Class: RubyVM::InstructionSequence

Inherits:
Object
  • Object
show all
Defined in:
iseq.c,
iseq.c

Overview

The InstructionSequence class represents a compiled sequence of instructions for the Virtual Machine used in MRI. Not all implementations of Ruby may implement this class, and for the implementations that implement it, the methods defined and behavior of the methods can change in any version.

With it, you can get a handle to the instructions that make up a method or a proc, compile strings of Ruby code down to VM instructions, and disassemble instruction sequences to strings for easy inspection. It is mostly useful if you want to learn how YARV works, but it also lets you control various settings for the Ruby iseq compiler.

You can find the source for the VM instructions in insns.def in the Ruby source.

The instruction sequence results will almost certainly change as Ruby changes, so example output in this documentation may be different from what you see.

Of course, this class is MRI specific.

Class Method Summary collapse

Instance Method Summary collapse

Class Method Details

.compile(source[, file[, path[, line[, options]]]]) ⇒ Object .new(source[, file[, path[, line[, options]]]]) ⇒ Object

Takes source, a String of Ruby code and compiles it to an InstructionSequence.

Optionally takes file, path, and line which describe the file path, real path and first line number of the ruby code in source which are metadata attached to the returned iseq.

file is used for ‘__FILE__` and exception backtrace. path is used for require_relative base. It is recommended these should be the same full path.

options, which can be true, false or a Hash, is used to modify the default behavior of the Ruby iseq compiler.

For details regarding valid compile options see ::compile_option=.

RubyVM::InstructionSequence.compile("a = 1 + 2")
#=> <RubyVM::InstructionSequence:<compiled>@<compiled>>

path = "test.rb"
RubyVM::InstructionSequence.compile(File.read(path), path, File.expand_path(path))
#=> <RubyVM::InstructionSequence:<compiled>@test.rb:1>

path = File.expand_path("test.rb")
RubyVM::InstructionSequence.compile(File.read(path), path, path)
#=> <RubyVM::InstructionSequence:<compiled>@/absolute/path/to/test.rb:1>


1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
# File 'iseq.c', line 1255

static VALUE
iseqw_s_compile(int argc, VALUE *argv, VALUE self)
{
    VALUE src, file = Qnil, path = Qnil, line = INT2FIX(1), opt = Qnil;
    int i;

    i = rb_scan_args(argc, argv, "1*:", &src, NULL, &opt);
    if (i > 4+NIL_P(opt)) rb_error_arity(argc, 1, 5);
    switch (i) {
      case 5: opt = argv[--i];
      case 4: line = argv[--i];
      case 3: path = argv[--i];
      case 2: file = argv[--i];
    }

    if (NIL_P(file)) file = rb_fstring_lit("<compiled>");
    if (NIL_P(path)) path = file;
    if (NIL_P(line)) line = INT2FIX(1);

    Check_Type(path, T_STRING);
    Check_Type(file, T_STRING);

    return iseqw_new(rb_iseq_compile_with_option(src, file, path, line, opt));
}

.compile_file(file[, options]) ⇒ Object

Takes file, a String with the location of a Ruby source file, reads, parses and compiles the file, and returns iseq, the compiled InstructionSequence with source location metadata set.

Optionally takes options, which can be true, false or a Hash, to modify the default behavior of the Ruby iseq compiler.

For details regarding valid compile options see ::compile_option=.

# /tmp/hello.rb
puts "Hello, world!"

# elsewhere
RubyVM::InstructionSequence.compile_file("/tmp/hello.rb")
#=> <RubyVM::InstructionSequence:<main>@/tmp/hello.rb>


