Project

General

Profile

Bug #15118

Method [] & []= does not respect frozen_string_literal: true comment

Added by chopraanmol1 (Anmol Chopra) 10 days ago. Updated 6 days ago.

Status:
Closed
Priority:
Normal
Assignee:
-
Target version:
-
ruby -v:
ruby 2.6.0dev (2018-09-13 trunk 64736) [x86_64-linux]
[ruby-core:89003]

Description

Calling ["something"] on object or proc (non-hash aref implementation) does not respect frozen_string_literal: true comment

Script:

# frozen_string_literal: true

require 'benchmark'
require 'memory_profiler'
class NopId
  def self.[](str)
    str.__id__
  end

  def self.[]=(str, val)
    str.__id__
  end
end

SampleHash = {"sometext" => 0}

NopProc = proc{|a| a.__id__}

N = 1_000_000

def method1
  NopId["sometext"]
end

def method2
  SampleHash["sometext"]
end

def method3
  NopProc["sometext"]
end

def method4
  NopId["sometext"] = 'othertext'
end

def print_iseq method_name
  puts "-+"*20
  puts RubyVM::InstructionSequence.disasm method(method_name)
  puts "-+"*20
end

def print_memory_profiler title, &block
  puts "-+"*20
  puts title
  MemoryProfiler.report{N.times(&block)}.pretty_print(detailed_report: false, allocated_strings: 0, retained_strings: 0)
  puts "-+"*20
end


print_iseq :method1
print_iseq :method2
print_iseq :method3
print_iseq :method4

Benchmark.bm(10) do |bm|
  bm.report("method[]"){ N.times{ method1 } }
  bm.report("hash[]"){ N.times{ method2 } }
  bm.report("proc[]"){ N.times{ method3 } }
  bm.report("method[]="){ N.times{ method4 } }
end

print_memory_profiler("method[]"){method1}
print_memory_profiler("hash[]"){method2}
print_memory_profiler("proc[]"){method3}
print_memory_profiler("method[]="){method4}

Output:

-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
== disasm: #<ISeq:method1@../test_aref.rb:21 (21,0)-(23,3)> (catch: FALSE)
0000 getinlinecache               7, <is:0>                           (  22)[LiCa]
0003 getconstant                  :NopId
0005 setinlinecache               <is:0>
0007 opt_aref_with                "sometext", <callinfo!mid:[], argc:1, ARGS_SIMPLE>, <callcache>
0011 leave                                                            (  23)[Re]
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
== disasm: #<ISeq:method2@../test_aref.rb:25 (25,0)-(27,3)> (catch: FALSE)
0000 getinlinecache               7, <is:0>                           (  26)[LiCa]
0003 getconstant                  :SampleHash
0005 setinlinecache               <is:0>
0007 opt_aref_with                "sometext", <callinfo!mid:[], argc:1, ARGS_SIMPLE>, <callcache>
0011 leave                                                            (  27)[Re]
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
== disasm: #<ISeq:method3@../test_aref.rb:29 (29,0)-(31,3)> (catch: FALSE)
0000 getinlinecache               7, <is:0>                           (  30)[LiCa]
0003 getconstant                  :NopProc
0005 setinlinecache               <is:0>
0007 opt_aref_with                "sometext", <callinfo!mid:[], argc:1, ARGS_SIMPLE>, <callcache>
0011 leave                                                            (  31)[Re]
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
== disasm: #<ISeq:method4@../test_aref.rb:33 (33,0)-(35,3)> (catch: FALSE)
0000 getinlinecache               7, <is:0>                           (  34)[LiCa]
0003 getconstant                  :NopId
0005 setinlinecache               <is:0>
0007 putobject                    "othertext"
0009 swap
0010 topn                         1
0012 opt_aset_with                "sometext", <callinfo!mid:[]=, argc:2, ARGS_SIMPLE>, <callcache>
0016 pop
0017 leave                                                            (  35)[Re]
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
                 user     system      total        real
method[]     0.120000   0.004000   0.124000 (  0.121883)
hash[]       0.088000   0.000000   0.088000 (  0.088723)
proc[]       0.132000   0.000000   0.132000 (  0.133687)
method[]=    0.128000   0.000000   0.128000 (  0.126702)
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
method[]
Total allocated: 40000000 bytes (1000000 objects)
Total retained:  0 bytes (0 objects)


-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
hash[]
Total allocated: 0 bytes (0 objects)
Total retained:  0 bytes (0 objects)


-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
proc[]
Total allocated: 40000000 bytes (1000000 objects)
Total retained:  0 bytes (0 objects)


-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
method[]=
Total allocated: 40000000 bytes (1000000 objects)
Total retained:  0 bytes (0 objects)


-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

As you can observe calling NopClass["something"] & NopProc["something"] does not respect frozen_string_literal: true comment

Patch:

https://github.com/ruby/ruby/pull/1957

After Patch Result:

