[ruby-core:87356] [Ruby trunk Feature#14808][Rejected] Last token of endless range should have EXPR_END

Issue #14808 has been updated by nobu (Nobuyoshi Nakada).

Status changed from Open to Rejected

That code is valid syntax in both versions, and it is *not* an endless range in the trunk as well as 2.5.
As `..` is an infix operator and requires the RHS, newlines following it are just ignored.
So it equals to:
```ruby
case 5
when (3..puts(true))
end
```

It results in an error at runtime in 2.5, not a syntax error.

```
$ ruby2.5 -rripper -e 'case 5' -e 'when 3..' -e  'puts(true)' -e end
true
Traceback (most recent call last):
-e:3:in `<main>': bad value for range (ArgumentError)
```

To put an endless range there, use parentheses.


----------------------------------------
Feature #14808: Last token of endless range should have EXPR_END
https://bugs.ruby-lang.org/issues/14808#change-72347

* Author: aycabta (aycabta .)
* Status: Rejected
* Priority: Normal
* Assignee: 
* Target version: 
----------------------------------------
In 2.5.1:

```ruby
irb(main):001:0> require 'ripper'
irb(main):002:0> pp Ripper.lex("case 5\nwhen 3..\n  puts(true)\nend\n")
[[[1, 0], :on_kw, "case", EXPR_BEG],
 [[1, 4], :on_sp, " ", EXPR_BEG],
 [[1, 5], :on_int, "5", EXPR_END|EXPR_ENDARG],
 [[1, 6], :on_nl, "\n", EXPR_BEG],
 [[2, 0], :on_kw, "when", EXPR_BEG],
 [[2, 4], :on_sp, " ", EXPR_BEG],
 [[2, 5], :on_int, "3", EXPR_END|EXPR_ENDARG],
 [[2, 6], :on_op, "..", EXPR_BEG],
 [[2, 8], :on_ignored_nl, "\n", EXPR_BEG],
 [[3, 0], :on_sp, "  ", EXPR_BEG],
 [[3, 2], :on_ident, "puts", EXPR_ARG],
 [[3, 6], :on_lparen, "(", EXPR_BEG|EXPR_LABEL],
 [[3, 7], :on_kw, "true", EXPR_END],
 [[3, 11], :on_rparen, ")", EXPR_ENDFN],
 [[3, 12], :on_nl, "\n", EXPR_BEG],
 [[4, 0], :on_kw, "end", EXPR_END],
 [[4, 3], :on_nl, "\n", EXPR_BEG]]
```

This is invalid code in 2.5.1, so I understand this result.

In 63451:

```ruby
irb(main):001:0> require 'ripper'
irb(main):002:0> pp Ripper.lex("case 5\nwhen 3..\n  puts(true)\nend\n")
[[[1, 0], :on_kw, "case", EXPR_BEG],
 [[1, 4], :on_sp, " ", EXPR_BEG],
 [[1, 5], :on_int, "5", EXPR_END],
 [[1, 6], :on_nl, "\n", EXPR_BEG],
 [[2, 0], :on_kw, "when", EXPR_BEG],
 [[2, 4], :on_sp, " ", EXPR_BEG],
 [[2, 5], :on_int, "3", EXPR_END],
 [[2, 6], :on_op, "..", EXPR_BEG],
 [[2, 8], :on_ignored_nl, "\n", EXPR_BEG],
 [[3, 0], :on_sp, "  ", EXPR_BEG],
 [[3, 2], :on_ident, "puts", EXPR_ARG],
 [[3, 6], :on_lparen, "(", EXPR_BEG|EXPR_LABEL],
 [[3, 7], :on_kw, "true", EXPR_END],
 [[3, 11], :on_rparen, ")", EXPR_ENDFN],
 [[3, 12], :on_nl, "\n", EXPR_BEG],
 [[4, 0], :on_kw, "end", EXPR_END],
 [[4, 3], :on_nl, "\n", EXPR_BEG]]
```

This is correct code in this revision.

I think that lex_state of the last token of endless range, `[[2, 6], :on_op, "..", EXPR_BEG]`, it should be EXPR_END. Because it's the end of an argument. It's important for REPL, RDoc, and so on.

But lex_state is parser matter in parse.y. How about this for the parser of Ruby compiler?



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

Unsubscribe: <mailto:[e-mail]?subject=unsubscribe>
<http://lists.ruby-lang.org/cgi-bin/mailman/options/ruby-core>