Aug 2019 Meeting

calendar_today August 29, 2019 confirmation_number 27 tickets discussed Source English

DevMeeting-2019-08-29

Please comment on 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 the ticket comments)

DO NOT discuss then on this ticket, please.

Date: 2019/08/29 (Thu) Time: 13:00-17:00 (JST) Place and Sign-up: https://connpass.com/event/139231/ log: https://docs.google.com/document/d/1XypDO1crRV9uNg1_ajxkljVdN8Vdyl5hnz462bDQw34/edit

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 the 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 are scheduled according to when/where we can reserve Matz’s time.

Agenda

Next dev-meeting

About 2.7 timeframe

Check security tickets

Discussion

Please comment on 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 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 this feature.

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

A short summary of a ticket is strongly recommended. We cannot read all discussion of the ticket in a limited time. A proposal is often changed during the discussion, so it is very helpful to summarize the latest/current proposal, post it as a comment in the ticket, and write a link to the comment.

Log

Next Date 9/19 (Fri) 13:00-17:00 @ moneyforward optional: 9/2 (Mon) 13:00-16:00 @ zoom Ann #release channel in ruby Slack mainly for release managers to progress their process About 2.7 timeframe preview 2 numbered parameter will be introduced pipeline operator is self-rejected foo\n# comment\n.bar should be fixed feature freeze (Nov.) Check security tickets Topics [Feature #11460] Unhelpful error message when naming a module with the same name as an existing class (nobu) Discussion: in short: “More helpful error message, please!” class A; end module A; end #=> TypeError: X is not a module # FILE:line: previous definition of X was here

matz: Go ahead. [Misc #15805] Let memory sizes of the various IMEMO object types be reflected correctly (methodmissing) Only the imemo_tmpbuf type’s auxiliary malloc heap buffer is factored into obj_memsize_of, other imemo types also alloc on the heap PR: https://github.com/ruby/ruby/pull/2140 Discussion: in short: “memsize_of of IMEMO objects is not precise!” already committed. done. [Misc #15806] Explicitly initialise encodings on init to remove branches on encoding lookup (methodmissing) Currently rb_enc_from_index and related methods explicitly check for encoding table init on each call - I think this should be initialised at boot time instead as even the simple case of loading the interpreter requires the encoding table to be loaded anyways (symbol.c) PR: https://github.com/ruby/ruby/pull/2128 Discussion: in short: “lazy initialization of rb_enc_init is really needed? nobu?” usa: It should not work on windows with Japanese path name. [Feature #12093] Eval InstructionSequence with binding (nobu) Discussion: ko1: do you want to allow ISeq#compile to accept a binding? Use case? Create another ticket. [Bug #11326] Defining a writer as a Struct member allowed? (jeremyevans0) Should we ban struct members ending in =? Do we also want to ban struct members ending in ! or ?? Discussion: in short: “Struct.new(:x=)” matz: accepted. usa: note that it’s incompatibility if we can do it now. but this does not mean I disagree this change, of course. [Feature #15588] String#each_chunk and #chunks (Glass_saga) name: #each_slice and #slices are better? size as characters or bytes should we support :strip parameter? Discussion: in short: “A feature to take every n chars” matz: It seem good to have this feature. I don’t like all names proposed so far (chunk* and slice*). shyouhei: I don’t see any use case. Without it, we cannot propose alternative names. matz: Indeed, use case is needed too.

