From: "kddnewton (Kevin Newton) via ruby-core" <ruby-core@...>
Date: 2024-03-14T13:12:32+00:00
Subject: [ruby-core:117168] [Ruby master Feature#20331] Should parser warn hash duplication and when clause?

Issue #20331 has been updated by kddnewton (Kevin Newton).


Thank you for the response Matz. I just want to clarify what you're asking for the implementation to be, based on what you're saying (let me know if I misunderstand).

> parser should warn only minimal and apparent cases, no base conversion, no bignum allocation. Handling _ in the number literals is on the border.

I think you're saying that the parser should warn for `{ 1 => foo, 1 => bar }` but it should not warn for `{ 1 => foo, 0b1 => bar }`. But we don't want to remove the warnings, so the compiler will need to warn for those cases? I think in that case the compiler will need to build the hash, check for duplicates, and then only warn if they have different bases. Is that the desired behavior?

> I don't want to add complicated checks since prism may be used for small devices too (e.g., PicoRuby)

To be clear, prism already runs on PicoRuby with this warning enabled, and it works just fine. Can you clarify what concerns you have here?

----------------------------------------
Feature #20331: Should parser warn hash duplication and when clause?
https://bugs.ruby-lang.org/issues/20331#change-107259

* Author: yui-knk (Kaneko Yuichiro)
* Status: Open
----------------------------------------
# Background

Right now, parser warns duplicated hash keys (#1) and when clause (#2).
For example,

```ruby
{1 => :a, 1 => :b}

# => warning: key 1 is duplicated and overwritten on line 1
```

```ruby
case 2
when 1, 1
else
end

# => test.rb:2: warning: duplicated `when' clause with line 2 is ignored
```

The parser compares different cardinality numbers.

```ruby
{
    1 => :a,
  0x1 => :b,
  0b1 => :b,
  0d1 => :b,
  0o1 => :b,
}

# => test.rb:2: warning: key 1 is duplicated and overwritten on line 3
# => test.rb:3: warning: key 1 is duplicated and overwritten on line 4
# => test.rb:4: warning: key 1 is duplicated and overwritten on line 5
# => test.rb:5: warning: key 1 is duplicated and overwritten on line 6
```

# Problem

Currently this is implemeted by converting string like `"123"` to Ruby Object and compare them.
It's needed to remove Ruby Object from parse.y for Universal Parser.
I created PR https://github.com/ruby/ruby/pull/10079 which implements bignum for parse.y without dependency on Ruby Object, however nobu and mame express concern about the cost and benefit of implmenting bignum for parser.
I want to discuss which is the best approach for this problem.

By the way, it's needed to calculate irreducible fraction for Rational key if we will keep warning messages.

```ruby
$ ruby -wc -e '{10.2r => :a, 10.2r => :b}'
-e:1: warning: key (51/5) is duplicated and overwritten on line 1
-e:1: warning: unused literal ignored
Syntax OK
```

# Options

## 1. Warnings on parser

Pros: 
* Users of Universal Parser don't need to implement warnings by themselves. I guess developers of other Ruby implementation may get benefit of reducing their effort.
* Warnings are shown by `ruby -wc`.

Cons:
* We need to maintain bignum implementation for parser.

There are two approaches for this option.

### 1-1. Implement bignum for parser

The PR is this approach, implementing sub set of Ruby bignum for parser.

### 1-2. Extract existing bignum implementation then use it

Make existing bignum implementation to be independent of Ruby Object and use it from both bignum.c and parse.y.

## 2. Moving warnings logic into compile phase

We can use Ruby Object in compile.c. Then moving the logic into compile.c solves this problem.

Pros:
* No need to implement bignum for parser.

Cons:
* Users of Universal Parser need to implement warnings by themselves.
* Warnings are not shown by `ruby -wc`.




-- 
https://bugs.ruby-lang.org/
 ______________________________________________
 ruby-core mailing list -- ruby-core@ml.ruby-lang.org
 To unsubscribe send an email to ruby-core-leave@ml.ruby-lang.org
 ruby-core info -- https://ml.ruby-lang.org/mailman3/postorius/lists/ruby-core.ml.ruby-lang.org/