Dec 2021 Meeting

calendar_today December 09, 2021 confirmation_number 19 tickets discussed Source English

DevelopersMeeting20211209Japan

https://bugs.ruby-lang.org/issues/18346

DateTime and location

  • 2021/12/09 (Thu) 13:00-17:00 JST @ Online

Next Date

  • 2022/01/13 (Thu) 13:00-17:00 JST @ Online

Announce

About release timeframe

  • preview 2?
    • no concrete plan yet.

Ordinary tickets

[Feature #18364] Add GC.stat_pool for Variable Width Allocation (peterzhu2118)

  • This feature will allow accessing stats for memory pools through Ruby. This will be especially useful for the Variable Width Allocation feature.
  • Should we extend GC.stat to return stats for memory pools or create a new method such as GC.stat_pool?
  • Is GC.stat_pool a good name?

Discussion:

  • ko1: GC.stat returns a flat hash (Symbol to Numeric map) now. This proposal changes it so that the method returns a nested hash.
# current
pp GC.stat
#=> {:count=>22,
#    :heap_allocated_pages=>146,
#    :heap_sorted_length=>146,
#    ...
#   }

pp GC.stat_pool(0) # statistics of heap pool whose unit size is 0
#=> {:heap_allocated_pages=>146,
#    :heap_sorted_length=>146,
#    ...
#   }

