to your skim invocation or `SKIM_DEFAULT_OPTIONS` to get it back. Unless you were using the `frizbee`, `fzy` or `arinae` matcher, this should not impact you.
+
+#### Removal of the `skim_v1` algorithm
+
+The `SkimV1` algorithm, skim's initial algorithm, has been deprecated for years now, and it has now been removed.
+
+#### SkimItem index (library only)
+
+`SkimItem::get_index` and `SkimItem::set_index` are gone, and all index handling is now done internally.
+
+### Benchmarks
+
+This benchmarks runs the interactive interface in a tmux session, and waits for the UI to stabilize.
+It uses a 10 million path-like ASCII items input file, and the query `test`.
+
+```
+=== Results: sk v4.0.0 [baseline] ===
+Completed runs: 50 / 50
+Average items matched: 2895782 / 10000000  (min: 2895782, max: 2895782)
+Average time: 3.827s  (min: 3.576s, max: 4.090s)
+Average items/second: 2615767  (min: 2445033, max: 2796365)
+Average peak memory usage: 1589.2 MB  (min: 1518.6 MB, max: 1661.2 MB)
+Average peak CPU usage: 528.9%  (min: 457.0%, max: 740.0%)
+
+=== Results: sk v3.7.0 ===
+Completed runs: 50 / 50
+Average items matched: 2895782 / 10000000  (min: 2895782, max: 2895782) +0.0%
+Average time: 3.930s  (min: 3.565s, max: 4.226s)  +2.7%
+Average items/second: 2548674  (min: 2366263, max: 2804816)  -2.6%
+Average peak memory usage: 1618.8 MB  (min: 1539.1 MB, max: 1680.6 MB) +1.9%
+Average peak CPU usage: 696.8%  (min: 608.0%, max: 875.0%)  +31.7%
+
+=== Results: fzf 0.70.0 ===
+Completed runs: 50 / 50
+Average items matched: 2895782 / 10000000  (min: 2895782, max: 2895782) +0.0%
+Average time: 5.421s  (min: 4.814s, max: 6.111s)  +41.7%
+Average items/second: 1848269  (min: 1636444, max: 2077385)  -29.3%
+Average peak memory usage: 2015.3 MB  (min: 1860.7 MB, max: 2173.9 MB) +26.8%
+Average peak CPU usage: 1301.1%  (min: 1229.0%, max: 1431.0%)  +146.0%
+
+=== Comparison Summary (vs baseline: sk v4.0.0) ===
+Binary      Avg time     Δ time  Avg rate     Δ rate
+------------------------------------------------------------------------------
+sk v4.0.0   3.827s    baseline   2615767     baseline
+sk v3.7.0   3.930s      +2.7%    2548674      -2.6%
+fzf 0.70.0  5.421s     +41.7%    1848269     -29.3%
+```
+
+### 🚀 Features
+
+- [**breaking**] Internally compute indexes at match time (removes get/set_index) (#1001)
+- [**breaking**] Use Arinae as default algorithm
+
+### ⚙️ Miscellaneous Tasks
+
+- [**breaking**] Default to disabled typos
+- Use python for bench script for comparison
+
 ## [3.7.0] - 2026-03-08
 
 This adds a new library API: `Skim::run_items`. Using this, you don't need to send the items, the library handles it for you.
@@ -40,8 +113,6 @@ let res = Skim::run_with(options, ["foo", "bar"]);
 
 It will automatically convert any iterator of <impl SkimItem> by adding an `index` field and then send it, before running skim.
 
-
-
 ### 🚀 Features
 
 - Add `scheme` for better filepath matching in Arinae
diff --git a/Cargo.lock b/Cargo.lock
diff --git a/Cargo.toml b/Cargo.toml
@@ -1,6 +1,6 @@
 [package]
 name = "skim"
-version = "3.7.0"
+version = "4.0.0"
 authors = ["Loric ANDRE", "Zhang Jinzhou <lotabout@gmail.com>"]
 description = "Fuzzy Finder in rust!"
 documentation = "https://docs.rs/skim"
diff --git a/man/man1/sk.1 b/man/man1/sk.1
@@ -1,6 +1,6 @@
 .ie \n(.g .ds Aq \(aq
 .el .ds Aq '
-.TH sk 1  "sk 3.7.0" 
+.TH sk 1  "sk 4.0.0" 
 .ie \n(.g .ds Aq \(aq
 .el .ds Aq '
 .SH NAME
@@ -79,11 +79,11 @@ Run in exact mode
 \fB\-\-regex\fR
 Start in regex mode instead of fuzzy\-match
 .TP
-\fB\-\-algo\fR \fI<ALGORITHM>\fR [default: skim_v2]
+\fB\-\-algo\fR \fI<ALGORITHM>\fR [default: arinae]
 Fuzzy matching algorithm
 
-skim_v2 Latest skim algorithm, should be better in almost any case
-skim_v1 Legacy skim algorithm
+arinae (ari) Latest algorithm
+skim_v2 Legacy skim algorithm
 clangd  Used in clangd for keyword completion
 fzy     Algorithm from fzy (https://github.com/jhawthorn/fzy)
 .br
@@ -92,17 +92,15 @@ fzy     Algorithm from fzy (https://github.com/jhawthorn/fzy)
 \fIPossible values:\fR
 .RS 14
 .IP \(bu 2
-skim_v1: Original skim fuzzy matching algorithm (v1)
-.IP \(bu 2
-skim_v2: Improved skim fuzzy matching algorithm (v2, default)
+skim_v2: Improved skim fuzzy matching algorithm (v2)
 .IP \(bu 2
 clangd: Clangd fuzzy matching algorithm
 .IP \(bu 2
 fzy: Fzy matching algorithm (https://github.com/jhawthorn/fzy)
 .IP \(bu 2
 frizbee: Frizbee matching algorithm, typo resistant
 .IP \(bu 2
-arinae: Arinae: typo\-resistant & natural algorithm
+arinae: Arinae: typo\-resistant & natural algorithm, default
 .RE
 .TP
 \fB\-\-case\fR \fI<CASE>\fR [default: smart]
@@ -122,7 +120,7 @@ ignore: Case\-insensitive matching
 smart: Smart case: case\-insensitive unless query contains uppercase
 .RE
 .TP
-\fB\-\-typos\fR [\fI<TYPOS>\fR] [default: smart]
+\fB\-\-typos\fR [\fI<TYPOS>\fR] [default: disabled]
 Enable typo\-tolerant matching
 
