Implement chilled strings

[Feature #20205] As a path toward enabling frozen string literals by default in the future, this commit introduce "chilled strings". From a user perspective chilled strings pretend to be frozen, but on the first attempt to mutate them, they lose their frozen status and emit a warning rather than to raise a `FrozenError`. Implementation wise, `rb_compile_option_struct.frozen_string_literal` is no longer a boolean but a tri-state of `enabled/disabled/unset`. When code is compiled with frozen string literals neither explictly enabled or disabled, string literals are compiled with a new `putchilledstring` instruction. This instruction is identical to `putstring` except it marks the String with the `STR_CHILLED (FL_USER3)` and `FL_FREEZE` flags. Chilled strings have the `FL_FREEZE` flag as to minimize the need to check for chilled strings across the codebase, and to improve compatibility with C extensions. Notes: - `String#freeze`: clears the chilled flag. - `String#-@`: acts as if the string was mutable. - `String#+@`: acts as if the string was mutable. - `String#clone`: copies the chilled flag. Co-authored-by: Jean Boussier <byroot@ruby-lang.org>
ruby · Mar 19, 2024 · 12be40a · 12be40a
1 parent 86b1531
commit 12be40a
Show file tree

Hide file tree

Showing 36 changed files with 714 additions and 282 deletions.
diff --git a/NEWS.md b/NEWS.md
@@ -7,6 +7,12 @@ Note that each entry is kept to a minimum, see links for details.
 
 ## Language changes
 
+* String literals in files without a `frozen_string_literal` comment now behave
+  as if they were frozen. If they are mutated a deprecation warning is emited.
+  These warnings can be enabled with `-W:deprecated` or by setting `Warning[:deprecated] = true`.
+  To disable this change you can run Ruby with the `--disable-frozen-string-literal` command line
+  argument. [Feature #20205]
+
 * `it` is added to reference a block parameter. [[Feature #18980]]
 
 * Keyword splatting `nil` when calling methods is now supported.

diff --git a/bootstraptest/test_ractor.rb b/bootstraptest/test_ractor.rb
@@ -1792,3 +1792,14 @@ class C8; def self.foo = 17; end
 }
 
 end # if !ENV['GITHUB_WORKFLOW']
+
+# Chilled strings are not shareable
+assert_equal 'false', %q{
+  Ractor.shareable?("chilled")
+}
+
+# Chilled strings can be made shareable
+assert_equal 'true', %q{
+  shareable = Ractor.make_shareable("chilled")
+  shareable == "chilled" && Ractor.shareable?(shareable)
+}
diff --git a/bootstraptest/test_yjit.rb b/bootstraptest/test_yjit.rb
@@ -4679,6 +4679,37 @@ def test(klass, args)
   test(KwInit, [Hash.ruby2_keywords_hash({1 => 1})])
 }
 
+# Chilled string setivar trigger warning
+assert_equal 'literal string will be frozen in the future', %q{
+  Warning[:deprecated] = true
+  $VERBOSE = true
+  $warning = "no-warning"
+  module ::Warning
+    def self.warn(message)
+      $warning = message.split("warning: ").last.strip
+    end
+  end
+
+  class String
+    def setivar!
+      @ivar = 42
+    end
+  end
+
+  def setivar!(str)
+    str.setivar!
+  end
+
+  10.times { setivar!("mutable".dup) }
+  10.times do
+    setivar!("frozen".freeze)
+  rescue FrozenError
+  end
+
+  setivar!("chilled") # Emit warning
+  $warning
+}
+
 # arity=-2 cfuncs
 assert_equal '["", "1/2", [0, [:ok, 1]]]', %q{
   def test_cases(file, chain)

diff --git a/class.c b/class.c
@@ -2244,7 +2244,10 @@ singleton_class_of(VALUE obj)
         return klass;
 
       case T_STRING:
-        if (FL_TEST_RAW(obj, RSTRING_FSTR)) {
+        if (CHILLED_STRING_P(obj)) {
+            CHILLED_STRING_MUTATED(obj);
+        }
+        else if (FL_TEST_RAW(obj, RSTRING_FSTR)) {
             rb_raise(rb_eTypeError, "can't define singleton");
         }
     }

diff --git a/compile.c b/compile.c
@@ -4723,7 +4723,7 @@ frozen_string_literal_p(const rb_iseq_t *iseq)
     return ISEQ_COMPILE_DATA(iseq)->option->frozen_string_literal > 0;
 }
 
