DevMeeting-2024-10-03

https://bugs.ruby-lang.org/issues/20717

DateTime and location

2024/10/03 (Thu) 13:00-17:00 JST @ Online

Next Date

2024/11/07 (Thu) 13:00-17:00 JST @ Online

Announce

About release timeframe

Ordinary tickets

[Misc #20746] Request to migrate the `json` gem from `flori/json` repository to `ruby/json` (byroot)

Current maintainers don’t have full administrative rights over the repo, which is problematic.
We’ve been trying to contact flori to try to resolve this situation with a transfer for over two years now, without success.

Discussion:

https://github.com/flori/json/pull/595
https://github.com/flori/json/pull/597
https://github.com/flori/json/pull/598

Conclusion:

hsbt: I will merge byroot’s PRs

[Misc #20728] Add eileencodes as a core committer (tenderlovemaking)

She’s made many useful contributions
She’s currently helping to maintain Prism / Prism Compiler and it would be helpful if she can merge her own commits

Conclusion:

matz: will reply

[Feature #20750] Expose ruby_thread_has_gvl_p in ruby/thread.h (eregon)

I think rb_thread_call_with_gvl() should be allowed if the GVL is already acquired, it seems a much nicer way to solve the use cases the OP pointed out (for a C function which can be called either with or without GVL).
@ko1 What do you think, how do you want to solve that issue?

Discussion:

ko1: it’s okay to introduce ruby_thread_has_gvl_p() . But I’m a bit negative to allow rb_thread_call_with_gvl() on acquiring GVL. Maybe becasuse I don’t want to abuse this functioin. For example, use with_gvl for any fnctions. I think this kind of difference (obtaining GVL or not) should be self-conscious.
matz: This reason is not reasonable for me. How about to document about it?

Conclusion:

matz: will reply

[Feature #10459] [PATCH] rfc3339 method for Time (hsbt)

https://github.com/rails/rails/blob/b88d9af34fbc1c84ce2769ba02584eab2c28ac6e/activesupport/lib/active_support/time_with_zone.rb#L152
It seems reasonable.
We introduce that with C ext same as xmlschema?

Conclusion:

akr: rfc3339 is slightly different from iso8601. I don’t want to introduce this alias

[Feature #17295] Feature: Create a directory and file with Pathname#touch (hsbt)

I found same use case for that today. How about this?
- https://github.com/BerkeleyLibrary/util/blob/main/spec/berkeley_library/util/files_spec.rb#L16
https://github.com/ruby/ruby/pull/3706

Discussion:

shyouhei: this was rejected before https://bugs.ruby-lang.org/issues/7361

Conclusion:

akr: negative

[Feature #17297] Feature: Introduce Pathname.mktmpdir (hsbt)

How about this?
https://github.com/ruby/ruby/pull/3709

Conclusion:

akr: okay
hsbt: I will merge

[Feature #17294] Feature: Allow method chaining with Pathname#mkpath Pathname#rmtree (hsbt)

How about this?
https://github.com/ruby/ruby/pull/3705

Discussion:

akr: This allows method chaining, though it does not fit with my taste
akr: But now I undrstand the feeling and culture of method chaining, so should I accept this?
nobu: (path = Pathname.new("/tmp/new")).mkpath?
akr: It is too cryptic
knu: I understand the feeling of those who want to write this. I often have to write LOG_DIR = Pathname("…/log"); PID_DIR = Pathname("…/pids"); OUTPUT_DIR = Pathname("…/out") followed by mkpath for each.
mame: I tend to write this

path = Pathname.new("/tmp/new")
path.mkpath

knu: I can’t think of a use case of rmtree returning self. path.rmtree.mkpath(mode: 700) may make sense.
akr: well, maybe it is ok too?

Conclusion:

akr: go ahead for both mkpath and rmtree
hsbt: I will proceed this

[Feature #17296] Feature: Pathname#chmod use FileUtils.chmod instead of File (hsbt)

It’s reasonable and useful.
Does anyone have concern?
https://github.com/ruby/ruby/pull/3708

Discussion:

akr: understandable
akr: is it ok to return 1?
nobu: lchmod should be also handled
nobu: For symlink, File.chmod changes the link target file. But FileUtils.chmod changes the symlink itself (only when lchmod is available). https://github.com/ruby/fileutils/blob/master/lib/fileutils.rb#L2209
akr: it is too incompatible, unacceptable

Conclusion:

akr: the incompatibility is unacceptable
akr: If this is really wanted, we should improve File.chmod itself

Implement Pathname#chdir

Does anyone know the reason why removing chdir?