1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
# File 'iseq.c', line 1300

static VALUE
iseqw_s_compile_file(int argc, VALUE *argv, VALUE self)
{
    VALUE file, line = INT2FIX(1), opt = Qnil;
    VALUE parser, f, exc = Qnil, ret;
    rb_ast_t *ast;
    rb_compile_option_t option;
    int i;

    i = rb_scan_args(argc, argv, "1*:", &file, NULL, &opt);
    if (i > 1+NIL_P(opt)) rb_error_arity(argc, 1, 2);
    switch (i) {
      case 2: opt = argv[--i];
    }
    FilePathValue(file);
    file = rb_fstring(file); /* rb_io_t->pathv gets frozen anyways */

    f = rb_file_open_str(file, "r");

    parser = rb_parser_new();
    rb_parser_set_context(parser, NULL, FALSE);
    ast = (rb_ast_t *)rb_parser_load_file(parser, file);
    if (!ast->body.root) exc = GET_EC()->errinfo;

    rb_io_close(f);
    if (!ast->body.root) {
	rb_ast_dispose(ast);
	rb_exc_raise(exc);
    }

    make_compile_option(&option, opt);

    ret = iseqw_new(rb_iseq_new_with_opt(&ast->body, rb_fstring_lit("<main>"),
					 file,
					 rb_realpath_internal(Qnil, file, 1),
					 line, NULL, 0, ISEQ_TYPE_TOP, &option));
    rb_ast_dispose(ast);
    return ret;
}

.compile_optionObject

Returns a hash of default options used by the Ruby iseq compiler.

For details, see InstructionSequence.compile_option=.



1388
1389
1390
1391
1392
# File 'iseq.c', line 1388

static VALUE
iseqw_s_compile_option_get(VALUE self)
{
    return make_compile_option_value(&COMPILE_OPTION_DEFAULT);
}

.compile_option=(options) ⇒ Object

Sets the default values for various optimizations in the Ruby iseq compiler.

Possible values for options include true, which enables all options, false which disables all options, and nil which leaves all options unchanged.

You can also pass a Hash of options that you want to change, any options not present in the hash will be left unchanged.

Possible option names (which are keys in options) which can be set to true or false include:

  • :inline_const_cache

  • :instructions_unification

  • :operands_unification

  • :peephole_optimization

  • :specialized_instruction

  • :stack_caching

  • :tailcall_optimization

Additionally, :debug_level can be set to an integer.

These default options can be overwritten for a single run of the iseq compiler by passing any of the above values as the options parameter to ::new, ::compile and ::compile_file.



1371
1372
1373
1374
1375
1376
1377
1378
# File 'iseq.c', line 1371

static VALUE
iseqw_s_compile_option_set(VALUE self, VALUE opt)
{
    rb_compile_option_t option;
    make_compile_option(&option, opt);
    COMPILE_OPTION_DEFAULT = option;
    return opt;
}

.disasm(body) ⇒ String .disassemble(body) ⇒ String

Takes body, a Method or Proc object, and returns a String with the human readable instructions for body.

For a Method object:

# /tmp/method.rb
def hello
  puts "hello, world"
end

puts RubyVM::InstructionSequence.disasm(method(:hello))

Produces:

== disasm: <RubyVM::InstructionSequence:hello@/tmp/method.rb>============
0000 trace            8                                               (   1)
0002 trace            1                                               (   2)
0004 putself
0005 putstring        "hello, world"
0007 send             :puts, 1, nil, 8, <ic:0>
0013 trace            16                                              (   3)
0015 leave                                                            (   2)

For a Proc:

# /tmp/proc.rb
p = proc { num = 1 + 2 }
puts RubyVM::InstructionSequence.disasm(p)

Produces:

== disasm: <RubyVM::InstructionSequence:block in <main>@/tmp/proc.rb>===
== catch table
| catch type: redo   st: 0000 ed: 0012 sp: 0000 cont: 0000
| catch type: next   st: 0000 ed: 0012 sp: 0000 cont: 0012
|------------------------------------------------------------------------
local table (size: 2, argc: 0 [opts: 0, rest: -1, post: 0, block: -1] s1)
[ 2] num
0000 trace            1                                               (   1)
0002 putobject        1
0004 putobject        2
0006 opt_plus         <ic:1>
0008 dup
0009 setlocal         num, 0
0012 leave

Overloads:



2585
2586
2587
2588
2589
2590
# File 'iseq.c', line 2585

static VALUE
iseqw_s_disasm(VALUE klass, VALUE body)
{
    VALUE iseqw = iseqw_s_of(klass, body);
    return NIL_P(iseqw) ? Qnil : rb_iseq_disasm(iseqw_check(iseqw));
}

.disasm(body) ⇒ String .disassemble(body) ⇒ String

Takes body, a Method or Proc object, and returns a String with the human readable instructions for body.