# alternative API
pp GC.stat
#=> {:count=>21,
#    :heap_allocated_pages=>99,
#    ...
#    :size_pools=>
#     [
#      {:slot_size=>40,
#       :heap_allocatable_pages=>4,
#       :heap_eden_pages=>40,
#       :heap_eden_slots=>1234,
#       :heap_tomb_pages=>0,
#       :heap_tomb_slots=>0},
#      {:slot_size=>80,
#       :heap_allocatable_pages=>80,
#       :heap_eden_pages=>14,
#       :heap_eden_slots=>123,
#       :heap_tomb_pages=>0,
#       :heap_tomb_slots=>0}
#      ...
#     ],
#     ...
#   }
  • matz: Is stat pool specific to MRI?
  • ko1: Yes.
  • matz: Is there any portability concern then?
  • ko1: This method already returns drastically different things among versions. There should be no problem.
  • ko1: Also jruby already extends this method so that it returns more than just a number.
  • matz: If so I guess we can follow jruby?
  • ko1: That breaks our CAPI (which expects size_ts as values).
  • matz: Hmm.
  • mame: I wonder what “pool” means in this :stat_pool key.
  • matz: There is only one “heap” in a ruby object space; it is split into two “generation”s; and each generation are split into several “page”s.
  • aaron: Peter made multiple pools which are sized.
  • matz: Maybe we should just have better terminologies.
  • aaron: size groups maybe?
  • matz: Or “arena”.
  • aaron: size_heap makes sense to me because the data sturcture is the same.
ko1@aluminium:~$ gem-codesearch 'rb_gc_stat\('
/srv/gems/airbnb-ruby-prof-0.0.1/ext/ruby_prof/rp_measure_allocations.c:size_t rb_gc_stat(VALUE key);
/srv/gems/airbnb-ruby-prof-0.0.1/ext/ruby_prof/rp_measure_allocations.c:    return rb_gc_stat(total_alloc_symbol);
/srv/gems/gc_tracer-1.5.1/ext/gc_tracer/gc_logging.c:   out_sizet(logging->out, rb_gc_stat(sym));
/srv/gems/gc_tracer-1.5.1/ext/gc_tracer/gc_logging.c:    rb_gc_stat(ID2SYM(rb_intern("count")));
/srv/gems/gctools-0.2.4/ext/gcprof/gcprof.c:  rb_gc_stat(obj);
/srv/gems/gctools-0.2.4/ext/gcprof/gcprof.c:  rb_gc_stat(_gcprof.start);
/srv/gems/gctools-0.2.4/ext/oobgc/oobgc.c:      _oobgc.start.total_allocated_objects = rb_gc_stat(sym_total_allocated_objects);
/srv/gems/gctools-0.2.4/ext/oobgc/oobgc.c:      _oobgc.start.heap_tomb_pages = rb_gc_stat(sym_heap_tomb_pages);
/srv/gems/gctools-0.2.4/ext/oobgc/oobgc.c:        rb_gc_stat(sym_heap_swept_slots) +
/srv/gems/gctools-0.2.4/ext/oobgc/oobgc.c:        (rb_gc_stat(sym_heap_tomb_pages) * _oobgc.heap_obj_limit) -
/srv/gems/gctools-0.2.4/ext/oobgc/oobgc.c:        rb_gc_stat(sym_heap_final_slots);
/srv/gems/gctools-0.2.4/ext/oobgc/oobgc.c:  size_t curr = rb_gc_stat(sym_total_allocated_objects);
/srv/gems/gctools-0.2.4/ext/oobgc/oobgc.c:    if ((rb_gc_stat(sym_old_objects) >= rb_gc_stat(sym_old_objects_limit)*0.97 ||
/srv/gems/gctools-0.2.4/ext/oobgc/oobgc.c:        rb_gc_stat(sym_remembered_wb_unprotected_objects) >= rb_gc_stat(sym_remembered_wb_unprotected_objects_limit)*0.97) &&
/srv/gems/heapfrag-0.1.0/ext/heapfrag/heapfrag.c:  size_t total_pages = rb_gc_stat(sym_heap_allocated_pages);
/srv/gems/heapfrag-0.1.0/ext/heapfrag/heapfrag.h:size_t rb_gc_stat(VALUE hash_or_sym);
/srv/gems/ish_lib_manager-0.0.1/test/dummy/vendor/bundle/ruby/2.3.0/extensions/x86_64-linux/2.3.0-static/ruby-prof-0.15.9/gem_make.out:checking for rb_gc_stat()... yes
/srv/gems/ish_lib_manager-0.0.1/test/dummy/vendor/bundle/ruby/2.3.0/extensions/x86_64-linux/2.3.0-static/ruby-prof-0.15.9/mkmf.log:have_func: checking for rb_gc_stat()... -------------------- yes
/srv/gems/ish_lib_manager-0.0.1/test/dummy/vendor/bundle/ruby/2.3.0/gems/ruby-prof-0.15.9/ext/ruby_prof/rp_measure_allocations.c:size_t rb_gc_stat(VALUE key);
/srv/gems/ish_lib_manager-0.0.1/test/dummy/vendor/bundle/ruby/2.3.0/gems/ruby-prof-0.15.9/ext/ruby_prof/rp_measure_allocations.c:    return rb_gc_stat(total_alloc_symbol);
/srv/gems/journalist-0.0.3/ext/journalist/garbage_collection.c:  rb_gc_stat(cookie.gc);
/srv/gems/journalist-0.0.3/ext/journalist/garbage_collection.c:  rb_gc_stat(cookie.gc);
/srv/gems/rblineprof-0.3.7/ext/rblineprof.c:    .allocated_objects = rb_gc_stat(sym_total_allocated_objects)
/srv/gems/rhodes-7.4.1/platform/shared/ruby/gc.c:rb_gc_stat(VALUE key)
/srv/gems/rhodes-7.4.1/platform/shared/ruby/include/ruby/intern.h:size_t rb_gc_stat(VALUE);
/srv/gems/ruby-compiler-0.1.1/vendor/ruby/doc/NEWS-2.1.0:* rb_gc_stat() added. This allows access to specific GC.stat() values from C
/srv/gems/ruby-compiler-0.1.1/vendor/ruby/gc.c:rb_gc_stat(VALUE key)
/srv/gems/ruby-prof-1.4.3/ext/ruby_prof/rp_measure_allocations.c:    return (double)rb_gc_stat(total_allocated_objects_key);
/srv/gems/tunemygc-1.0.71/ext/tunemygc/extconf.rb:        f.puts "    record->#{k} = rb_gc_stat(sym_gc_stat[#{i}]);"

Conclusion:

  • matz: Let’s return nested data structure for Ruby.
  • matz: We can discuss about the key for this new value set.
  • ko1: Maybe we can also refine our CAPI.
  • ko1: What should be returned when VWA is off?
  • aaron: It doesn’t change the output.
  • ko1: Make sure 3.1 without VWA outputs the same thing as 3.0.