-static inline int
+static inline bool
 static_literal_node_p(const NODE *node, const rb_iseq_t *iseq, bool hash_key)
 {
     switch (nd_type(node)) {
@@ -10365,12 +10365,18 @@ iseq_compile_each0(rb_iseq_t *iseq, LINK_ANCHOR *const ret, const NODE *const no
         debugp_param("nd_lit", get_string_value(node));
         if (!popped) {
             VALUE lit = get_string_value(node);
-            if (!frozen_string_literal_p(iseq)) {
+            switch (ISEQ_COMPILE_DATA(iseq)->option->frozen_string_literal) {
+              case ISEQ_FROZEN_STRING_LITERAL_UNSET:
+                lit = rb_fstring(lit);
+                ADD_INSN1(ret, node, putchilledstring, lit);
+                RB_OBJ_WRITTEN(iseq, Qundef, lit);
+                break;
+              case ISEQ_FROZEN_STRING_LITERAL_DISABLED:
                 lit = rb_fstring(lit);
                 ADD_INSN1(ret, node, putstring, lit);
                 RB_OBJ_WRITTEN(iseq, Qundef, lit);
-            }
-            else {
+                break;
+              case ISEQ_FROZEN_STRING_LITERAL_ENABLED:
                 if (ISEQ_COMPILE_DATA(iseq)->option->debug_frozen_string_literal || RTEST(ruby_debug)) {
                     VALUE debug_info = rb_ary_new_from_args(2, rb_iseq_path(iseq), INT2FIX(line));
                     lit = rb_str_dup(lit);
@@ -10382,6 +10388,9 @@ iseq_compile_each0(rb_iseq_t *iseq, LINK_ANCHOR *const ret, const NODE *const no
                 }
                 ADD_INSN1(ret, node, putobject, lit);
                 RB_OBJ_WRITTEN(iseq, Qundef, lit);
+                break;
+              default:
+                rb_bug("invalid frozen_string_literal");
             }
         }
         break;

diff --git a/error.c b/error.c
@@ -3860,6 +3860,12 @@ void
 rb_error_frozen_object(VALUE frozen_obj)
 {
     rb_yjit_lazy_push_frame(GET_EC()->cfp->pc);
+
+    if (CHILLED_STRING_P(frozen_obj)) {
+        CHILLED_STRING_MUTATED(frozen_obj);
+        return;
+    }
+
     VALUE debug_info;
     const ID created_info = id_debug_created_info;
     VALUE mesg = rb_sprintf("can't modify frozen %"PRIsVALUE": ",

diff --git a/ext/objspace/objspace_dump.c b/ext/objspace/objspace_dump.c
@@ -476,6 +476,8 @@ dump_object(VALUE obj, struct dump_config *dc)
             dump_append(dc, ", \"embedded\":true");
         if (FL_TEST(obj, RSTRING_FSTR))
             dump_append(dc, ", \"fstring\":true");
+        if (CHILLED_STRING_P(obj))
+            dump_append(dc, ", \"chilled\":true");
         if (STR_SHARED_P(obj))
             dump_append(dc, ", \"shared\":true");
         else

diff --git a/gems/bundled_gems b/gems/bundled_gems
@@ -5,8 +5,12 @@
 # - repository-url: URL from where clone for test
 # - revision: revision in repository-url to test
 #   if `revision` is not given, "v"+`version` or `version` will be used.
-minitest        5.22.3  https://github.com/minitest/minitest
-power_assert    2.0.3   https://github.com/ruby/power_assert
+
+# Waiting for https://github.com/minitest/minitest/pull/991
+minitest        5.22.3  https://github.com/Shopify/minitest b5f5202575894796e00109a8f8a5041b778991ee
+
+# Waiting for https://github.com/ruby/power_assert/pull/48
+power_assert    2.0.3   https://github.com/ruby/power_assert 78dd2ab3ccd93796d83c0b35b978c39bfabb938c
 rake            13.1.0  https://github.com/ruby/rake
 test-unit       3.6.2   https://github.com/test-unit/test-unit
 rexml           3.2.6   https://github.com/ruby/rexml
@@ -17,8 +21,8 @@ net-pop         0.1.2   https://github.com/ruby/net-pop
 net-smtp        0.4.0.1 https://github.com/ruby/net-smtp
 matrix          0.4.2   https://github.com/ruby/matrix
 prime           0.1.2   https://github.com/ruby/prime
-rbs             3.4.4   https://github.com/ruby/rbs 61b412bc7ba00519e7d6d08450bd384990d94ea2
-typeprof        0.21.11 https://github.com/ruby/typeprof
+rbs             3.4.4   https://github.com/ruby/rbs ba7872795d5de04adb8ff500c0e6afdc81a041dd
+typeprof        0.21.11 https://github.com/ruby/typeprof b19a6416da3a05d57fadd6ffdadb382b6d236ca5
 debug           1.9.1   https://github.com/ruby/debug 2d602636d99114d55a32fedd652c9c704446a749
 racc            1.7.3   https://github.com/ruby/racc
 mutex_m         0.2.0   https://github.com/ruby/mutex_m

