[#98645] [Ruby master Misc#16933] DevelopersMeeting20200618Japan — mame@...

Issue #16933 has been reported by mame (Yusuke Endoh).

14 messages 2020/06/04

[#98663] [Ruby master Bug#16936] `make check TESTS="-n !/Foo#method/"` not skipping the test case — jaruga@...

Issue #16936 has been reported by jaruga (Jun Aruga).

13 messages 2020/06/05

[#98772] [Ruby master Bug#16959] Weakmap has specs and third-party usage despite being a private API — headius@...

Issue #16959 has been reported by headius (Charles Nutter).

13 messages 2020/06/12

[#98826] [Ruby master Feature#16963] Remove English.rb from Ruby 2.8/3.0 — hsbt@...

Issue #16963 has been reported by hsbt (Hiroshi SHIBATA).

9 messages 2020/06/16

[#98920] [Ruby master Bug#16978] Ruby should not use realpath for __FILE__ — v.ondruch@...

Issue #16978 has been reported by vo.x (Vit Ondruch).

24 messages 2020/06/23

[#98947] [Ruby master Feature#16986] Anonymous Struct literal — ko1@...

Issue #16986 has been reported by ko1 (Koichi Sasada).

66 messages 2020/06/26

[#98964] [Ruby master Feature#16989] Sets: need ♥️ — marcandre-ruby-core@...

Issue #16989 has been reported by marcandre (Marc-Andre Lafortune).

33 messages 2020/06/26

[#98965] [Ruby master Feature#16990] Sets: operators compatibility with Array — marcandre-ruby-core@...

Issue #16990 has been reported by marcandre (Marc-Andre Lafortune).

11 messages 2020/06/26

[#98968] [Ruby master Feature#16993] Sets: from hash keys using Hash#key_set — marcandre-ruby-core@...

Issue #16993 has been reported by marcandre (Marc-Andre Lafortune).

10 messages 2020/06/26

[#98997] [Ruby master Feature#17000] 2.7.2 turns off deprecation warnings by deafult — mame@...

Issue #17000 has been reported by mame (Yusuke Endoh).

16 messages 2020/06/30

[ruby-core:98761] [Ruby master Feature#16955] A new mode `Warning[:keyword_into_rest_arg] = true` for 2.7

From: eregontp@...
Date: 2020-06-12 11:11:39 UTC
List: ruby-core #98761
Issue #16955 has been updated by Eregon (Benoit Daloze).


I think it can be useful to debug delegation issues,
and the killer feature is this will show the warning on the first conversion,
which means the warning if not a false positive will show the line to fix, and not having to guess which of the many callers forgot to use `ruby2_keywords`.

IMHO, `opt = args.pop if args.last.is_a?(Hash)` is very old style and old Ruby code, which would be much nicer as `def foo(*args, **kwargs)`.
Of course such code doesn't need to change, but I think it's OK to warn it in this "debug delegation mode".

I took a look at the first few warnings, and they all seem like they could all make better use of keyword arguments and are very old code.

Regarding duplicated warnings from generated code, how does it look like if we override Warning.warn to remove all duplicates?

----------------------------------------
Feature #16955: A new mode `Warning[:keyword_into_rest_arg] = true` for 2.7
https://bugs.ruby-lang.org/issues/16955#change-86119

* Author: mame (Yusuke Endoh)
* Status: Open
* Priority: Normal
----------------------------------------
## Problem

Please see #16954.  This ticket is for discussing the second solution.

## Proposal

The problem is caused by automatic conversion from keywords to positional arguments, especially, when the keyword is implicitly absorbed into rest arguments.

```ruby
def app_proxy(*args)
  # args.last is a normal (non-ruby2_keyword-flagged) Hash, { :k => 42 }, converted from the keyword "k: 42"
  ...
end

app_proxy(k: 42)     # this call is troublesome
```

So I propose a new mode `Warning[:keyword_into_rest_arg] = true` to notice such an event.

```ruby
1: Warning[:keyword_into_rest_arg] = true
2:
3: def app_proxy(*args)
4: end
5:
6: app_proxy(k: 42)

# t.rb:6: warning: The keyword argument is absorbed into the rest parameter
# t.rb:3: warning: The called method `app_proxy' is defined here
```

An experimental patch is [here](https://gist.githubusercontent.com/mame/d5fbddc94a4a02b708653d1948836bb9/raw/cb89be7eaa20eb57be8811d3617663d7459e80d7/keyword_into_rest_arg.patch).  (The new mode is enabled by default in the patch just to make the experiment easy.)

## Important note: false positives

Unfortunately, this brings many false positives.  Typically, the following code is completely innocent and will work great in 3.0, but is warned:

```ruby
1: def foo(*args)
2:   opt = args.pop if args.last.is_a?(Hash)
3:   ...
4: end
5:
6: foo(k: 42)

# t.rb:6: warning: The keyword argument is absorbed into the rest parameter
# t.rb:1: warning: The called method `foo' is defined here
```

By an experiment with `make test-all`, [tons of warnings are observed](https://gist.github.com/mame/d5fbddc94a4a02b708653d1948836bb9).  Note that duplicated warings are already filtered out.  (Some tests generates a code dynamically, which makes the duplication filter not work.)

Many warnings occur within standard libraries.  I think that it is possible to fix these warnings, but we need to introduce tons of fixes in stdlib of 2.7.2.  I'm afraid if it is acceptable.

However, @eregon said that it might be still usable with combination of a dedicated warning filter mechanism (such as [deprecation_toolkit](https://github.com/Shopify/deprecation_toolkit)).

## Discussion points

* Is this really helpful?
* We need to add this change into Ruby 2.7.2.  Is it acceptable?  (Matz had already agreed with this change.  But after that, it turned out that it brings so many false positives.)
* Is the name `:keyword_into_rest_arg` okay?
* Should the commandline option `-W:keyword_into_rest_arg` be added?




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

Unsubscribe: <mailto:ruby-core-request@ruby-lang.org?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>

In This Thread