For a Method object:

# /tmp/method.rb
def hello
  puts "hello, world"
end

puts RubyVM::InstructionSequence.disasm(method(:hello))

Produces:

== disasm: <RubyVM::InstructionSequence:hello@/tmp/method.rb>============
0000 trace            8                                               (   1)
0002 trace            1                                               (   2)
0004 putself
0005 putstring        "hello, world"
0007 send             :puts, 1, nil, 8, <ic:0>
0013 trace            16                                              (   3)
0015 leave                                                            (   2)

For a Proc:

# /tmp/proc.rb
p = proc { num = 1 + 2 }
puts RubyVM::InstructionSequence.disasm(p)

Produces:

== disasm: <RubyVM::InstructionSequence:block in <main>@/tmp/proc.rb>===
== catch table
| catch type: redo   st: 0000 ed: 0012 sp: 0000 cont: 0000
| catch type: next   st: 0000 ed: 0012 sp: 0000 cont: 0012
|------------------------------------------------------------------------
local table (size: 2, argc: 0 [opts: 0, rest: -1, post: 0, block: -1] s1)
[ 2] num
0000 trace            1                                               (   1)
0002 putobject        1
0004 putobject        2
0006 opt_plus         <ic:1>
0008 dup
0009 setlocal         num, 0
0012 leave

Overloads:



2585
2586
2587
2588
2589
2590
# File 'iseq.c', line 2585

static VALUE
iseqw_s_disasm(VALUE klass, VALUE body)
{
    VALUE iseqw = iseqw_s_of(klass, body);
    return NIL_P(iseqw) ? Qnil : rb_iseq_disasm(iseqw_check(iseqw));
}

.load(*args) ⇒ Object

:nodoc:



1023
1024
1025
1026
1027
1028
1029
# File 'iseq.c', line 1023

static VALUE
iseq_s_load(int argc, VALUE *argv, VALUE self)
{
    VALUE data, opt=Qnil;
    rb_scan_args(argc, argv, "11", &data, &opt);
    return iseq_load(data, NULL, opt);
}

.RubyVM::InstructionSequence.load_from_binary(binary) ⇒ Object

Load an iseq object from binary format String object created by RubyVM::InstructionSequence.to_binary.

This loader does not have a verifier, so that loading broken/modified binary causes critical problem.

You should not load binary data provided by others. You should use binary data translated by yourself.



3427
3428
3429
3430
3431
# File 'iseq.c', line 3427

static VALUE
iseqw_s_load_from_binary(VALUE self, VALUE str)
{
    return iseqw_new(rb_iseq_ibf_load(str));
}

.RubyVM::InstructionSequence.load_from_binary_extra_data(binary) ⇒ String

Load extra data embed into binary format String object.

Returns:



3439
3440
3441
3442
3443
# File 'iseq.c', line 3439

static VALUE
iseqw_s_load_from_binary_extra_data(VALUE self, VALUE str)
{
    return rb_iseq_ibf_load_extra_data(str);
}

.compile(source[, file[, path[, line[, options]]]]) ⇒ Object .new(source[, file[, path[, line[, options]]]]) ⇒ Object

Takes source, a String of Ruby code and compiles it to an InstructionSequence.

Optionally takes file, path, and line which describe the file path, real path and first line number of the ruby code in source which are metadata attached to the returned iseq.

file is used for ‘__FILE__` and exception backtrace. path is used for require_relative base. It is recommended these should be the same full path.

options, which can be true, false or a Hash, is used to modify the default behavior of the Ruby iseq compiler.

For details regarding valid compile options see ::compile_option=.

