Project

General

Profile

Actions

Feature #16994

closed

Sets: shorthand for frozen sets of symbols / strings

Added by marcandre (Marc-Andre Lafortune) over 3 years ago. Updated over 3 years ago.

Status:
Feedback
Assignee:
-
Target version:
-
[ruby-core:98969]

Description

I would like a shorthand syntax for frozen Sets of symbols or of strings.

I am thinking of:

%ws{hello world} # => Set['hello', 'world'].freeze
%is{hello world} # => Set[:hello, :world].freeze

The individual strings would be frozen. These literals would be created once at parse time (like Regex are):

def foo
  p %ws{hello world}.object_id
end
foo
foo # => prints the same id twice

We should consider these sets to return a unique frozen to_a.

Reminder: Ruby has literal notations for Rational and Complex. I've sadly never had to use either.
I would venture to say that Complex is much less used than Sets, and that sets are underused.

Reminder: previous discussion for builtin syntax was not for frozen literal, strings or symbols specifically: https://bugs.ruby-lang.org/issues/5478

For builtin notations for generic sets (i.e. unfrozen or containing other than string/symbol), please discuss in another issue.


Related issues 1 (1 open0 closed)

Related to Ruby master - Feature #16989: Sets: need ♥️Openknu (Akinori MUSHA)Actions
Actions #1

Updated by marcandre (Marc-Andre Lafortune) over 3 years ago

Updated by Dan0042 (Daniel DeLorme) over 3 years ago

+1
I think this is more important than having a general Set syntax as discussed in #5478. Being able to use %ws[foo bar].include?(str) is a double-plus of not creating a new object each time and having O(1) efficiency.

Updated by Dan0042 (Daniel DeLorme) over 3 years ago

I just thought of something...
In the same way that "str".freeze is optimized to be deduplicated, %w[a b].include?(obj) could be optimized so it becomes equivalent to obj == -"a" || obj == -"b", or something around those lines. This would have the advantage that all existing ruby code that uses this pattern would automatically become faster, without having to convert to a new literal syntax.

Updated by Eregon (Benoit Daloze) over 3 years ago

Dan0042 (Daniel DeLorme) wrote in #note-3:

I just thought of something...
In the same way that "str".freeze is optimized to be deduplicated, %w[a b].include?(obj) could be optimized so it becomes equivalent to obj == -"a" || obj == -"b", or something around those lines.

That already works on TruffleRuby (and for more than this specific case), it needs a JIT, inlining (also through builtins like #include?) and escape analysis.

Updated by matz (Yukihiro Matsumoto) over 3 years ago

  • Status changed from Open to Feedback

We are going to introduce built-in set, but not in 3.0 (too little time to implement it before 3.0 release).
After merging built-in set, we will seriously consider this proposal.

Remaining issues:

  • Name? %ws would be the first two character specifier after %. Is it reasonable? Or should we seek another name?
  • Frozen? %w returns non frozen array of non frozen strings. How should %ws behave?

Matz.

Updated by normalperson (Eric Wong) over 3 years ago

wrote:

Remaining issues:

  • Name? %ws would be the first two character specifier after %. Is it reasonable? Or should we seek another name?
  • Frozen? %w returns non frozen array of non frozen strings. How should %ws behave?

How about suffix notation similar to Regexp modifiers?

[ 'foo', 'bar' ]s

Or with ability to specify ordering:

[ 'foo', 'bar' ]os  # ordered set
[ 'foo', 'bar' ]us  # unordered set

Fwiw, I sometimes wish I could use unordered hash to save space:

{ 'foo' => 'bar' }u

And maybe 'f' modifier for frozen strings of values

https://bugs.ruby-lang.org/issues/16994#change-87685

Actions

Also available in: Atom PDF

Like0
Like0Like0Like0Like0Like0Like0