Feature #15896: Symbol#+

• Status: Rejected
• Assignee: -
• Target version: -

[zeus (Zeus 81)]

I've made this today

class Symbol
  @@cache_add = Hash.new {|h1,k1| h1[k1] = Hash.new {|h2,k2| h2[k2] = :"#{k1}#{k2}"}}
  def +(s) @@cache_add[self][s] end
end

:abc + :def
# => :abcdef

I thought it was so awesome it should become standard.

[Eregon (Benoit Daloze)]

FWIW, there is no need for a cache, since Symbols are interned (and an explicit cache would leak memory).

I think Symbols are generally different than Strings, so #+ feels a bit out of place for me here.
Can you show a realistic use case?

[jeremyevans0 (Jeremy Evans)]

I am against adding this to core. Symbols represent identifiers, not data/text. Strings should be used for data/text. I don't think we should add methods to Symbol that further blur the line between String and Symbol.

I would rather see :"#{s1}#{s2}" for explicitly combining two symbols in the few cases that need to do so. That would be slower than the cached approach as it currently results in 3 string allocations, but I don't think it is likely to be a bottleneck.

One issue with the cached approach you used is that the cache doesn't use a weak map, so dynamic symbols that use + would never get garbage collected.

[shevegen (Robert A. Heiler)]

I agree with benoit and jeremy, also in regards to the use case description; even speed/cache considerations aside, matz
possibly may consider the proposal to not fit into/to class Symbol by default. It is actually already awesome that class
Symbol can be modified like this :) - I am not sure if it is required to add '+' to class Symbol as a method, by default.
I can understand it somewhat to some extent (people love Symbols ever since the old pickaxe, I guess), but I think what
benoit wrote is quite true - it may be better to be conservative here and keep Symbols and Strings distinct as separate
entities. (See also what matz wrote about the origin of Symbols for ruby).

[ko1 (Koichi Sasada)]

• Status changed from Open to Rejected

Ut seems difficult to introduce.
Please reopen it with your opinion if it should be discussed more.