[Feature #18349] Let –jit enable YJIT on supported platforms (k0kubun)

  • Any concerns? Can we do this from Ruby 3.1?

Discussion:

  • shyouhei: Is --jit still experimental?
  • matz: I don’t remember I declared it is mature.
  • matz: I have no reason to stop this proposal, given this is requested by @k0kubun.
  • n0kada: What happens when there is no YJIT for the platform?
  • naruse: I think it’s better to kill JIT then.
  • matz: Users can be explicit by passing --mjit/--yjit if they really need specific one.
  • naruse: Understood.
  • hsbt: --yjit will be kept?
  • matz: yes

Conclusion:

  • matz: Leave it to @k0kubun.
  • naruse: okay for Ruby 3.1

[Feature #18351] Support anonymous rest and keyword rest argument forwarding (jeremyevans0)

  • I think this is a natural addition after the addition of anonymous block forwarding in #11256.
  • Is it OK to support this syntax? If so, is the patch acceptable?

Preliminary discussion:

def foo(*)
  bar(*)
end

def baz(**)
  quux(**)
end
  • mame: I am not sure if this syntactic complexity is worth or not
  • mame: I am keen to remove anonymous block forwarding :-)
  • ko1: Is this so useful than ... forwarding?

Discussion:

  • matz: (Reading the pull request…)
  • matz: I think it’s possible.
  • ko1: For 3.1?
  • matz: That’s the point. It’s already December and it’s too late for 3.1.
  • knu: Maybe you want to write ary = [*] but it seems it is not covered in this implementation. (cf. Lua’s ...)
  • matz: That seems confusing.
  • nobu: You could use Array.new for that.
  • knu: But you can pass * to []/[]= because they are method calls, right?
def foo(*)
  [*]
end
p foo(1, 2, 3)

def foo(*)
  p(*, *)
end

foo(1, 2) #=> 1, 2, 1, 2

def bar(*, **)
  foo(*, **)
  foo(*, *, **)
end

def foo *
  bar
end
#=>
# 3.0 and earlier
# and after this proposal
def foo(*bar)
end

def foo(*)
  bar *
  baz
end
#=>
def foo(*)
  bar * baz
end

Conclusion:

  • matz: Interesting. Accepted for 3.2. Wait for a while.

[Feature #16663] Add block or filtered forms of Kernel#caller to allow early bail-out (jeremyevans0)

  • Two possible method names are find_caller and each_caller.
  • find_caller optimizes for the case where the first matching frame will be returned.
  • each_caller is more general, always iterating unless there is a non-local exit (break/return).
  • Are either of these names acceptable? If so, which method name should be used?

Discussion:

  • mame: New name candidates are propsed.
  • ko1: Can we add new method into Kernel for this purpose? (I’m not sure it’s so important). I prefer to extend caller if possible.
# Kernel methods
find_caller do |frame|
  frame.path =~ /.../
end #=> frame or nil

each_caller do |frame|
  break frame if frame.path =~ /.../
end #=> frame or ???

each_caller do |frame|
  frame.path =~ /.../
end #=> void

callers.each do |frame|
  # ...
end

# Thread class methods
Thread::Backtrace.each_caller do |frame|
end

Thread.each_caller do |frame|
end

Thread.callers.each do |frame|
end

# Thread instance methods
Thread.current.each_caller do |frame|
end
other_thread.each_caller do |frame| # may read frames from the other thread
end
  • knu: each_caller is confusing to me becasue what is returned from the method is not that obvious. What you need may be callers that returns an enumerable object.
  • nobu: Maybe we can add this to Thread::Backtrace instead of Kernel.
  • matz: each_something is not a good name because Enumerable#each’s block is not a filter. find_something sounds like it stops after first truthy block evaluation.
  • akr: What the OP wants is actually the each semantics, though.
  • ko1: how about keyword parameter like:
caller(filter: -> {|path| /pat/ =~ path})
  • matz: I don’t think it’s a wise idea for a method to return different types depending on its keyword arguments.

Conclusion:

  • matz: No conclusion now.