Discussion:

hsbt: it was obsoleted at Ruby 1.8 and removed at Ruby 1.9. Why?
akr: I cannot remember. I guess I disliked the process-global state
knu: If a pathname were a relative path, it would no longer point to the directory after calling chdir. A method call making the receiver invalid may sound strange.
mame: Now what do you think?
akr: Hmm… I don’t use chdir so much. When do users want to use chdir?
akr: I think it is not needed unless there is an explained use case

[GH-2974] Pathname#mountpoint? to support bind mount

akr: We need to check if the platform is Linux, and the spec of mountinfo.
znz: mountinfo includes /tmp/foo\040bar when the mount path includes a space
akr: We need to handle the escaping
znz: Here is the document: https://docs.kernel.org/filesystems/proc.html#proc-pid-mountinfo-information-about-mounts

[GH-2024] each_address should resolve for AAAA first

Does anyone have any concern?
akr: Looks good

[Feature #18127] Ractor-local version of Singleton (hsbt)

What’s the blocker for this?

Conclusion:

hsbt: it is already merged
matz: confirmed the spec, Okay

[Feature #6012] Proc#source_location also return the column (eregon)

How about {Proc,Method,UnboundMethod}.source_location #=> [path, start_line, start_column, start_offset, end_line, end_column, end_offset] ?
How about {Proc,Method,UnboundMethod}.code_location #=> object with path, start_line, start_column, start_offset, end_line, end_column, end_offset and code ?
It would avoid having to use RubyVM to find this information.
If some do not like these APIs, what do they suggest instead?

Discussion:

matz: adding column information is okay
matz: I am not sure if offset information is needed
matz: I am not sure about the incompatibility of changing the array size

Conclusion:

matz: {Proc,Method,UnboundMethod}.source_location #=> [path, start_line, start_column, end_line, end_column] looks good. If offset is really needed, please persuade me.

misc

https://bugs.ruby-lang.org/issues/15554 warn/error passing a block to a method which never use a block

akr: command-line argument is sometimes unuseful. Warning[...] is better
mame: If we want to pass the command-line argument to rake test, we need to write Rake::TestTask#options= accessor

require "rake/testtask"

Rake::TestTask.new(:test) do |t|
  t.libs << "test"
  t.libs << "lib"
  t.options = "--warn-unused-block"
  t.test_files = FileList["test/**/*_test.rb"]
end

https://bugs.ruby-lang.org/issues/20620 singleton_method undefined for module using “extend self”

matz: looks good, I will reply “go ahead”

https://bugs.ruby-lang.org/issues/20564 Switch default parser to Prism

matz: If any showstopper appears, we may revert. Currently, AFAIK, there is no showstopper. I will reply

https://bugs.ruby-lang.org/issues/20738 Removing a specific entry from a hash literal

func(
  foo: 1,
  bar: 2 if cond?,
  baz: 3,
)

matz: The following looks good enough. I don’t see the need of a new syntax

h = {}
h[:foo] = 1
h[:bar] = 2 if bar?
h

kw = {}
kw[:opt] = 1 if cond?
foo(**kw)

foo(
  opt: 1 if cond?
)

foo(
  ?opt: (cond? ? 1 : nil)
)

matz: I am not positive.

https://bugs.ruby-lang.org/issues/20745 `IO::Buffer#copy` triggers UB when src/dest buffers overlap

matz: I hope we can avoid undefined behavior unless there is a very special reason.

https://bugs.ruby-lang.org/issues/20769 Add `Hash#transform_value`

mutated_hash = hash.transform_value(:image) { |url| download(url) }
mutated_hash = { **hash, image: download(hash[:image]) }

h = { :a => 1, :b => 2, :c => 3 }
h.transform_value(:b) {|n| n * 100 } #=> { a: 1, b: 200, c: 3 }

matz: I don’t see the need.