diff --git a/include/ruby/internal/fl_type.h b/include/ruby/internal/fl_type.h
@@ -916,6 +916,9 @@ static inline void
 RB_OBJ_FREEZE_RAW(VALUE obj)
 {
     RB_FL_SET_RAW(obj, RUBY_FL_FREEZE);
+    if (TYPE(obj) == T_STRING) {
+        RB_FL_UNSET_RAW(obj, FL_USER3); // STR_CHILLED
+    }
 }
 
 RUBY_SYMBOL_EXPORT_BEGIN

diff --git a/include/ruby/internal/intern/error.h b/include/ruby/internal/intern/error.h
@@ -190,7 +190,6 @@ RBIMPL_ATTR_NONNULL(())
  */
 void rb_error_frozen(const char *what);
 
-RBIMPL_ATTR_NORETURN()
 /**
  * Identical  to  rb_error_frozen(),  except  it takes  arbitrary  Ruby  object
  * instead of C's string.

diff --git a/insns.def b/insns.def
@@ -375,7 +375,17 @@ putstring
 ()
 (VALUE val)
 {
-    val = rb_ec_str_resurrect(ec, str);
+    val = rb_ec_str_resurrect(ec, str, false);
+}
+
+/* put chilled string val. string will be copied but frozen in the future. */
+DEFINE_INSN
+putchilledstring
+(VALUE str)
+()
+(VALUE val)
+{
+    val = rb_ec_str_resurrect(ec, str, true);
 }
 
 /* put concatenate strings */

diff --git a/internal/string.h b/internal/string.h
@@ -17,6 +17,7 @@
 
 #define STR_NOEMBED      FL_USER1
 #define STR_SHARED       FL_USER2 /* = ELTS_SHARED */
+#define STR_CHILLED      FL_USER3
 
 #ifdef rb_fstring_cstr
 # undef rb_fstring_cstr
@@ -77,7 +78,7 @@ VALUE rb_id_quote_unprintable(ID);
 VALUE rb_sym_proc_call(ID mid, int argc, const VALUE *argv, int kw_splat, VALUE passed_proc);
 
 struct rb_execution_context_struct;
-VALUE rb_ec_str_resurrect(struct rb_execution_context_struct *ec, VALUE str);
+VALUE rb_ec_str_resurrect(struct rb_execution_context_struct *ec, VALUE str, bool chilled);
 
 #define rb_fstring_lit(str) rb_fstring_new((str), rb_strlen_lit(str))
 #define rb_fstring_literal(str) rb_fstring_lit(str)
@@ -108,6 +109,26 @@ STR_SHARED_P(VALUE str)
     return FL_ALL_RAW(str, STR_NOEMBED | STR_SHARED);
 }
 
+static inline bool
+CHILLED_STRING_P(VALUE obj)
+{
+    return RB_TYPE_P(obj, T_STRING) && FL_TEST_RAW(obj, STR_CHILLED);
+}
+
+static inline void
+CHILLED_STRING_MUTATED(VALUE str)
+{
+    rb_category_warn(RB_WARN_CATEGORY_DEPRECATED, "literal string will be frozen in the future");
+    FL_UNSET_RAW(str, STR_CHILLED | FL_FREEZE);
+}
+
+static inline void
+STR_CHILL_RAW(VALUE str)
+{
+    // Chilled strings are always also frozen
+    FL_SET_RAW(str, STR_CHILLED | RUBY_FL_FREEZE);
+}
+
 static inline bool
 is_ascii_string(VALUE str)
 {

