Support mangle.nth_identifer to customize base54 behavior (#1038)

* Support nth_identifier option to customize base54

* Ensure base54 is operational immediately

README.md

@@ -898,6 +898,12 @@ If you happen to need the source map as a raw object, set `sourceMap.asObject` t
 - `module` (default `false`) -- Pass `true` an ES6 modules, where the toplevel
   scope is not the global scope. Implies `toplevel`.
 
+- `nth_identifier` (default: an internal mangler that weights based on character
+  frequency analysis) -- Pass an object with a `get(n)` function that converts an
+  ordinal into the nth most favored (usually shortest) identifier.
+  Optionally also provide `reset()`, `sort()`, and `consider(chars, delta)` to
+  use character frequency analysis of the source code.
+
 - `reserved` (default `[]`) -- Pass an array of identifiers that should be
   excluded from mangling. Example: `["foo", "bar"]`.
 
@@ -944,6 +950,12 @@ await minify(code, { mangle: { toplevel: true } }).code;
   - `false` -- `obj["prop"]` is mangled.
   - `true` -- `obj.prop` is mangled unless there is `obj["prop"]` elsewhere in the code.
 
+- `nth_identifer` (default: an internal mangler that weights based on character
+  frequency analysis) -- Pass an object with a `get(n)` function that converts an
+  ordinal into the nth most favored (usually shortest) identifier.
+  Optionally also provide `reset()`, `sort()`, and `consider(chars, delta)` to
+  use character frequency analysis of the source code.
+
 - `regex` (default: `null`) — Pass a [RegExp literal or pattern string](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/RegExp) to only mangle property matching the regular expression.
 
 - `reserved` (default: `[]`) — Do not mangle property names listed in the

lib/compress/index.js

@@ -390,7 +390,8 @@ class Compressor extends TreeWalker {
         var passes = +this.options.passes || 1;
         var min_count = 1 / 0;
         var stopping = false;
-        var mangle = { ie8: this.option("ie8") };
+        var nth_identifier = this.mangle_options && this.mangle_options.nth_identifier || base54;
+        var mangle = { ie8: this.option("ie8"), nth_identifier: nth_identifier };
         for (var pass = 0; pass < passes; pass++) {
             this._toplevel.figure_out_scope(mangle);
             if (pass === 0 && this.option("drop_console")) {
@@ -1889,7 +1890,7 @@ def_optimize(AST_Switch, function(self, compressor) {
     function is_inert_body(branch) {
         return !aborts(branch) && !make_node(AST_BlockStatement, branch, {
             body: branch.body
-        }).has_side_effects(compressor)
+        }).has_side_effects(compressor);
     }
 });
 
@@ -2262,6 +2263,7 @@ def_optimize(AST_Call, function(self, compressor) {
             argnames: [],
             body: []
         }).optimize(compressor);
+        var nth_identifier = compressor.mangle_options && compressor.mangle_options.nth_identifier || base54;
         if (self.args.every((x) => x instanceof AST_String)) {
             // quite a corner-case, but we can handle it:
             //   https://github.com/mishoo/UglifyJS2/issues/203
@@ -2271,14 +2273,13 @@ def_optimize(AST_Call, function(self, compressor) {
                     return arg.value;
                 }).join(",") + "){" + self.args[self.args.length - 1].value + "})";
                 var ast = parse(code);
-                var mangle = { ie8: compressor.option("ie8") };
+                var mangle = { ie8: compressor.option("ie8"), nth_identifier: nth_identifier };
                 ast.figure_out_scope(mangle);
                 var comp = new Compressor(compressor.options, {
                     mangle_options: compressor.mangle_options
                 });
                 ast = ast.transform(comp);
                 ast.figure_out_scope(mangle);
-                base54.reset();
                 ast.compute_char_frequency(mangle);
                 ast.mangle_names(mangle);
                 var fun;

lib/minify.js

@@ -11,7 +11,6 @@ import { AST_Toplevel, AST_Node } from "./ast.js";
 import { parse } from "./parse.js";
 import { OutputStream } from "./output.js";
 import { Compressor } from "./compress/index.js";
-import { base54 } from "./scope.js";
 import { SourceMap } from "./sourcemap.js";
 import {
     mangle_properties,
@@ -203,7 +202,6 @@ async function minify(files, options) {
     if (options.mangle) toplevel.figure_out_scope(options.mangle);
     if (timings) timings.mangle = Date.now();
     if (options.mangle) {
-        base54.reset();
         toplevel.compute_char_frequency(options.mangle);
         toplevel.mangle_names(options.mangle);
     }

lib/propmangle.js

@@ -146,12 +146,15 @@ function mangle_properties(ast, options) {
         cache: null,
         debug: false,
         keep_quoted: false,
+        nth_identifier: base54,
         only_cache: false,
         regex: null,
         reserved: null,
         undeclared: false,
     }, true);
 