RubyVM::InstructionSequence.compile("a = 1 + 2")
#=> <RubyVM::InstructionSequence:<compiled>@<compiled>>

path = "test.rb"
RubyVM::InstructionSequence.compile(File.read(path), path, File.expand_path(path))
#=> <RubyVM::InstructionSequence:<compiled>@test.rb:1>

path = File.expand_path("test.rb")
RubyVM::InstructionSequence.compile(File.read(path), path, path)
#=> <RubyVM::InstructionSequence:<compiled>@/absolute/path/to/test.rb:1>


1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
# File 'iseq.c', line 1255

static VALUE
iseqw_s_compile(int argc, VALUE *argv, VALUE self)
{
    VALUE src, file = Qnil, path = Qnil, line = INT2FIX(1), opt = Qnil;
    int i;

    i = rb_scan_args(argc, argv, "1*:", &src, NULL, &opt);
    if (i > 4+NIL_P(opt)) rb_error_arity(argc, 1, 5);
    switch (i) {
      case 5: opt = argv[--i];
      case 4: line = argv[--i];
      case 3: path = argv[--i];
      case 2: file = argv[--i];
    }

    if (NIL_P(file)) file = rb_fstring_lit("<compiled>");
    if (NIL_P(path)) path = file;
    if (NIL_P(line)) line = INT2FIX(1);

    Check_Type(path, T_STRING);
    Check_Type(file, T_STRING);

    return iseqw_new(rb_iseq_compile_with_option(src, file, path, line, opt));
}

.of(body) ⇒ Object

Returns the instruction sequence containing the given proc or method.

For example, using irb:

# a proc > p = proc { num = 1 + 2 } > RubyVM::InstructionSequence.of(p) > #=> <RubyVM::InstructionSequence:block in irb_binding@(irb)>

# for a method > def foo(bar); puts bar; end > RubyVM::InstructionSequence.of(method(:foo)) > #=> <RubyVM::InstructionSequence:foo@(irb)>

Using ::compile_file:

# /tmp/iseq_of.rb def hello

puts "hello, world"

end

$a_global_proc = proc { str = ‘a’ + ‘b’ }

# in irb > require ‘/tmp/iseq_of.rb’

# first the method hello > RubyVM::InstructionSequence.of(method(:hello)) > #=> #<RubyVM::InstructionSequence:0x007fb73d7cb1d0>

# then the global proc > RubyVM::InstructionSequence.of($a_global_proc) > #=> #<RubyVM::InstructionSequence:0x007fb73d7caf78>



2511
2512
2513
2514
2515
2516
2517
2518
2519
2520
2521
2522
2523
2524
2525
2526
2527
2528
2529
2530
2531
# File 'iseq.c', line 2511

static VALUE
iseqw_s_of(VALUE klass, VALUE body)
{
    const rb_iseq_t *iseq = NULL;

    if (rb_obj_is_proc(body)) {
        iseq = vm_proc_iseq(body);

        if (!rb_obj_is_iseq((VALUE)iseq)) {
            iseq = NULL;
        }
    }
    else if (rb_obj_is_method(body)) {
        iseq = rb_method_iseq(body);
    }
    else if (rb_typeddata_is_instance_of(body, &iseqw_data_type)) {
        return body;
    }

    return iseq ? iseqw_new(iseq) : Qnil;
}

Instance Method Details

#absolute_pathObject

Returns the absolute path of this instruction sequence.

nil if the iseq was evaluated from a string.

For example, using ::compile_file:

# /tmp/method.rb def hello

puts "hello, world"

end

# in irb > iseq = RubyVM::InstructionSequence.compile_file(‘/tmp/method.rb’) > iseq.absolute_path #=> /tmp/method.rb



1496
1497
1498
1499
1500
# File 'iseq.c', line 1496

static VALUE
iseqw_absolute_path(VALUE self)
{
    return rb_iseq_realpath(iseqw_check(self));
}

#base_labelObject

Returns the base label of this instruction sequence.

For example, using irb:

iseq = RubyVM::InstructionSequence.compile(‘num = 1 + 2’) #=> <RubyVM::InstructionSequence:<compiled>@<compiled>> iseq.base_label #=> “<compiled>”

Using ::compile_file:

# /tmp/method.rb def hello

puts "hello, world"

end

# in irb > iseq = RubyVM::InstructionSequence.compile_file(‘/tmp/method.rb’) > iseq.base_label #=> <main>



1551
1552
1553
1554
1555
# File 'iseq.c', line 1551

static VALUE
iseqw_base_label(VALUE self)
{
    return rb_iseq_base_label(iseqw_check(self));
}

