Issue #11116 has been updated by mame (Yusuke Endoh).


Eregon (Benoit Daloze) wrote:
> Does that preserve the encoding of the String though?

The short answer: yes.

If the encoding is ASCII-compatible, it is preserved via its encoding of the resulting string.

```
s = "Hello こんにちは".encode("Windows-31J")
s = s.dump
puts s                 #=> "Hello \x82\xB1\x82\xF1\x82\xC9\x82\xBF\x82\xCD"
p s.encoding           #=> #<Encoding:Windows-31J>
s = eval(s)
p s.encoding           #=> #<Encoding:Windows-31J>
puts s.encode("UTF-8") #=> Hello こんにちは
```

If the encoding is not ASCII-compatible, the dumped string has an explicit `force_encoding`.

```
s = "Hello こんにちは".encode("UTF-16LE")
s = s.dump
puts s                 #=> "H\x00e\x00l\x00l\x00o\x00 \x00S0\x930k0a0o0".dup.force_encoding("UTF-16LE")
s = eval(s)
puts s.encode("UTF-8") #=> Hello こんにちは
```

I guess that there might be other subtle edge cases that I don't know.  It is difficult for me to write the detailed document.


> What about String#inspect, does it also eval() to itself?

No, it is not guaranteed.  `String#inspect` is just for human, so there is no annoying hack like the above non-ASCII-compatible encoding.  (I have encountered another subtle case, but I cannot remember...)

----------------------------------------
Bug #11116: The spec of String#dump
https://bugs.ruby-lang.org/issues/11116#change-76450

* Author: mame (Yusuke Endoh)
* Status: Closed
* Priority: Normal
* Assignee: mame (Yusuke Endoh)
* Target version: 
* ruby -v: ruby 2.2.1p85 (2015-02-26 revision 49769) [x86_64-linux]
* Backport: 2.0.0: UNKNOWN, 2.1: UNKNOWN, 2.2: UNKNOWN
----------------------------------------
The current spec says:

     call-seq:
       str.dump   -> new_str

    Produces a version of +str+ with all non-printing characters replaced by
    <code>\nnn</code> notation and all special characters escaped.

      "hello \n ''".dump  #=> "\"hello \\n ''\"

`\nnn` must be `\xnn` now.

In addition, I've expected String#dump to return a string that evaluates to an original string (except singleton methods, object id, etc.) when `eval`ed.  Is this a right expectation?  If so, it would be good to officially include the mention in the spec.  What do you think?

-- 
Yusuke Endoh <mame / ruby-lang.org>



-- 
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>