[Feature #12084] Class#instance (jeremyevans0)

  • This should be simple to implement, since a singleton class keeps a reference to the instance.
  • Are we OK adding this feature? If so, is the method name acceptable, or do we want singleton_instance?
  • If called on a non-singleton class, should TypeError be raised?
  • For TrueClass, FalseClass, and NilClass, should it return true, false, and nil?

Discussion:

Array.singleton_class.instance # => Array
"foo".singleton_class.instance # => "foo"
  • ko1: Forget about edge cases first.
    • Do we want this?
    • Name?
  • ko1: This is easy to do in C but fairly difficult to do in Ruby currently. One need to scan the entire ObjectSpace.
  • mame: There seems no practical use-case…

Conclusion:

  • matz: I don’t think it is needed.

[Bug #14891] Pathname#join has different behaviour to File.join (jeremyevans0)

  • I agree with @znz that this behavior is expected.
  • Can this be closed?

Preliminary discussion:

  • nobu: Go ahead

Discussion:

  • mame: What do you mean by “go ahead”, nobu?

Conclusion:

  • nobu: Already closed
  • akr: Maybe we can improve documents.

[Bug #14434] IO#reopen fails after EPIPE (jeremyevans0)

  • I’m not sure whether this behavior is a bug.
  • The behavior has existed since Ruby 1.9 (no failure in 1.8.7).
  • Can we decide whether this behavior is expected or a bug?

Discussion:

  • akr: So the IO is failing to flush.
  • mame: The patch tries to ignore errors.
  • nobu: … But the patch breaks our CI.

Conclusion:

  • nobu: I think it is a bug but maybe we cannot fix it soon, by Ruby 3.1

[Feature #18370] Call Exception#full_message to print exceptions reaching the top-level (eregon)

  • Improve consistency in how exceptions are shown, and makes it easier to evolve exception formatting.
  • OK to do?

Preliminary discussion:

  • mame: I am not sure if a consistency is a good enough reason to change this, but I’m okay if matz is okay

Discussion:

  • mame: This request is unrelated to did_you_mean/error_highlight according to eregon. The OP wants consistency.
  • matz: We need a clear benefit for this.
  • akr: I think this is a bug of full_message, it should escape the message too (if the escaping behavior is really needed)

Conclusion:

  • matz: Show us clear benefits of this, so that I can :+1: to this idea.

[Feature #18367] Stop the interpreter from escaping error messages (mame)

  • mame: Strongly related to the previous topic
  • mame: TBH I’m a bit afraid if raise "\0" can show NUL character to the terminal

Discussion:

  • akr: Do we ignore the reason we introduced the escapes?
  • mame: Other languages consider this as application matters.
  • shyouhei: That means they insist for instance docker(1) should escape its containers’ outputs?
  • mame: I guess the “application” here means their scripts.
  • naruse: That is harsh. That forces everyone to properly escape their puts arguments, which is not a practical requirement.
  • akr: Yes.. This request is moving the accountability demarcation point.
$ ruby -e '"\\".no_method'
-e:1:in `<main>': undefined method `no_method' for "\\\\":String (NoMethodError)

"\\".no_method
   ^^^^^^^^^^
  • naruse: Why do we have to escape backslashes at the first place?
  • mame: If the escape is removed, raise "\\x00" and raise "\x00" would be undistinguishable, nobu said
raise "\\x00" #=> test.rb:1:in `<main>': \x00
raise "\x00"  #=> test.rb:1:in `<main>': \x00
              #                          ^^^^ highlighted by ANSI escape sequence

raise "\e[31mRed\e[0m"
  • akr: Are recent terminals vulnerable? If not, the escaping behavior can be removed
    • xterm WindowsTerminal iterm2 libvte mlterm attri
  • matz: My gut feeling is we might not need to escape.