#disasmString #disassembleString

Returns the instruction sequence as a String in human readable form.

puts RubyVM::InstructionSequence.compile('1 + 2').disasm

Produces:

== disasm: <RubyVM::InstructionSequence:<compiled>@<compiled>>==========
0000 trace            1                                               (   1)
0002 putobject        1
0004 putobject        2
0006 opt_plus         <ic:1>
0008 leave

Overloads:



2360
2361
2362
2363
2364
# File 'iseq.c', line 2360

static VALUE
iseqw_disasm(VALUE self)
{
    return rb_iseq_disasm(iseqw_check(self));
}

#disasmString #disassembleString

Returns the instruction sequence as a String in human readable form.

puts RubyVM::InstructionSequence.compile('1 + 2').disasm

Produces:

== disasm: <RubyVM::InstructionSequence:<compiled>@<compiled>>==========
0000 trace            1                                               (   1)
0002 putobject        1
0004 putobject        2
0006 opt_plus         <ic:1>
0008 leave

Overloads:



2360
2361
2362
2363
2364
# File 'iseq.c', line 2360

static VALUE
iseqw_disasm(VALUE self)
{
    return rb_iseq_disasm(iseqw_check(self));
}

#each_child {|child_iseq| ... } ⇒ Object

Iterate all direct child instruction sequences. Iteration order is implementation/version defined so that people should not rely on the order.

Yields:

  • (child_iseq)


2430
2431
2432
2433
2434
2435
2436
# File 'iseq.c', line 2430

static VALUE
iseqw_each_child(VALUE self)
{
    const rb_iseq_t *iseq = iseqw_check(self);
    iseq_iterate_children(iseq, yield_each_children, NULL);
    return self;
}

#evalObject

Evaluates the instruction sequence and returns the result.

RubyVM::InstructionSequence.compile("1 + 2").eval #=> 3

Returns:



1423
1424
1425
1426
1427
# File 'iseq.c', line 1423

static VALUE
iseqw_eval(VALUE self)
{
    return rb_iseq_eval(iseqw_check(self));
}

#first_linenoObject

Returns the number of the first source line where the instruction sequence was loaded from.

For example, using irb:

iseq = RubyVM::InstructionSequence.compile(‘num = 1 + 2’) #=> <RubyVM::InstructionSequence:<compiled>@<compiled>> iseq.first_lineno #=> 1



1567
1568
1569
1570
1571
# File 'iseq.c', line 1567

static VALUE
iseqw_first_lineno(VALUE self)
{
    return rb_iseq_first_lineno(iseqw_check(self));
}

#inspectObject

Returns a human-readable string representation of this instruction sequence, including the #label and #path.



1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
# File 'iseq.c', line 1433

static VALUE
iseqw_inspect(VALUE self)
{
    const rb_iseq_t *iseq = iseqw_check(self);
    const struct rb_iseq_constant_body *const body = iseq->body;
    VALUE klass = rb_class_name(rb_obj_class(self));

    if (!body->location.label) {
	return rb_sprintf("#<%"PRIsVALUE": uninitialized>", klass);
    }
    else {
	return rb_sprintf("<%"PRIsVALUE":%"PRIsVALUE"@%"PRIsVALUE":%d>",
			  klass,
			  body->location.label, rb_iseq_path(iseq),
			  FIX2INT(rb_iseq_first_lineno(iseq)));
    }
}

#labelObject

Returns the label of this instruction sequence.

<main> if it’s at the top level, <compiled> if it was evaluated from a string.

For example, using irb:

iseq = RubyVM::InstructionSequence.compile(‘num = 1 + 2’) #=> <RubyVM::InstructionSequence:<compiled>@<compiled>> iseq.label #=> “<compiled>”

Using ::compile_file:

# /tmp/method.rb def hello

puts "hello, world"

end

# in irb > iseq = RubyVM::InstructionSequence.compile_file(‘/tmp/method.rb’) > iseq.label #=> <main>



1525
1526
1527
1528
1529
# File 'iseq.c', line 1525

static VALUE
iseqw_label(VALUE self)
{
    return rb_iseq_label(iseqw_check(self));
}

#marshal_dumpObject (private)

#marshal_loadObject (private)

#pathObject

Returns the path of this instruction sequence.

