Nov 2018 Meeting

calendar_today November 22, 2018 confirmation_number 24 tickets discussed Source English

DevMeeting-2018-11-22

Please comment your favorite ticket numbers you want to ask to discuss with your SHORT comment or summary. (your summary/comment will help us because we don’t need to read all of ticket comments)

DO NOT discuss then on this ticket, please.

Date: 2018/11/22 (Thu) Time: 14:00-18:00 (JST) Place: pixiv Inc. (Tokyo, Japan) Sign-up: https://connpass.com/event/105368/

NOTES

  • Dev meeting IS NOT a decision making place. All decisions should be done at the bug tracker.
  • Dev meeting is a place we can ask Matz, nobu, nurse and other developers directly.
  • Matz is a very busy person. Take this opportunity to ask him. If you can not attend, other attendees can ask instead of you (if attendees can understand your issue).
  • We will write a log about discussion to a file or to each ticket in English.
  • All activities are best-effort (keep in mind that most of us are volunteer developers).
  • The date, time and place is scheduled according to when/where we can reserve Matz’s time.

Agenda

Next dev-meeting

About 2.6 timeframe

Check security tickets

Carry-over from previous meeting(s)

From Attendees

  • [Feature #15230] RubyVM.resolve_feature_path (mame)
    • I’d like this feature to investigate what will be loaded by require(feature).
  • [Feature #15231] Remove Object#=~ (mame)
    • The method looks useless, and made a trouble at least for me. I’d like to hear opinions from other committers.
  • [Feature #11689] Add methods allow us to get visibility from Method and UnboundMethod object. (yui-knk)
    • I want to introduce this feature to help meta programming. I’d like to hear opinions from other committers.
  • [Feature #15286] Proposal: Add Kernel.#expand(*args) (aycabta)
  • [Feature #15289] Accept “target” keyword on TracePoint#enable (ko1)
  • [Feature #15287] New TracePoint events to support loading features (ko1)
  • [Bug #6087] How should inherited methods deal with return values of their own subclass? (mame)
    • 6 years ago, matz decided that class A < Array; end; A.new.flatten.class #=> A in 2.X, Array in 3.0. Just confirm: has the decision been still unchanged?

From non-attendees

  • [Bug #14127] CSV generating UTF-16LE encoded file without BOM (nobu)
    • Although this is not a bug of csv.rb, I’d suggest to enable “bom|” flag when writing instead. https://github.com/nobu/ruby/tree/bug/14127-bom-header
  • [Bug #15285] lambda return behavior regression from #14639 (nobu)
  • [Feature #15220] Adding OpenSSL 1.1.1 on Travis CI gcc-8 case (jaruga)
    • To detect an issue for the latest OpenSSL early and guarantee a Ruby version supporting a OpenSSL version.
  • [Feature #15301] Symbol#call, returning method bound with arguments (zverok)
  • [Feature #15302] Proc#with and Proc#by, for partial function application and currying (Ritchie Buitre)
    • Convenient methods for functional programming.
  • [Bug #14968] make all pipes and sockets non-blocking by default (normalperson)
  • [Feature #13618] Thread::Light for 2.6 [ruby-core:89900] (normalperson)
  • [Feature #15317] How to deal with obsolete property values in Unicode 11.0.0 (duerst)
    • A clear idea on this is needed to upgrade to Unicode 11.0.0.
  • [Feature #13890] Allow a regexp as an argument to ‘count’, to count more interesting things than single characters (duerst)
    • I find this missing regularly, and writing it by hand in Ruby is clumsy, so addition would be valuable.
  • [Feature #12698] Method to delete a substring by regex match (duerst)
  • [Feature #10548] remove callcc (Callcc is now going obsoleted. Please use Fiber.) (ioquatix)
    • It’s unofficially deprecated, unsupported in most implementations of Ruby, and not used very much. Do you think it’s a good idea to drop support by 3.0?
  • [Feature #15327] Proposal: Enable refinements to #respond_to? (osyo)
  • [Feature #15326] Proposal: Enable refinements to #public_send (osyo)
  • [Feature #15114] Symbol#to_proc does not work with refinements? (osyo)
    • I want to more refinements!
  • [Feature #15330] Proposal: autoload_relative (marcandre)
    • This feature is actually more useful than autoload and should be added to 2.6

Please comment your favorite ticket we need to discuss with the following format.