-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
== disasm: #<ISeq:method1@../test_aref.rb:21 (21,0)-(23,3)> (catch: FALSE)
0000 getinlinecache               7, <is:0>                           (  22)[LiCa]
0003 getconstant                  :NopId
0005 setinlinecache               <is:0>
0007 putobject                    "sometext"
0009 opt_aref                     <callinfo!mid:[], argc:1, ARGS_SIMPLE>, <callcache>
0012 leave                                                            (  23)[Re]
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
== disasm: #<ISeq:method2@../test_aref.rb:25 (25,0)-(27,3)> (catch: FALSE)
0000 getinlinecache               7, <is:0>                           (  26)[LiCa]
0003 getconstant                  :SampleHash
0005 setinlinecache               <is:0>
0007 putobject                    "sometext"
0009 opt_aref                     <callinfo!mid:[], argc:1, ARGS_SIMPLE>, <callcache>
0012 leave                                                            (  27)[Re]
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
== disasm: #<ISeq:method3@../test_aref.rb:29 (29,0)-(31,3)> (catch: FALSE)
0000 getinlinecache               7, <is:0>                           (  30)[LiCa]
0003 getconstant                  :NopProc
0005 setinlinecache               <is:0>
0007 putobject                    "sometext"
0009 opt_aref                     <callinfo!mid:[], argc:1, ARGS_SIMPLE>, <callcache>
0012 leave                                                            (  31)[Re]
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
== disasm: #<ISeq:method4@../test_aref.rb:33 (33,0)-(35,3)> (catch: FALSE)
0000 putnil                                                           (  34)[LiCa]
0001 getinlinecache               8, <is:0>
0004 getconstant                  :NopId
0006 setinlinecache               <is:0>
0008 putobject                    "sometext"
0010 putobject                    "othertext"
0012 setn                         3
0014 opt_aset                     <callinfo!mid:[]=, argc:2, ARGS_SIMPLE>, <callcache>
0017 pop
0018 leave                                                            (  35)[Re]
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
                 user     system      total        real
method[]     0.092000   0.004000   0.096000 (  0.094751)
hash[]       0.088000   0.000000   0.088000 (  0.087865)
proc[]       0.116000   0.000000   0.116000 (  0.117047)
method[]=    0.096000   0.000000   0.096000 (  0.096765)
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
method[]
Total allocated: 0 bytes (0 objects)
Total retained:  0 bytes (0 objects)


-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
hash[]
Total allocated: 0 bytes (0 objects)
Total retained:  0 bytes (0 objects)


-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
proc[]
Total allocated: 0 bytes (0 objects)
Total retained:  0 bytes (0 objects)


-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+
method[]=
Total allocated: 0 bytes (0 objects)
Total retained:  0 bytes (0 objects)


-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+

Associated revisions

Revision 64745
Added by nobu (Nobuyoshi Nakada) 9 days ago

Use opt_{aref,aset} over opt_{aref,aset}_with

  • compile.c (iseq_compile_each0): Use opt_aref/opt_aset over opt_aref_with/opt_aset_with when frozen_string_literal: true, not to resurrect the index string on non-Hash receiver.

[Fix GH-1957]

From: chopraanmol1 chopraanmol1@gmail.com

History

#1 Updated by chopraanmol1 (Anmol Chopra) 10 days ago

  • Description updated (diff)

#2 [ruby-core:89005] Updated by shevegen (Robert A. Heiler) 10 days ago

Interesting.

I just tested with this code:

#!/System/Index/bin/ruby -w
# Encoding: ISO-8859-1
# frozen_string_literal: true
# =========================================================================== #
NopProc = proc {|a| a }

def foobar
  result = NopProc["sometext"]
  result
end

x = foobar

puts x.class
puts x.frozen?

y = 'abc'
puts y.frozen?

# Result:
#   String
#   false
#   true

And I think you are right. I am testing .frozen? twice; the 'abc'
string is indeed frozen whereas the variant returned by [] does
not seem to honour the instruction in the "magic" comment section.

Very good catch. May I ask, for curiosity, how you discovered it?
I assume you may have tested somehow systematically or something?

#3 [ruby-core:89007] Updated by chopraanmol1 (Anmol Chopra) 9 days ago

shevegen (Robert A. Heiler) wrote:

Very good catch. May I ask, for curiosity, how you discovered it?
I assume you may have tested somehow systematically or something?

I was working on reducing memory allocation on roo gem. While profiling (with memory_profiler gem) I observed string literal being allocated multiple time at specific locations even after adding magic string literal comment. After debugging for while I discovered calling this method https://github.com/sparklemotion/nokogiri/blob/7b8cd0f5b15a926e92c869b450dd6f71cdd17b61/lib/nokogiri/xml/node.rb#L120 in cell_xml['r'] fashion resulted into above behavior.

#4 Updated by chopraanmol1 (Anmol Chopra) 9 days ago

  • Description updated (diff)

#5 Updated by chopraanmol1 (Anmol Chopra) 9 days ago

  • Subject changed from Method [] does not respect frozen_string_literal: true comment to Method [] & []= does not respect frozen_string_literal: true comment

#7 [ruby-core:89040] Updated by Hanmac (Hans Mackowiak) 7 days ago

nobu (Nobuyoshi Nakada) : should this ticket be closed or not yet?

it seems fixed in 64745

#8 [ruby-core:89045] Updated by nobu (Nobuyoshi Nakada) 6 days ago

  • Backport changed from 2.3: UNKNOWN, 2.4: UNKNOWN, 2.5: UNKNOWN to 2.3: REQUIRED, 2.4: REQUIRED, 2.5: REQUIRED
  • Status changed from Open to Closed
  • Description updated (diff)

Thank you, I missed the reference to this ticket in the commit log.

Also available in: Atom PDF