<compiled> if the iseq was evaluated from a string.

For example, using irb:

iseq = RubyVM::InstructionSequence.compile(‘num = 1 + 2’) #=> <RubyVM::InstructionSequence:<compiled>@<compiled>> iseq.path #=> “<compiled>”

Using ::compile_file:

# /tmp/method.rb def hello

puts "hello, world"

end

# in irb > iseq = RubyVM::InstructionSequence.compile_file(‘/tmp/method.rb’) > iseq.path #=> /tmp/method.rb



1474
1475
1476
1477
1478
# File 'iseq.c', line 1474

static VALUE
iseqw_path(VALUE self)
{
    return rb_iseq_path(iseqw_check(self));
}

#to_aArray

Returns an Array with 14 elements representing the instruction sequence with the following data:

magic

A string identifying the data format. Always YARVInstructionSequence/SimpleDataFormat.

major_version

The major version of the instruction sequence.

minor_version

The minor version of the instruction sequence.

format_type

A number identifying the data format. Always 1.

misc

A hash containing:

:arg_size

the total number of arguments taken by the method or the block (0 if iseq doesn’t represent a method or block)

[+:local_size+]

the number of local variables + 1

[+:stack_max+]

used in calculating the stack depth at which a SystemStackError is thrown.

#label

The name of the context (block, method, class, module, etc.) that this instruction sequence belongs to.

<main> if it’s at the top level, <compiled> if it was evaluated from a string.

#path

The relative path to the Ruby file where the instruction sequence was loaded from.

<compiled> if the iseq was evaluated from a string.

#absolute_path

The absolute path to the Ruby file where the instruction sequence was loaded from.

nil if the iseq was evaluated from a string.

#first_lineno

The number of the first source line where the instruction sequence was loaded from.

type

The type of the instruction sequence.

Valid values are :top, :method, :block, :class, :rescue, :ensure, :eval, :main, and plain.

locals

An array containing the names of all arguments and local variables as symbols.

params

An Hash object containing parameter information.

More info about these values can be found in vm_core.h.

catch_table

A list of exceptions and control flow operators (rescue, next, redo, break, etc.).

bytecode

An array of arrays containing the instruction names and operands that make up the body of the instruction sequence.

Note that this format is MRI specific and version dependent.

Returns:



1656
1657
1658
1659
1660
1661
# File 'iseq.c', line 1656

static VALUE
iseqw_to_a(VALUE self)
{
    const rb_iseq_t *iseq = iseqw_check(self);
    return iseq_data_to_ary(iseq);
}

#to_binary(extra_data = nil) ⇒ Object

Returns serialized iseq binary format data as a String object. A corresponding iseq object is created by RubyVM::InstructionSequence.load_from_binary() method.

String extra_data will be saved with binary data. You can access this data with RubyVM::InstructionSequence.load_from_binary_extra_data(binary).

Note that the translated binary data is not portable. You can not move this binary data to another machine. You can not use the binary data which is created by another version/another architecture of Ruby.



3407
3408
3409
3410
3411
3412
# File 'iseq.c', line 3407

static VALUE
iseqw_to_binary(int argc, VALUE *argv, VALUE self)
{
    VALUE opt = !rb_check_arity(argc, 0, 1) ? Qnil : argv[0];
    return rb_iseq_ibf_dump(iseqw_check(self), opt);
}

#trace_pointsArray

Return trace points in the instruction sequence. Return an array of [line, event_symbol] pair.

Returns:



2459
2460
2461
2462
2463
2464
2465
2466
2467
2468
2469
2470
2471
2472
2473
2474
# File 'iseq.c', line 2459

static VALUE
iseqw_trace_points(VALUE self)
{
    const rb_iseq_t *iseq = iseqw_check(self);
    const struct rb_iseq_constant_body *const body = iseq->body;
    unsigned int i;
    VALUE ary = rb_ary_new();

    for (i=0; i<body->insns_info.size; i++) {
	const struct iseq_insn_info_entry *entry = &body->insns_info.body[i];
	if (entry->events) {
	    push_event_info(iseq, entry->events, entry->line_no, ary);
	}
    }
    return ary;
}