diff --git a/iseq.c b/iseq.c
@@ -720,18 +720,20 @@ finish_iseq_build(rb_iseq_t *iseq)
 }
 
 static rb_compile_option_t COMPILE_OPTION_DEFAULT = {
-    OPT_INLINE_CONST_CACHE, /* int inline_const_cache; */
-    OPT_PEEPHOLE_OPTIMIZATION, /* int peephole_optimization; */
-    OPT_TAILCALL_OPTIMIZATION, /* int tailcall_optimization */
-    OPT_SPECIALISED_INSTRUCTION, /* int specialized_instruction; */
-    OPT_OPERANDS_UNIFICATION, /* int operands_unification; */
-    OPT_INSTRUCTIONS_UNIFICATION, /* int instructions_unification; */
-    OPT_FROZEN_STRING_LITERAL,
-    OPT_DEBUG_FROZEN_STRING_LITERAL,
-    TRUE,			/* coverage_enabled */
+    .inline_const_cache = OPT_INLINE_CONST_CACHE,
+    .peephole_optimization = OPT_PEEPHOLE_OPTIMIZATION,
+    .tailcall_optimization = OPT_TAILCALL_OPTIMIZATION,
+    .specialized_instruction = OPT_SPECIALISED_INSTRUCTION,
+    .operands_unification = OPT_OPERANDS_UNIFICATION,
+    .instructions_unification = OPT_INSTRUCTIONS_UNIFICATION,
+    .frozen_string_literal = OPT_FROZEN_STRING_LITERAL,
+    .debug_frozen_string_literal = OPT_DEBUG_FROZEN_STRING_LITERAL,
+    .coverage_enabled = TRUE,
 };
 
-static const rb_compile_option_t COMPILE_OPTION_FALSE = {0};
+static const rb_compile_option_t COMPILE_OPTION_FALSE = {
+    .frozen_string_literal = -1, // unspecified
+};
 
 int
 rb_iseq_opt_frozen_string_literal(void)
@@ -770,9 +772,11 @@ set_compile_option_from_ast(rb_compile_option_t *option, const rb_ast_body_t *as
 {
 #define SET_COMPILE_OPTION(o, a, mem) \
     ((a)->mem < 0 ? 0 : ((o)->mem = (a)->mem > 0))
-    SET_COMPILE_OPTION(option, ast, frozen_string_literal);
     SET_COMPILE_OPTION(option, ast, coverage_enabled);
 #undef SET_COMPILE_OPTION
+    if (ast->frozen_string_literal >= 0) {
+        option->frozen_string_literal = ast->frozen_string_literal;
+    }
     return option;
 }
 
@@ -814,13 +818,14 @@ make_compile_option_value(rb_compile_option_t *option)
         SET_COMPILE_OPTION(option, opt, specialized_instruction);
         SET_COMPILE_OPTION(option, opt, operands_unification);
         SET_COMPILE_OPTION(option, opt, instructions_unification);
-        SET_COMPILE_OPTION(option, opt, frozen_string_literal);
         SET_COMPILE_OPTION(option, opt, debug_frozen_string_literal);
         SET_COMPILE_OPTION(option, opt, coverage_enabled);
         SET_COMPILE_OPTION_NUM(option, opt, debug_level);
     }
 #undef SET_COMPILE_OPTION
 #undef SET_COMPILE_OPTION_NUM
+    VALUE frozen_string_literal = option->frozen_string_literal == -1 ? Qnil : RBOOL(option->frozen_string_literal);
+    rb_hash_aset(opt, ID2SYM(rb_intern("frozen_string_literal")), frozen_string_literal);
     return opt;
 }
 
@@ -1248,7 +1253,7 @@ pm_iseq_compile_with_option(VALUE src, VALUE file, VALUE realpath, VALUE line, V
     pm_parse_result_t result = { 0 };
     pm_options_line_set(&result.options, NUM2INT(line));
 
-    pm_options_frozen_string_literal_set(&result.options, option.frozen_string_literal);
+    pm_options_frozen_string_literal_init(&result, option.frozen_string_literal);
 
     VALUE error;
     if (RB_TYPE_P(src, T_FILE)) {

diff --git a/iseq.h b/iseq.h
@@ -47,6 +47,10 @@ extern const ID rb_iseq_shared_exc_local_tbl[];
 