* [Ticket ref] Ticket title (your name)
  * your comment why you want to put this ticket here if you want to add.

Your comment is very important if you are no attendee because we can not ask why you want to discuss about it.

Example:

* [Feature #14609] `Kernel#p` without args shows the receiver (ko1)
  * I feel this feature is very useful and some people say :+1: so let discuss about this feature.

I don’t guarantee to put tickets in agenda if the comment violate the format (because it is hard to copy&paste).

Log

Next dev-meeting

  • 2018/12/12 (Wed) 13:30-17:00 pixiv 6F (Tokyo, Japan)

About 2.6 timeframe

  • RC1 will be released on the end of November
  • Code frozen around 12/20
  • 2.6.0 at 12/25

Check security tickets (confidencial)

(not logged)

Carry-over from previous meeting(s)

  • It seems no carry-over issue was there.

From Attendees

  • Mame: How about changing the meeting time? 14:00-18:00 → 13:00-17:00

  • OK but 12/12 is 13:30-.

[Feature #15144] Enumerator#chain (zverok)

  • A.k.a. Enumerator#+

  • knu posted an implementation

  • Matz: why not Enumerable#chain ?
  • Knu: definitely agreed!

[Bug #14127] generating UTF-16LE encoded file without BOM (nobu)

  • Although this is not a bug of csv.rb, I’d suggest to enable “bom ” flag when writing instead. https://github.com/nobu/ruby/tree/bug/14127-bom-header
  • Usa: can this be made for “r+” and / or “a”?
  • Nobu: the proposed patch writes bom only when the pos is zero.
  • Naruse: I want it to avoid writing bom twice, when the file already has one.
  • Knu: what should happen for StringIO…?
  • Knu: We need a concrete specification first as to when a BOM is written.
  • Ko1: file a new ticket for this.

[Feature #15230] RubyVM.resolve_feature_path (mame)

  • I’d like this feature to investigate what will be loaded by require(feature).

  • Mame: I want it because I am writing a static analyzer and that wants this kind of thing.
  • Shyouhei: I have concerns that this method can be absued.
  • Mame: true, but I don’t plan to abuse this.
  • Shyouhei: I know you don’t but once this is merged people should start using it anyways.
  • Usa: I think it’s a good feature in general; the proposal seems like a rough cut though.
  • Matz: I have no strong objection.

[Feature #15231] Remove Object#=~ (mame)

  • The method looks useless, and made a trouble at least for me. I’d like to hear opinions from other committers.

  • (nobody is against removal)
  • Ko1: NilClass#=~ should remain.
  • Mame: what about !~
  • Matz: why not delete it also?
  • Usa: does it bother?
  • Knu: people might have their own =~, so deleting !~ affect them.
  • Akira: it seems sequel defines =~ and not for !~

[Feature #11689] Add methods allow us to get visibility from Method and UnboundMethod object. (yui-knk)

  • I want to introduce this feature to help meta programming. I’d like to hear opinions from other committers.

  • Matz: the concept is not if a method is “visible” or not, but if a method is “callable” or not.  So I feel a bit NG to call it visibility.

[Feature #15286] Proposal: Add Kernel.#expand(*args) (aycabta)

  • Matz: I don’t like the name.
  • Ko1: only the name?
  • Matz: I don’t like the feature itself, neither.
  • Knu: I have a feeling that it should not call a method.
  • Knu: what about Binding#expand ?
  • Mrkn: { a: a } is much shorter than binding.expand(a) …
  • Naruse: tell us a more realistic use-case.

[Bug #15285] lambda return behavior regression from #14639 (nobu)

[Feature #15289] Accept “target” keyword on TracePoint#enable (ko1)

  • Matz: OK, do it yourself.

[Feature #15287] New TracePoint events to support loading features (ko1)

  • Matz: Ok except naming “loaded”. Koichi should name good name.

[Bug #6087] How should inherited methods deal with return values of their own subclass? (mame)

  • 6 years ago, matz decided that class A < Array; end; A.new.flatten.class #=> A in 2.X, Array in 3.0. Just confirm: has the decision been still unchanged?

  • Matz: It is not obvious in each case if the operation is collecting elements (i.e. making an Array object) or making an altered version of the container object.
  • Knu: select() is a good example.  It may be considered a method for obtaining a subset or a method that collects matching elements into an array.  Hash#select() changed to return a Hash because it seemed more useful than returning an alist.
  • Nobu: Methods defined by Enumerable all return an array.
  • Knu: Making all Array methods return an Array object would make it harder to define your own array-like container by subclassing Array.