Conclusion:

  • No change for 3.1, either way.
  • Research recent terminal emulators in 3.2.
  • SyntaxError contains many info often.
$ ruby -e 'eval("def foo(nil = 1)")'
Traceback (most recent call last):
        1: from -e:1:in `<main>'
-e:1:in `eval': (eval):1: syntax error, unexpected `nil', expecting ')' (SyntaxError)
def foo(nil = 1)
        ^~~
(eval):1: Can't assign to nil
def foo(nil = 1)
        ^~~
(eval):1: syntax error, unexpected ')', expecting end-of-input
def foo(nil = 1)
               ^

$ ruby -e 'begin; eval("def foo(nil = 1)"); rescue SyntaxError; p $!.message; end'
"(eval):1: syntax error, unexpected `nil', expecting ')'\ndef foo(nil = 1)\n        ^~~\n(eval):1: Can't assign to nil\ndef foo(nil = 1)\n        ^~~\n(eval):1: syntax error, unexpected ')', expecting end-of-input\ndef foo(nil = 1)\n               ^\n"
  • SyntaxError#message #=> “syntax error”
  • SyntaxError#additional_message(highlight: true) #=> “(eval):1:…”
-e:1:in `eval': syntax error (SyntaxError)     # #message part
| (eval):1:...                                 # #additional_message part

[Feature #17881] Add a Module#const_added callback (byroot)

  • Would allow the Zeitwerk autoloader to no longer use tracepoint.
  • The main problem with Zeitwerk using tracepoint is that since tracepoint callbacks are not “reentrant” Zeitwerk is partially broken when using a debugger, e.g. https://github.com/ruby/debug/issues/408
  • This cause lots of confusion for users.
  • Could alternatively be Module#module_defined, to only be called for Module/Class

Preliminary discussion:

  • ko1: const_added can be overwritten by some gems. Is it really okay for Zeitwerk usecase? I will ask if the proposed API is really useful for Zeitwerk use case

Discussion:

  • ko1: If we had this hook Zeitwerk can get rid of trace points.
  • mame: Why we want to avoid trace point is debugger also uses it.
  • ko1: … And JIT breaks for trace pont.
  • ko1: Basically trace point shall be used for traces. Zeitwerk’s usage is kind of abuse.
  • akr: This can be used to autoload ::A::B::C, in case ::A::B is not loaded yet.
  • ko1: If Zeitwerk want to use it every one of modules must interface well with its const_defined hook.
  • akr: What has been done for ther hooks?
  • ko1: Maybe nobody cared?
  • nobu: I can try extending autoload to support nested ones, but not in time frame of 3.1.

Conclusion:

  • ko1: Let’s postpone it for 3.2
  • matz: Agreed. I accept the feature.

[Feature #15912] Allow some reentrancy during TracePoint event (byroot)

  • Alternative to Module#const_added for fixing the Zeitwerk and debuggers combination.
  • @ko1 proposed TracePoint#reopen(&block) so that debuggers can yield to the user with that.
  • Potentially with a list of reopened events: tp.reopen(:class) { "do stuff" }

Preliminary discussion:

  • ko1: TracePoint.allow_reentrance do ... end?

Discussion:

  • ko1: Trace points do not nest. But debuggers can evaluate in middle of Zeitwerk load, which doesn’t work.
  • ko1: Also breaks byebug.
  • ko1: The proposal explicitly allows a region of trace point call backs to fire other trace points.
  • ko1: Basically this is a dangerous operation (can e.g. loops indefinitely). This one is very explicit.
  • matz: For instance signal handlers can register itself for reentrancy.
  • akr: Doubt. Current POSIX manipulates signal masks for that purpose.
  • ko1: Unlike signal mask (which is per signal no), trace points blocks/allows everything at once.

Conclusion:

  • matz: One of TracePoint.allow_reentran{t,ce,cy} accepted.

[Feature #12737] Module#defined_refinements (shugo)

  • Is the name Module#refinements acceptable?
    • It’s consistent with Module#constants.
    • Other candidates: configured_refinements, contained_refinements, defined_refinements
  • Is it OK to return an Array instead of Hash?
    • The target class of a refinement can be obtained by another new method Refinement#refined_class.
  • I’d like to add Module#refinements and Refinement#refined_class after the release of Ruby 3.1.

Discussion:

  • mame: The OP comes to a new name.
module M
  refine String do
    $M_String = self
  end

  refine Integer do
    $M_Integer = self
  end
end

p M.refinements #=> [$M_String, $M_Integer]
p M.refinements[0].refined_class #=> String
p M.refinements[1].refined_class #=> Integer
  • matz: Got it.

Conclusion:

  • matz: Refinement#refined_class accepted.
  • matz: Module#refinements accepted.
  • shyouhei: Wait for 3.2.