 When passed without a value (\-\-typos), uses adaptive formula (pattern_length / 4). When passed with a value (e.g. \-\-typos=2), uses that exact number as the maximum allowed typos. \-\-typos=0 explicitly disables typo tolerance. Applies to both fzy and frizbee matchers.
@@ -1062,4 +1060,4 @@ When using `sk \-\-remote`, pipe in action chains (see the KEYBINDS section), fo
 .ie \n(.g .ds Aq \(aq
 .el .ds Aq '
 .SH VERSION
-v3.7.0
+v4.0.0
diff --git a/shell/completion.bash b/shell/completion.bash
@@ -62,7 +62,7 @@ _sk() {
                     return 0
                     ;;
                 --algo)
-                    COMPREPLY=($(compgen -W "skim_v1 skim_v2 clangd fzy frizbee arinae" -- "${cur}"))
+                    COMPREPLY=($(compgen -W "skim_v2 clangd fzy frizbee arinae" -- "${cur}"))
                     return 0
                     ;;
                 --case)
diff --git a/shell/completion.fish b/shell/completion.fish
@@ -14,12 +14,11 @@ pathname\t''
 complete -c sk -s n -l nth -d 'Fields to be matched' -r
 complete -c sk -l with-nth -d 'Fields to be transformed' -r
 complete -c sk -s d -l delimiter -d 'Delimiter between fields' -r
-complete -c sk -l algo -d 'Fuzzy matching algorithm' -r -f -a "skim_v1\t'Original skim fuzzy matching algorithm (v1)'
-skim_v2\t'Improved skim fuzzy matching algorithm (v2, default)'
+complete -c sk -l algo -d 'Fuzzy matching algorithm' -r -f -a "skim_v2\t'Improved skim fuzzy matching algorithm (v2)'
 clangd\t'Clangd fuzzy matching algorithm'
 fzy\t'Fzy matching algorithm (https://github.com/jhawthorn/fzy)'
 frizbee\t'Frizbee matching algorithm, typo resistant'
-arinae\t'Arinae: typo-resistant & natural algorithm'"
+arinae\t'Arinae: typo-resistant & natural algorithm, default'"
 complete -c sk -l case -d 'Case sensitivity' -r -f -a "respect\t'Case-sensitive matching'
 ignore\t'Case-insensitive matching'
 smart\t'Smart case: case-insensitive unless query contains uppercase'"
diff --git a/shell/completion.nu b/shell/completion.nu
@@ -5,7 +5,7 @@ module completions {
   }
 
   def "nu-complete sk algorithm" [] {
-    [ "skim_v1" "skim_v2" "clangd" "fzy" "frizbee" "arinae" ]
+    [ "skim_v2" "clangd" "fzy" "frizbee" "arinae" ]
   }
 
   def "nu-complete sk case" [] {
diff --git a/shell/completion.zsh b/shell/completion.zsh
@@ -23,12 +23,11 @@ _sk() {
 '*--with-nth=[Fields to be transformed]:WITH_NTH:_default' \
 '-d+[Delimiter between fields]:DELIMITER:_default' \
 '--delimiter=[Delimiter between fields]:DELIMITER:_default' \
-'--algo=[Fuzzy matching algorithm]:ALGORITHM:((skim_v1\:"Original skim fuzzy matching algorithm (v1)"
-skim_v2\:"Improved skim fuzzy matching algorithm (v2, default)"
+'--algo=[Fuzzy matching algorithm]:ALGORITHM:((skim_v2\:"Improved skim fuzzy matching algorithm (v2)"
 clangd\:"Clangd fuzzy matching algorithm"
 fzy\:"Fzy matching algorithm (https\://github.com/jhawthorn/fzy)"
 frizbee\:"Frizbee matching algorithm, typo resistant"
-arinae\:"Arinae\: typo-resistant & natural algorithm"))' \
+arinae\:"Arinae\: typo-resistant & natural algorithm, default"))' \
 '--case=[Case sensitivity]:CASE:((respect\:"Case-sensitive matching"
 ignore\:"Case-insensitive matching"
 smart\:"Smart case\: case-insensitive unless query contains uppercase"))' \
diff --git a/shell/version.txt b/shell/version.txt
@@ -1 +1 @@
-3.7.0
+4.0.0

Original file line number	Diff line number	Diff line change
`@@ -5,7 +5,7 @@ module completions {`
`5`	`5`	`}`
`6`	`6`
`7`	`7`	`def "nu-complete sk algorithm" [] {`
`8`		`- [ "skim_v1" "skim_v2" "clangd" "fzy" "frizbee" "arinae" ]`
	`8`	`+ [ "skim_v2" "clangd" "fzy" "frizbee" "arinae" ]`
`9`	`9`	`}`
`10`	`10`
`11`	`11`	`def "nu-complete sk case" [] {`