+    var nth_identifier = options.nth_identifier;
+
     var reserved_option = options.reserved;
     if (!Array.isArray(reserved_option)) reserved_option = [reserved_option];
     var reserved = new Set(reserved_option);
@@ -312,7 +315,7 @@ function mangle_properties(ast, options) {
             // either debug mode is off, or it is on and we could not use the mangled name
             if (!mangled) {
                 do {
-                    mangled = base54(++cname);
+                    mangled = nth_identifier.get(++cname);
                 } while (!can_mangle(mangled));
             }
 
@@ -324,7 +327,7 @@ function mangle_properties(ast, options) {
     function mangle_private(name) {
         let mangled = private_cache.get(name);
         if (!mangled) {
-            mangled = base54(++cprivate);
+            mangled = nth_identifier.get(++cprivate);
             private_cache.set(name, mangled);
         }
 

lib/scope.js

@@ -667,8 +667,9 @@ AST_Scope.DEFMETHOD("def_variable", function(symbol, init) {
 
 function next_mangled(scope, options) {
     var ext = scope.enclosed;
+    var nth_identifier = options.nth_identifier;
     out: while (true) {
-        var m = base54(++scope.cname);
+        var m = nth_identifier.get(++scope.cname);
         if (ALL_RESERVED_WORDS.has(m)) continue; // skip over "do"
 
         // https://github.com/mishoo/UglifyJS2/issues/242 -- do not
@@ -744,6 +745,7 @@ AST_Symbol.DEFMETHOD("global", function() {
 AST_Toplevel.DEFMETHOD("_default_mangler_options", function(options) {
     options = defaults(options, {
         eval        : false,
+        nth_identifier : base54,
         ie8         : false,
         keep_classnames: false,
         keep_fnames : false,
@@ -765,6 +767,7 @@ AST_Toplevel.DEFMETHOD("_default_mangler_options", function(options) {
 
 AST_Toplevel.DEFMETHOD("mangle_names", function(options) {
     options = this._default_mangler_options(options);
+    var nth_identifier = options.nth_identifier;
 
     // We only need to mangle declaration nodes.  Special logic wired
     // into the code generator will display the mangled name if it's
@@ -816,7 +819,7 @@ AST_Toplevel.DEFMETHOD("mangle_names", function(options) {
         if (node instanceof AST_Label) {
             let name;
             do {
-                name = base54(++lname);
+                name = nth_identifier.get(++lname);
             } while (ALL_RESERVED_WORDS.has(name));
             node.mangled_name = name;
             return true;
@@ -878,9 +881,12 @@ AST_Toplevel.DEFMETHOD("find_colliding_names", function(options) {
 });
 
 AST_Toplevel.DEFMETHOD("expand_names", function(options) {
-    base54.reset();
-    base54.sort();
     options = this._default_mangler_options(options);
+    var nth_identifier = options.nth_identifier;
+    if (nth_identifier.reset && nth_identifier.sort) {
+        nth_identifier.reset();
+        nth_identifier.sort();
+    }
     var avoid = this.find_colliding_names(options);
     var cname = 0;
     this.globals.forEach(rename);
@@ -892,7 +898,7 @@ AST_Toplevel.DEFMETHOD("expand_names", function(options) {
     function next_name() {
         var name;
         do {
-            name = base54(cname++);
+            name = nth_identifier.get(cname++);
         } while (avoid.has(name) || ALL_RESERVED_WORDS.has(name));
         return name;
     }
@@ -919,30 +925,37 @@ AST_Sequence.DEFMETHOD("tail_node", function() {
 
 AST_Toplevel.DEFMETHOD("compute_char_frequency", function(options) {
     options = this._default_mangler_options(options);
+    var nth_identifier = options.nth_identifier;
+    if (!nth_identifier.reset || !nth_identifier.consider || !nth_identifier.sort) {
+        // If the identifier mangler is invariant, skip computing character frequency.
+        return;
+    }
+    nth_identifier.reset();
+
     try {
         AST_Node.prototype.print = function(stream, force_parens) {
             this._print(stream, force_parens);
             if (this instanceof AST_Symbol && !this.unmangleable(options)) {
-                base54.consider(this.name, -1);
+                nth_identifier.consider(this.name, -1);
             } else if (options.properties) {
                 if (this instanceof AST_DotHash) {
-                    base54.consider("#" + this.property, -1);
+                    nth_identifier.consider("#" + this.property, -1);
                 } else if (this instanceof AST_Dot) {
-                    base54.consider(this.property, -1);
+                    nth_identifier.consider(this.property, -1);
                 } else if (this instanceof AST_Sub) {
                     skip_string(this.property);
                 }
             }
         };
-        base54.consider(this.print_to_string(), 1);
+        nth_identifier.consider(this.print_to_string(), 1);
     } finally {
         AST_Node.prototype.print = AST_Node.prototype._print;
     }
-    base54.sort();
+    nth_identifier.sort();
 