usa: just idea, a use case for static length records

str.each_slice(7, 10, 20) do |zip, tel, name| … end

[Feature #15553] Addrinfo.getaddrinfo supports timeout (Glass_saga) I’d like to have some feedback for patch2.diff Discussion: in short: “A patch to make getaddrinfo interruptible, please review” usa: akr should review this, but he does not attend. Skip. [Misc #15723] Reconsider numbered parameters (sikachu) Reading from description, I think there are a few supporters of removing @1, @2 in favor of just having @ refer to the first argument. I also believe having just @ is enough for major use cases, and for the readability of the code when multiple implicit receivers are used. Proposal: Remove @1 and @2 and have only @ for the first argument in Ruby 2.7+ Discussion: in short: “single1 @” matz: I’d like to pick _1, _2, _3, and so on. _0 is equivalent to |_0|. Prohibit use of both _0 and _1 simultaneously. nobu: Will do. osyo: If there is a variable named _0, …? matz: Always warn such a use case. durest: is @ rejected? matz: Yes. ko1: _0 and _1 is sequential but different meaning(|_0| and |_1,|). it is confusing. But I have no counterproposal, though. Obsolete _0, … plan: local variables (parameters and assigned variable) → force warning (Don’t use) on Ruby 2.7 and syntax error on Ruby 3. method invocation vcall: x = _0 # expect _0() outside block. → force warning on Ruby 2.7 and syntax error on Ruby 3. vcall: 1.times{ _0 } → block parameter (incompatibility) vcall: 1.times{|i| _0 } → force warning on Ruby 2.7 and syntax error on Ruby 3. method invocation (x = _0(), x = foo._0) → no warning method name (def _0(); end) → no warning [Feature #15998] Allow String#-@ to deduplicate tainted string, but return an untainted one (ko1) taint removal schedule Discussion: in short: “When will taint be removed?” ko1: warn $SAFE = 1 and all relevant methods (taint, tainted?) remove the feature itself (and warn) shyouhei: setuid’ed ruby automatically sets $SAFE = 1. naruse: I created https://bugs.ruby-lang.org/issues/16131 [Feature #15868] Implement File.absolute_path? (nobu) Discussion: matz: go ahead [Feature #16029] Expose fstring related APIs to C-extensions (nobu) Discussion: ko1, nobu: now it is difficult to expose them because of current implementation. [Feature #15778] caller_locations(debug: true) to access bindings of the stack. (eregon) Can we introduce the API? If not, why not? If not, please propose a way to support such functionality on other Ruby implementations (e.g., JRuby, TruffleRuby). Discussion: in short: “Wants a legitimate way to get caller bindings (for debug)” ko1: I’m against introducing this feature because it would break caller-callee relationship; a callee should not depend on what the caller has in its scope. At the very least it should not be used in production code. matz: if JRuby developers accepts this, I accept this. But are they okay? I’d like eregon to confirm them. matz: I’m okay for debug. It represents debug-purpose clearly. [Feature #16035] Allow non-finalizable objects such as Integer, static Symbol etc in ObjectSpace::WeakMap (nobu) Discussion: ko1: It makes the object non-collectable. Is it intended? Need to confirm. matz: I think it seems needed. I’ll comment and accept. [Feature #16038] Provide a public WeakMap which compare by equality rather than identity (nobu) Discussion: matz: want OP to elaborate euse case [Bug #15908] Detecting BOM with non-UTF encoding (nobu) Discussion: in short: On Windows, often text files may be Unicode(UTF-16LE-BOM) or OEM(Codepage). Currently there is no easy way to open them properly. naruse: I understand there theorecally exists a situation this feature is useful. But I think in practice it doesn’t exist. If needs be, the user should have already managed to do that on their own. usa: I often use “BOMed UTF-8 or CP932” CSV file for Excel. But, it’s only generating. I’ve never written a script which reads such CSV file… [Feature #14781] Enumerator.generate (zverok) Matz seems to have liked the idea and name proposed; implementation patch provided; the name should be decided upon: either generate, or, maybe, produce (looking like more-or-less oppose to reduce) Discussion: in short: Enumerator.generate(1, &:succ).take(5), the names generate and produce are proposed knu: I think this is a nice feature to add, and the newly proposed name “produce” sounds good. I withdraw the name “from”, and the idea of optionally giving multiple seeds. knu: As I commented in the ticket, I’d like to add support for StopIteration so the proc itself can define an end without requiring user to always use take_while to stop the stream before getting a NoMethodError. matz: produce sounds reasonable. [Feature #14784] Comparable#clamp with a range (zverok) The issue generally agreed on, except for behavior with upper bound with excluding-end range. ArgumentError is proposed for this case. Discussion: in short: “Raising an exception is proposed for clamp(1…2)” 3 .clamp(1 …2 ) #=> ArgumentError 3r.clamp(1r…2r) #=> ArgumentError

