Is this really wnat you'd want? ARGF and IO (and possibly Zlib's readers) can stream data lazily in lines/chars/bytes/codepoints, where forcing them to an Array would have to eagerly read everything (and potentially block).

irb(main):001:0> io = File.open('Makefile')
=> #File:Makefile
irb(main):002:0> lines = io.lines
=> #<Enumerator: #File:Makefile:lines>
irb(main):003:0> lines.next
=> "SHELL = /bin/sh\n"
irb(main):004:0> io.pos
=> 16
irb(main):005:0> lines.next
=> "NULLCMD = :\n"
irb(main):006:0> io.pos
=> 28
irb(main):007:0> lines.to_a; nil
=> nil
irb(main):008:0> io.pos
=> 12737

With the exception of StringIO hey also support infinite streams. I think the current Enumerator is best.

drbrain (Eric Hodel) wrote:

With the exception of StringIO hey also support infinite streams. I think the current Enumerator is best.

Strongly agreed. It could also increase a lot the memory usage (while on String we anyway already have the whole String in memory so it is less of a concern).

File.open('/usr/share/dict/words').lines.each_with_object(Hash.new(0)) { |word, count| count[word[0].downcase] += 1 }

Eregon (Benoit Daloze) wrote:

Strongly agreed. It could also increase a lot the memory usage (while on String we anyway already have the whole String in memory so it is less of a concern).

File.open('/usr/share/dict/words').lines.each_with_object(Hash.new(0)) { |word, count| count[word[0].downcase] += 1 }

In that case you can use IO#each_line, which still returns Enumerator.

 File.open('/usr/share/dict/words').each_line.with_object(Hash.new(0)) { |word, count| count[word[0].downcase] += 1 }

But in other words, we need to replace uses of IO#lines to IO#each_line, or our existing Ruby program may have performance problem with 2.0.0...

Ticket #6670 is reopened. Please continue discussion there http://bugs.ruby-lang.org/issues/6670 ¶