     function skip_string(node) {
         if (node instanceof AST_String) {
-            base54.consider(node.value, -1);
+            nth_identifier.consider(node.value, -1);
         } else if (node instanceof AST_Conditional) {
             skip_string(node.consequent);
             skip_string(node.alternative);
@@ -966,19 +979,20 @@ const base54 = (() => {
             frequency.set(ch, 0);
         });
     }
-    base54.consider = function(str, delta) {
+    function consider(str, delta) {
         for (var i = str.length; --i >= 0;) {
             frequency.set(str[i], frequency.get(str[i]) + delta);
         }
-    };
+    }
     function compare(a, b) {
         return frequency.get(b) - frequency.get(a);
     }
-    base54.sort = function() {
+    function sort() {
         chars = mergeSort(leading, compare).concat(mergeSort(digits, compare));
-    };
-    base54.reset = reset;
+    }
+    // Ensure this is in a usable initial state.
     reset();
+    sort();
     function base54(num) {
         var ret = "", base = 54;
         num++;
@@ -990,7 +1004,13 @@ const base54 = (() => {
         } while (num > 0);
         return ret;
     }
-    return base54;
+
+    return {
+        get: base54,
+        consider,
+        reset,
+        sort
+    };
 })();
 
 export {

test/compress.js

@@ -211,7 +211,6 @@ async function run_compress_tests() {
             var output = cmp.compress(input);
             output.figure_out_scope(test.mangle);
             if (test.mangle) {
-                base54.reset();
                 output.compute_char_frequency(test.mangle);
                 (function(cache) {
                     if (!cache) return;

test/compress/properties.js

@@ -202,6 +202,31 @@ mangle_unquoted_properties: {
     }
 }
 
+mangle_nth_identifier: {
+    mangle = {
+        properties: {
+            nth_identifier: (function () {
+                function get(n) {
+                    return "zyxwvutsrq"[n];
+                }
+                return {
+                    get
+                };
+            })()
+        },
+    }
+    input: {
+        var a = {};
+        a.foo = "bar";
+        x = { baz: "ban" };
+    }
+    expect: {
+        var a = {};
+        a.v = "bar";
+        x = { u: "ban" };
+    }
+}
+
 mangle_debug: {
     mangle = {
         properties: {

test/compress/rename.js

@@ -362,6 +362,36 @@ mangle_catch_redef_2_ie8_toplevel: {
     expect_stdout: "undefined"
 }
 
+mangle_catch_nth_identifier: {
+    rename = true
+    options = {
+        ie8: false,
+        toplevel: true,
+    }
+    mangle = {
+        ie8: false,
+        toplevel: true,
+        nth_identifier: (function () {
+            function get(n) {
+                return "foo";
+            }
+            return {
+                get
+            };
+        })()
+    }
+    input: {
+        try {
+            throw "FAIL1";
+        } catch (a) {
+            var a = "FAIL2";
+        }
+        console.log(a);
+    }
+    expect_exact: 'try{throw"FAIL1"}catch(foo){var foo="FAIL2"}console.log(foo);'
+    expect_stdout: "undefined"
+}
+
 issue_2120_1: {
     rename = true
     mangle = {

tools/terser.d.ts

@@ -80,16 +80,52 @@ export interface MangleOptions {
     keep_classnames?: boolean | RegExp;
     keep_fnames?: boolean | RegExp;
     module?: boolean;
+    nth_identifier?: SimpleIdentifierMangler | WeightedIdentifierMangler;
     properties?: boolean | ManglePropertiesOptions;
     reserved?: string[];
     safari10?: boolean;
     toplevel?: boolean;
 }
 
+/**
+ * An identifier mangler for which the output is invariant with respect to the source code.
+ */
+export interface SimpleIdentifierMangler {
+    /**
+     * Obtains the nth most favored (usually shortest) identifier to rename a variable to.
+     * The mangler will increment n and retry until the return value is not in use in scope, and is not a reserved word.
+     * This function is expected to be stable; Evaluating get(n) === get(n) should always return true.
+     * @param n The ordinal of the identifier.
+     */
+    get(n: number): string;
+}
+
+/**
+ * An identifier mangler that leverages character frequency analysis to determine identifier precedence.
+ */
+export interface WeightedIdentifierMangler extends SimpleIdentifierMangler {
+    /**
+     * Modifies the internal weighting of the input characters by the specified delta.
+     * Will be invoked on the entire printed AST, and then deduct mangleable identifiers.
+     * @param chars The characters to modify the weighting of.
+     * @param delta The numeric weight to add to the characters.
+     */
+    consider(chars: string, delta: number): number;
+    /**
+     * Resets character weights.
+     */
+    reset(): void;
+    /**
+     * Sorts identifiers by character frequency, in preparation for calls to get(n).
+     */
+    sort(): void;
+}
+
 export interface ManglePropertiesOptions {
     builtins?: boolean;
     debug?: boolean;
     keep_quoted?: boolean | 'strict';
+    nth_identifier?: SimpleIdentifierMangler | WeightedIdentifierMangler;
     regex?: RegExp | string;
     reserved?: string[];
 }