[Feature #14332] Module.used_refinements to list refinement modules (shugo)

  • eregon suggested to add it soon, but It may be too late to add it in Ruby 3.1.

Discussion:

  • mame: What was the previous status of this issue?

Conclusion:

  • matz: Wait for 3.2.

[Feature #16038] Provide a public WeakMap that compares by equality rather than by identity (eregon)

  • This is necessary to have access to a weak-keys map on TruffleRuby/JRuby. There are many use cases:, notably this PR in Rails and all the reasons WeakHashMap is used in Java code.

Preliminary discussion:

  • mame: I’m unsure what is proposed. Moving ObjectSpace::WeakMap to ::WeakMap and adding WeakMap#compare_by_equality? Or introducing ObjectSpace::WeakKeysMap?
  • mame: I think I got it
    • JVM cannot implement WeakMap
    • So if it is used in Rails, JRuby/TruffuleRuby cannot support it well
    • eregon wants WeakKeysMap or WeakValuesMap because:
      • it is enough for Rails use case
      • it is possible to implement on JVM
  • mame: I feel it is a different story from the original issue
# use case: object deduplication
class Foo
  # @@cache = ObjectSpace::WeakKeysMap.new
  @@cache = ObjectSpace::WeakValuesMap.new
  @@cache.compare_by_equality

  def self.new(*args)
    @@cache[args] ||= super
  end
end

foo1 = Foo.new(1, 2, 3)
GC.start
foo2 = Foo.new(1, 2, 3)

# compare_by_equality is needed
p foo1.equal?(foo2) # satisfied by WeakValuesMap
                    # not necessarily satisfied by WeakKeysMap
  • mame: I don’t know this use case is the same as Rails use case
  • mame: What’s happen if we do hash[obj] = obj when hash is Java’s WeakHashMap? Is obj collected?

Discussion:

  • mame: Original proposal was to use WeakMap as object deduplication. In order to do so one have to use compare_by_identity now. The OP wanted to use equality method.
  • mame: @matz understood the motivation but was not sure if WeakMap was the right solution.
  • mame: Besides JVM languages have difficulties implementing their own WeakMap.
  • matz: Yes because a weak map in general has to interact with underlying GC; which is what they cannot.
  • mame: According to @eregon there is no weak-key weak-value map for JVM; there are only weak-key or weak-value map.
  • akr: It sounds what is needed is a weak set. A weak key map should suffice.
  • mame: That works only when the registered keys themselves can be queried.
  • ko1: It seems Java’s WeakHashMap lacks that feature.

Conclusion:

  • matz: Let me summarize and have a feedback.

[Feature #18033] Time.new or Time.at with string (nobu)

  • Time.new("2021-12-09") or Time.at("2021-12-09").
  • Other new methods: Time.try_convert or Time.iso.

Preliminary discussion:

[master]$ ruby -ve 'p Time.new("2021-12-09")'
ruby 3.0.2p107 (2021-07-07 revision 0db68f0233) [x86_64-linux]
2021-01-01 00:00:00 +0900

[master]$ ruby -ve 'p Time.new("2021-12-09")'
ruby 3.1.0dev (2021-11-26T02:39:59Z master 7f7c3a0a75) [x86_64-linux]
<internal:timev>:310:in `initialize': invalid value for Integer(): "2021-12-09" (ArgumentError)

[master]$ ruby -ve 'p Time.new("2021-12-09")'
ruby 3.X.0 (nobu's patch)
2021-12-09 00:00:00 +0900