[Feature #15317] How to deal with obsolete property values in Unicode 11.0.0 (duerst)

  • A clear idea on this is needed to upgrade to Unicode 11.0.0.

  • Martin: one property value has disappeared.
  • Naruse: program error is preferred here because we can be aware of such breakage before actually deploying the code.
  • Mame: what about issuing a warning instead?

[Feature #13890] Allow a regexp as an argument to ‘count’, to count more interesting things than single characters (duerst)

  • I find this missing regularly, and writing it by hand in Ruby is clumsy, so addition would be valuable.

[Feature #12698] Method to delete a substring by regex match (duerst)

From non-attendees

[Feature #15220] Adding OpenSSL 1.1.1 on Travis CI gcc-8 case (jaruga)

  • To detect an issue for the latest OpenSSL early and guarantee a Ruby version supporting a OpenSSL version.

  • Shyouhei: I had already added one.

[Feature #15301] Symbol#call, returning method bound with arguments (zverok)

  • Matz: I’m very faintly against it.

[Feature #15302] Proc#with and Proc#by, for partial function application and currying (Ritchie Buitre)

  • Convenient methods for functional programming.

  • Matz: which is which? (looks confused at the first glance)
  • Matz: I don’t like the name.
  • Ko1: Is the naming only concern?
  • Matz: Not against the feature but I feel the notation is not right.

[Bug #14968] make all pipes and sockets non-blocking by default (normalperson)

  • Shyouhei: AFAIK there is a problem on Windows.
  • Martin: Greg is working on it.
  • Shyouhei: Is mingw32 the same with MSVC?
  • Naruse: No, but in this case they are not that different because it is about kernel.
  • Shyouhei: Greg says it works for him.
  • Matz: so the timer thread is already gone?
  • Ko1: yes.
  • Shyouhei: Eric says that caused race condition around blocking regions.
  • Usa: so it is already backwards incompatible? No?
  • Matz: maybe we should try it and fix or back out as necessary?
  • Naruse: so Eric should commit it today.

[Feature #13618] Thread::Light for 2.6 ruby-core:89900

  • Matz: will respond.

[Feature #10548] remove callcc (Callcc is now going obsolete. Please use Fiber.) (ioquatix)

  • It’s unofficially deprecated, unsupported in most implementations of Ruby, and not used very much. Do you think it’s a good idea to drop support by 3.0?

  • Ko1: callcc has many issues.  We have some workarounds, but that must be insufficient.
  • Mame: I don’t think it’s worth deleting because nobody actually uses this feature in production, thus deletion has no practical benefit.
  • Matz: deleting codes that concern reentrancy might benefit for speed?
  • Usa: for 3x3?
  • Matz: may not make it 3x faster.

[Feature #15327] Proposal: Enable refinements to #respond_to? (osyo)

  • Matz: Mmm.
  • Ko1: when respond_to says false, we can send that method anyways?
  • Usa: yes.
  • Knu: there are implicit and explicit respond_to. Is it for explicit one?
  • Usa: yes.
  • Matz: respond_to does not interact with refinements right now is/was intentional because refinements are file-scope; it should be immediately obvious what method is callable or not at looking at the source code.
  • Ko1: what about respond_to_missing?
  • Matz: redefining respond_to_missing at a refinements should be against local rebinding.
  • Shyouhei: but I think we can undef a method inside of a refinement.
  • (everybody started thinking…)
  • Ko1: OK, let’s not think about it for a while.

[Feature #15326] Proposal: Enable refinements to #public_send (osyo)

  • Matz: this one makes sense. Accepted.

[Feature #15114] Symbol#to_proc does not work with refinements? (osyo)

  • I want to more refinements!

  • Matz: sounds like a bug?
  • assign nobu

[Feature #15330] Proposal: autoload_relative (marcandre)

  • This feature is actually more useful than autoload and should be added to 2.6

  • Matz: why people love autoload, why…?
  • Naruse: What is new in this issue is not why, but how much people love autoloads.
  • Ko1: Marc-Andre recently changed requires in stdlib to require_relative.  I guess it is natural for him to think about autoloads next.
  • Shyouhei: why not make a new method called autoload_relative, which is an alias of require_relative? :trollface:
  • Knu: It is understandable to me that this may cover most popular use cases of autoload.