 #define ISEQ_FLIP_CNT(iseq) ISEQ_BODY(iseq)->variable.flip_count
 
+#define ISEQ_FROZEN_STRING_LITERAL_ENABLED 1
+#define ISEQ_FROZEN_STRING_LITERAL_DISABLED 0
+#define ISEQ_FROZEN_STRING_LITERAL_UNSET -1
+
 static inline rb_snum_t
 ISEQ_FLIP_CNT_INCREMENT(const rb_iseq_t *iseq)
 {
@@ -227,7 +231,7 @@ struct rb_compile_option_struct {
     unsigned int specialized_instruction: 1;
     unsigned int operands_unification: 1;
     unsigned int instructions_unification: 1;
-    unsigned int frozen_string_literal: 1;
+    signed int frozen_string_literal: 2; /* -1: not specified, 0: false, 1: true */
     unsigned int debug_frozen_string_literal: 1;
     unsigned int coverage_enabled: 1;
     int debug_level;

diff --git a/lib/ruby_vm/rjit/insn_compiler.rb b/lib/ruby_vm/rjit/insn_compiler.rb
@@ -57,6 +57,7 @@ def compile(jit, ctx, asm, insn)
       when :putobject then putobject(jit, ctx, asm)
       when :putspecialobject then putspecialobject(jit, ctx, asm)
       when :putstring then putstring(jit, ctx, asm)
+      when :putchilledstring then putchilledstring(jit, ctx, asm)
       when :concatstrings then concatstrings(jit, ctx, asm)
       when :anytostring then anytostring(jit, ctx, asm)
       when :toregexp then toregexp(jit, ctx, asm)
@@ -776,6 +777,27 @@ def putstring(jit, ctx, asm)
 
       asm.mov(C_ARGS[0], EC)
       asm.mov(C_ARGS[1], to_value(put_val))
+      asm.mov(C_ARGS[2], 0)
+      asm.call(C.rb_ec_str_resurrect)
+
+      stack_top = ctx.stack_push(Type::TString)
+      asm.mov(stack_top, C_RET)
+
+      KeepCompiling
+    end
+
+    # @param jit [RubyVM::RJIT::JITState]
+    # @param ctx [RubyVM::RJIT::Context]
+    # @param asm [RubyVM::RJIT::Assembler]
+    def putchilledstring(jit, ctx, asm)
+      put_val = jit.operand(0, ruby: true)
+
+      # Save the PC and SP because the callee will allocate
+      jit_prepare_routine_call(jit, ctx, asm)
+
+      asm.mov(C_ARGS[0], EC)
+      asm.mov(C_ARGS[1], to_value(put_val))
+      asm.mov(C_ARGS[2], 1)
       asm.call(C.rb_ec_str_resurrect)
 
       stack_top = ctx.stack_push(Type::TString)

diff --git a/mini_builtin.c b/mini_builtin.c
@@ -28,16 +28,16 @@ builtin_iseq_load(const char *feature_name, const struct rb_builtin_function *ta
     }
     vm->builtin_function_table = table;
     static const rb_compile_option_t optimization = {
-        TRUE, /* unsigned int inline_const_cache; */
-        TRUE, /* unsigned int peephole_optimization; */
-        FALSE,/* unsigned int tailcall_optimization; */
-        TRUE, /* unsigned int specialized_instruction; */
-        TRUE, /* unsigned int operands_unification; */
-        TRUE, /* unsigned int instructions_unification; */
-        TRUE, /* unsigned int frozen_string_literal; */
-        FALSE, /* unsigned int debug_frozen_string_literal; */
-        FALSE, /* unsigned int coverage_enabled; */
-        0, /* int debug_level; */
+        .inline_const_cache = TRUE,
+        .peephole_optimization = TRUE,
+        .tailcall_optimization = FALSE,
+        .specialized_instruction = TRUE,
+        .operands_unification = TRUE,
+        .instructions_unification = TRUE,
+        .frozen_string_literal = TRUE,
+        .debug_frozen_string_literal = FALSE,
+        .coverage_enabled = FALSE,
+        .debug_level = 0,
     };
     const rb_iseq_t *iseq = rb_iseq_new_with_opt(&ast->body, name_str, name_str, Qnil, 0, NULL, 0, ISEQ_TYPE_TOP, &optimization);
     GET_VM()->builtin_function_table = NULL;