ko1@aluminium:~$ gem-codesearch '\bTime\.new\(\"'
/srv/gems/cosmos-4.5.1/lib/cosmos/tools/cmd_tlm_server/limits_groups_background_task.rb:    FUTURE_TIME = Time.new("3000")
/srv/gems/cosmos-4.5.1/lib/cosmos/tools/cmd_tlm_server/limits_groups_background_task.rb:    PAST_TIME = Time.new("1900")
/srv/gems/feldtruby-0.4.12/lib/feldtruby/optimize/stdout_logger.rb:                     @last_report_time = Hash.new(Time.new("1970-01-01")) # Maps event strings to the last time they were reported on, used by anote.
/srv/gems/harvestime-0.0.3/lib/harvestime/invoice.rb:      Time.new("").to_date
/srv/gems/harvestime-0.0.3/spec/harvestime/invoice_spec.rb:        expectation = Time.new("").to_date
/srv/gems/logstash-input-multi-rds-0.0.1/lib/logstash/inputs/multirds.rb:          return Time.new("1999-01-01")
/srv/gems/logstash-input-rds-0.17.0/lib/logstash/inputs/rds.rb:          return Time.new("1999-01-01")
/srv/gems/marvell-0.1.1/spec/spec_helper.rb:    Time.new("2014-02-09 20:24:26 -0600")
/srv/gems/puppet-7.12.1/spec/unit/network/formats_spec.rb:        tm = Time.new("2016-01-27T19:30:00")
/srv/gems/roadforest-0.7/spec/rdfa-handler.rb:          @graph << [EX.a, EX.b, RDF::Literal::Time.new("12:34:56")]

ko1@aluminium:~$ gem-codesearch "\bTime\.new\(\'"
/srv/gems/digi_moji-0.0.3/lib/digi_moji/cli.rb:      t = Time.new('00:00:00') + time_in_sec(sec, unit)
/srv/gems/distribution-api-client-0.1.0/spec/card_url_spec.rb:    t = Time.new('2015-10-30T21:00:00+001')
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time1 = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time2 = Time.new('2014-01-02T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-0.1.0/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time1 = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time2 = Time.new('2014-01-02T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/fluent-plugin-mixpanel-enchanced-0.0.13/test/plugin/test_out_mixpanel.rb:    time = Time.new('2014-01-01T01:23:45+00:00').to_i
/srv/gems/frodata-0.9.3/spec/frodata/properties/time_spec.rb:  let(:subject) { FrOData::Properties::Time.new('Timely', '21:00:00-08:00') }
/srv/gems/frodo-0.12.8/spec/frodo/properties/time_spec.rb:  let(:subject) { Frodo::Properties::Time.new('Timely', '21:00:00-08:00') }
/srv/gems/grooveshark-0.2.14/spec/grooveshark/user_spec.rb:    expect(user.feed(Time.new('20141122')))
/srv/gems/logstash-output-s3-leprechaun-fork-1.0.15/spec/outputs/s3_spec.rb:      allow(Time).to receive(:now) { Time.new('2015-10-09-09:00') }
/srv/gems/looky-lu-0.0.5/spec/generators/state_generator_spec.rb:    time = Time.new('2013-05-10 19:58:31 +0000')
/srv/gems/odata-0.6.18/spec/odata/properties/time_spec.rb:  let(:subject) { OData::Properties::Time.new('Timely', '21:00:00-08:00') }
/srv/gems/odata4-0.9.1/spec/odata4/properties/time_spec.rb:  let(:subject) { OData4::Properties::Time.new('Timely', '21:00:00-08:00') }
/srv/gems/picky-4.31.3/spec/lib/loggers/verbose_spec.rb:        Time.stub :now => Time.new('zeros')
/srv/gems/picky-4.31.3/spec/lib/loggers/verbose_spec.rb:        Time.stub :now => Time.new('zeros')
/srv/gems/puppet-7.12.1/spec/unit/util/storage_spec.rb:          { :checked => Time.new('2018-08-08 15:28:25.546999000 -07:00') }
/srv/gems/ruby-cldr-0.5.0/README.textile:       format   = Cldr::Format::Time.new('HH:mm:ss z', calendar)
/srv/gems/ruby-cldr-0.5.0/test/format/datetime_test.rb:    time   = Cldr::Format::Time.new('HH:mm', @calendar)
/srv/gems/ruby-nmap-0.10.0/spec/scanner_spec.rb:    let(:start_time) { Time.new('Sat Jul 20 23:55:27 2013') }
/srv/gems/rubycom-0.4.3/test/rubycom/arg_parse_test.rb:    time = Time.new('2013-05-08 00:00:00 -0500')
/srv/gems/saper-0.5.3/spec/items/time_spec.rb:      expect { Saper::Items::Time.new('') }.to raise_error Saper::InvalidItem
/srv/gems/scraperd-0.0.3/test/activity_test.rb:              :pubDate => Time.new('2015-03-14 19:45:50 +0100'),
/srv/gems/scraperd-0.0.3/test/activity_test.rb:    assert_equal activity.added_at, Time.new('2015-03-14 19:45:50 +0100')
/srv/gems/scrapperd-0.0.1/test/activity_test.rb:              :pubDate => Time.new('2015-03-14 19:45:50 +0100'),
/srv/gems/scrapperd-0.0.1/test/activity_test.rb:    assert_equal activity.added_at, Time.new('2015-03-14 19:45:50 +0100')
/srv/gems/wavefront-client-3.6.2/spec/wavefront/cli/batch_write_spec.rb:    expect(k.valid_timestamp?(Time.new('1999-12-31').to_i)).to be false
/srv/gems/zentool-0.8.0/lib/zentool/zendesk_ticket.rb:    @start_time = Time.new('2016-01-01').to_i
  • Or Kernel#Time, Time#of?