h = { :a => 1, :b => 2, :c => 3 }
h2 = h.dup
h2[:b] = h2[:b] * 100

matz: I will reject

https://bugs.ruby-lang.org/issues/20770 A new pipe operator proposal

add(value, 3) |> 3.times { square(it) }
add(value, 3) |> self.square(it)
add(value, 3) |> mul(it, it)
add(value, 3) |> foo(&it)
add(value, 3) |> it.foo
add(value, 3) |> it * it
add(value, 3) |> "#{ it }"
add(value, 3) |> [it]
add(value, 3) |> x if it == 0
add(value, 3) |> foo(it) { bar }

matz: Instead of introducing then-specific syntax, I am rather interested in an one-line block syntax without a closing brace.

times |> foo(it)
times do =|x| foo(x)
times do = foo(x)
times do |x| = 10.times do |y| = foo(x, y)
times do |x|> 10.times do |y|> foo(x, y)

obj.map { foo(it) }.select { cond(it) }

add(value, 3).then |> square(it).then |> half(it)

add(value, 3) |> square(it) |> half(it)
add(value, 3).then { square(it) }.then { half(it) }

add(value, 3) => x
square(x) => x
half(x) => x

x = add(value, 3)
x = square(x)
x = half(x)

add(value, 3) => square(it) => half(it) # No

https://bugs.ruby-lang.org/issues/20778 ruby/repl_type_completor as a bundled gem

matz: Positive. I will reply

https://bugs.ruby-lang.org/issues/17326 Add Kernel#must! to the standard library

annotation?

Matz proposed annotation for Ruby 4 in EuRuKo. Let’s do brain storming

type annotation
DI（Java）
routing info for app web server (like Python’s flask)
mark a test function (like Rust)
mark a module that is only activated on a specific platform/configuration (like Rust)
documentation
memoization
synchronized
https://github.com/rails/thor/wiki
info for optimization (pure, noreturn, etc.)
rubocop:disable

def foo
  @_foo ||= begin
    ...
  end
end

@doc(<<END)
  これはこういうメソッドです
END
def foo ... end

DevMeeting-2024-10-03

DateTime and location

Next Date

Announce

About release timeframe

Ordinary tickets

[Misc #20746] Request to migrate the json gem from flori/json repository to ruby/json (byroot)

[Misc #20728] Add eileencodes as a core committer (tenderlovemaking)

[Feature #20750] Expose ruby_thread_has_gvl_p in ruby/thread.h (eregon)

[Feature #10459] [PATCH] rfc3339 method for Time (hsbt)

[Feature #17295] Feature: Create a directory and file with Pathname#touch (hsbt)

[Feature #17297] Feature: Introduce Pathname.mktmpdir (hsbt)

[Feature #17294] Feature: Allow method chaining with Pathname#mkpath Pathname#rmtree (hsbt)

[Feature #17296] Feature: Pathname#chmod use FileUtils.chmod instead of File (hsbt)

Implement Pathname#chdir

[GH-2974] Pathname#mountpoint? to support bind mount

[GH-2024] each_address should resolve for AAAA first

[Feature #18127] Ractor-local version of Singleton (hsbt)

[Feature #6012] Proc#source_location also return the column (eregon)

misc

https://bugs.ruby-lang.org/issues/15554 warn/error passing a block to a method which never use a block

https://bugs.ruby-lang.org/issues/20620 singleton_method undefined for module using “extend self”

https://bugs.ruby-lang.org/issues/20564 Switch default parser to Prism

https://bugs.ruby-lang.org/issues/20738 Removing a specific entry from a hash literal

https://bugs.ruby-lang.org/issues/20745 IO::Buffer#copy triggers UB when src/dest buffers overlap

https://bugs.ruby-lang.org/issues/20769 Add Hash#transform_value

https://bugs.ruby-lang.org/issues/20770 A new pipe operator proposal

https://bugs.ruby-lang.org/issues/20778 ruby/repl_type_completor as a bundled gem

https://bugs.ruby-lang.org/issues/17326 Add Kernel#must! to the standard library

annotation?

[Misc #20746] Request to migrate the `json` gem from `flori/json` repository to `ruby/json` (byroot)

https://bugs.ruby-lang.org/issues/20745 `IO::Buffer#copy` triggers UB when src/dest buffers overlap

https://bugs.ruby-lang.org/issues/20769 Add `Hash#transform_value`