JS: use average character frequency of sample of JS files for variabl…

…e renaming, reduces GZIP size
tdewolff · Jan 4, 2022 · f2684cf · f2684cf
1 parent fb5ce73
commit f2684cf
Show file tree

Hide file tree

Showing 3 changed files with 17 additions and 11 deletions.
diff --git a/js/js.go b/js/js.go
@@ -23,9 +23,10 @@ var DefaultMinifier = &Minifier{}
 
 // Minifier is a JS minifier.
 type Minifier struct {
-	Precision         int // number of significant digits
-	KeepVarNames      bool
-	NoNullishOperator bool
+	Precision           int // number of significant digits
+	KeepVarNames        bool
+	useAlphabetVarNames bool
+	NoNullishOperator   bool
 }
 
 // Minify minifies JS data, it reads from r and writes to w.
@@ -56,7 +57,7 @@ func (o *Minifier) Minify(_ *minify.M, w io.Writer, r io.Reader, _ map[string]st
 	m := &jsMinifier{
 		o:       o,
 		w:       w,
-		renamer: newRenamer(!o.KeepVarNames),
+		renamer: newRenamer(!o.KeepVarNames, !o.useAlphabetVarNames),
 	}
 	m.hoistVars(&ast.BlockStmt)
 	ast.List = m.optimizeStmtList(ast.List, functionBlock)

diff --git a/js/js_test.go b/js/js_test.go
@@ -776,7 +776,7 @@ func TestJSVarRenaming(t *testing.T) {
 	}
 
 	m := minify.New()
-	o := Minifier{}
+	o := Minifier{useAlphabetVarNames: true}
 	for _, tt := range jsTests {
 		t.Run(tt.js, func(t *testing.T) {
 			r := bytes.NewBufferString(tt.js)
@@ -804,7 +804,7 @@ func TestWriterError(t *testing.T) {
 }
 
 func TestRenamerIndices(t *testing.T) {
-	renamer := newRenamer(true)
+	renamer := newRenamer(true, true)
 	for _, i := range []int{0, 1, 2, 53, 54, 55, 117, 118} {
 		name := renamer.getName([]byte{' '}, i)
 		j := renamer.getIndex(name)

diff --git a/js/vars.go b/js/vars.go
@@ -7,23 +7,28 @@ import (
 	"github.com/tdewolff/parse/v2/js"
 )
 
-// TODO: use character frequency for variable name mangling to improve gzip compression
-
 type renamer struct {
 	identStart    []byte
 	identContinue []byte
 	reserved      map[string]struct{}
 	rename        bool
 }
 
-func newRenamer(rename bool) *renamer {
+func newRenamer(rename, useCharFreq bool) *renamer {
 	reserved := make(map[string]struct{}, len(js.Keywords))
 	for name := range js.Keywords {
 		reserved[name] = struct{}{}
 	}
+	identStart := []byte("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ_$")
+	identContinue := []byte("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ_$0123456789")
+	if useCharFreq {
+		// sorted based on character frequency of a collection of JS samples (incl. the var names!)
+		identStart = []byte("etnsoiarclduhmfpgvbjy_wOxCEkASMFTzDNLRPHIBV$WUKqYGXQZJ")
+		identContinue = []byte("etnsoiarcldu14023hm8f6pg57v9bjy_wOxCEkASMFTzDNLRPHIBV$WUKqYGXQZJ")
+	}
 	return &renamer{
-		identStart:    []byte("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ_$"),
-		identContinue: []byte("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ_$0123456789"),
+		identStart:    identStart,
+		identContinue: identContinue,
 		reserved:      reserved,
 		rename:        rename,
 	}