knu: Kernel#rand(range) accepts an exclusive range, but that’s because rand() only deals with Float and Integer which always have a predecessor. (clamp() deals with Rational in addition to Float and Integer) matz: Accepted. [Feature #15815] Add option to raise NoMethodError for OpenStruct (mtsmfm) We can use Symbol#to_proc with enumerable stuff when we use OpenStruct instead of Hash. For example: OpenStruct.new(JSON.parse(users)).map(&:id) But it can’t prevent typo. For Hash, it has Hash#fetch to raise KeyError. Discussion: in short: “Raise NoMethodError for undefined key of OpenStruct” matz: seems reasonable. mame: OpenStruct.new(a: 1, exception: 1) should make a key exception for compatibility. (i.e. when giving an exception keyword argument, the first Hash argument is required) [Feature #16103] Make the dot-colon method reference frozen (nobu) Discussion: ko1: sounds reasonable matz: sounds reasonable [Feature #15864] Proposal: Add methods to determine if it is an infinite range (mrkn) I proposed some candidates from mathematical terms Discussion: in short: “Alternative name proposal of Range#endless?: upper_unbounded?, right_unbounded?, or unbounded_above?” matz: use case is needed. The proposed names have chance. [Feature #14183] “Real” keyword argument (mame) Jeremy and I agreed with Jeremy’s proposal. I’d like to ask for matz’s final confirmation. Proposal summary: keyword arguments are separated from positional arguments Always need to use foo(kw: 1) or foo(opt) to pass a keyword argument Always need to use def foo(kw: 1) or def foo(opt) to accept a keyword argument But for compatibility, “Jeremy’s hack” remains (in Ruby 3) Only when a method does not accept keyword arguments, and when a call passes keyword arguments, the keyword arguments was converted to a Hash object as the last argument def foo(opt = {}) end foo(kw: 1) # OK

Discussion: in short: “Jeremy made the branch of keyword separation ready. Let’s merge it!” matz: Go ahead. [Feature #15955] UnboundMethod#apply (mame) The ticket proposes a shortcut to unbound_method.bind(obj).call(args…) without allocation of a Method object. There are some use cases, and looks reasonable to me. What do you think? Discussion: in short: “Faster shorthand for unbound_method.bind(obj).call(args…). Use cases: pp, sorbet-runtime, zeitwerk” …naming discussion… mame: How about UnboundMethod#bind_call ? matz: Go ahead. [Feature #16115] Keyword arguments from method calls or ignore extra hash keys in splat (mame) It proposes a triple splat (foo(opt)) to pass a keyword hash with filtering out unknown keywords. What do you think? def foo(kw: 1) end opt = { kw: 2, unknown: 3 } foo(opt)

Discussion: in short: “A feature to ignore unknown keywords” ko1: The argument spec should be selected by one who define the method. I’m skeptical to allow mame: Adding a parameter to an existing method definition is casual when version up. But client code cannot use the new feature if it works with both old and new versions. amatsuda, naruse: Skeptical. ko1: Do you need a method to get the list of supported keywords of a method? Would it suffice? [Feature #15991] Allow questionmarks in variable names (osyo) Proposal: Allow ? in variable names In #5781, matz is explicitly against an instance variable that ends with ? How about allowing only local variables ? Discussion: in short: Allow foo? = 1 nobu: Seems to me it’s too difficult to implement it. It will likely conflict with the ternary operator, multiple assignment and so on. [Feature #16113] Partial application: fetch(urls).map(&JSON.:parse.w(symbolize_names: true)) (zverok) Discussion: matz: I don’t like it, will comment. many: As for this example, what’s wrong with the normal block notation? You could do it in much shorter code. [Feature #16122] Struct::Value: simple immutable value object (zverok) Discussion: in short: “immutable variant of Struct.” knu: def initialize(*); super.freeze; end is not enough? knu: Why not Enumerable? I’d write a plain old ruby class. mame: The API should be Struct.new(:a, :b, mutable: false) or what not.