Discussion:

  • nobu: @eregon wants iso8601 because Time.new already takes a string.
  • nobu: In fact takig non-integer string is not accepted already.
zsh % nocorrect sudo docker run -i --rm rubylang/all-ruby env ALL_RUBY_SINCE=1.9.0 LC_ALL=C.UTF-8 ./all-ruby -e 'p Time.new("2021-12-09")'
ruby-1.9.0-0        -e:1:in `initialize': wrong number of arguments(1 for 0) (ArgumentError)
                        from -e:1:in `new'
                        from -e:1:in `<main>'
                exit 1
...
ruby-1.9.1-p431     -e:1:in `initialize': wrong number of arguments(1 for 0) (ArgumentError)
                        from -e:1:in `new'
                        from -e:1:in `<main>'
                exit 1
ruby-1.9.2-preview1 2021-01-01 00:00:00 +0000
...
ruby-3.0.2          2021-01-01 00:00:00 +0000
ruby-3.1.0-preview1 <internal:timev>:306:in `initialize': invalid value for Integer(): "2021-12-09" (ArgumentError)
                        from -e:1:in `new'
                        from -e:1:in `<main>'
                exit 1
  • shyouhei: That sounds NG to me.
  • ko1: Looking at the code search it seems there are use cases Time.new to take a String. They are mostly tests.
  • mame: Should we
    • Revert the behaviour to 3.0?
    • Keep the current 3.1 implementation?
    • Merge nobu’s patch?
  • akr: I have some code concerns about nobu’s patch. Too many optional time components.
  • akr: It seems the current behavior is a bug fix to me.
  • knu: I agree. The month/mday and the time part should not be optional. Only sub-sec part and maybe timezone part should be optional. I also agree that the change fro m to_i to Integer() should be considered a bug fix.

Conclusion:

  • matz: Not this time.

[Feature #18190] Split Random::Formatter from securerandom (nobu)

Preliminary discussion:

Conclusion:

  • matz: Accepted

[Feature #18183] make SecureRandom.choose public (nobu)

  • How about Random::Formatter#random_string if choose is not a good name
SecureRandom.choose([*"1".."3"], 5) #=> 12312 or 23131 or something
SecureRandom.choose([*"A".."C"], 5) #=> ABCBA or BCACA or something

SecureRandom.random_string([*"1".."3"], 5) #=> 12312 or 23131 or something
SecureRandom.random_string([*"A".."C"], 5) #=> ABCBA or BCACA or something

ALPHANUMERIC = [*'A'..'Z', *'a'..'z', *'0'..'9']
def alphanumeric(n=nil)
  n = 16 if n.nil?
  choose(ALPHANUMERIC, n)
end

Conclusion:

  • matz: I don’t like the name, random_string

[Feature #18181] Introduce Enumerable#min_with_value, max_with_value, and minmax_with_value (ko1)

%w(abcde fg hijk).lazy.map {|str| [str, str.size] }.min_by {|str, size| size }
#=> ["fg", 2]

Conclusion:

  • matz: A better name is required

[Feature #18397] Remove documentation that Qfalse == 0 in extension.rdoc, instead encourage use of RTEST

Conclusion:

  • matz: Accepted as a documentation change