Bug #10795: to_s returns references to self if called on string

• Status: Rejected
• Assignee: -
• Target version: -
• ruby -v: ruby 2.1.5p273 (2014-11-13 revision 48405) [x86_64-darwin13.0]
• Backport: 2.0.0: UNKNOWN, 2.1: UNKNOWN, 2.2: UNKNOWN

[fra.boffa (Francesco Boffa)]

This is not actually a bug, but rather a strong violation of the Principle of least astonishment.

Lately I found a bug in the Rails project I work to. There was a method concatenating two float values with a comma, converting them to string like this:

def coords_to_string
  latitude.to_s << "," << longitude.to_s
end

The bug happened when we passed latitude and longitude as strings instead of floats.

Actually, 5.12345.to_s returns a new string object and any operation made on that string wont affect the original 5.12345. Instead "5.12345".to_s will return the same instance of that string, so that the << method will affect the original string.

So we found that our latitude variable was growing on each call and after a while it become "5.12345,13.12345,13.12345,13.12345,13.12345,13.12345"

This was probably made like this for performance reasons, however we found it a clear violation of the POLA, whereas we expected to_s to always return a new instance of that string.

Hope, this may help.
Francesco Boffa

[marcandre (Marc-Andre Lafortune)]

• Status changed from Open to Rejected

Remember, the principle of least astonishment applies to Matz only :-)

Changing this would, among other things, be a source of incompatibilities.

In any case, your code is better written

def coords_to_string
  "#{latitude},#{longitude}"
end

[sawa (Tsuyoshi Sawada)]

Marc-Andre Lafortune wrote:

In any case, your code is better written

def coords_to_string
  "#{latitude},#{longitude}"
end

or

def coords_to_string
  [latitude, longitude].join(",")
end

[fra.boffa (Francesco Boffa)]

For sure, our code was not the ideal for that task, and indeed, I already had changed it to one of your sane alternatives and the original committer has been appropriately thrown by the window :).

I insist, however, that this should at least be made clearer in the documentation of all the to_s methods available in the standard library.

Thanks for your attention.

[marcandre (Marc-Andre Lafortune)]

Francesco Boffa wrote:

I insist, however, that this should at least be made clearer in the documentation of all the to_s methods available in the standard library.

Not sure what you mean by "all" to_s methods. Only String#to_s returns the receiver.

The document is quite explicit, both in the interface and the description. OTOH, it didn't mention the effect for subclasses, so I modified it to follow that of Array#to_a.