Why does the =~ operator only sometimes have side effects?

Question

I've noticed a side effect in Ruby/Oniguruma that is only present in 1 out of 4 seemingly equivalent statements. Why is the variable day defined in 009, but not in 003, 005 or 007?

irb(main):001:0> r = /(?<day>\d\d):(?<mon>\d\d)/
=> /(?<day>\d\d):(?<mon>\d\d)/

irb(main):002:0> r =~ "24:12"
=> 0
irb(main):003:0> day
NameError: undefined local variable or method `day' 

irb(main):004:0> "24:12" =~ r
=> 0
irb(main):005:0> day
NameError: undefined local variable or method `day'


irb(main):006:0> "24:12" =~ /(?<day>\d\d):(?<mon>\d\d)/
=> 0
irb(main):007:0> day
NameError: undefined local variable or method `day'


irb(main):008:0> /(?<day>\d\d):(?<mon>\d\d)/ =~ "24:12"
=> 0
irb(main):009:0> day
=> "24"

nb#1: It's the same regex and the same string in all four cases.

nb#2: I've verified the behavior in MS Windows and Ubuntu Linux.

Note: although this is not the case here, you should be careful when playing around with local variables in IRb. They may behave slightly differently in IRb than in a script, due to the way the code is being evaluated in IRb. Always write scripts to confirm. — Jörg W Mittag, May 25 '11 at 15:25

Vasiliy Ermolovich · Accepted Answer · 2011-05-25T13:03:20.370

13

When you call "24:12" =~ r you actually call "24:12".=~(r). So, String#=~ just returns the position the match starts, or nil if there is no match.

But when you call /(?<day>\d\d):(?<mon>\d\d)/ =~ "24:12" you actually call Regexp#=~

And as the documentation says

If =~ is used with a regexp literal with named captures, captured strings (or nil) is assigned to local variables named by the capture names.

what about 003:

The assignment is not occur if the regexp is not a literal.

   re = /(?<lhs>\w+)\s*=\s*(?<rhs>\w+)/
   re =~ "  x = y  "
   p lhs    # undefined local variable
   p rhs    # undefined local variable

and

The assignment is not occur if the regexp is placed at right hand side.
" x = y " =~ /(?<lhs>\w+)\s*=\s*(?<rhs>\w+)/
p lhs, rhs # undefined local variable

edited May 25 '11 at 13:03

answered May 25 '11 at 12:49

Vasiliy Ermolovich

24,459
5
79
77

Thanks Nash. A very good answer. But can you also explain `003` in the question -- i.e. a compiled regex (not a regex literal) receives the `=~` message and it doesn't assign the local variable. – Staffan Nöteberg May 25 '11 at 12:59
1

"Captured strings is assigned...", "The assignment is not occur..."? Strange grammar... – Tim Pietzcker May 25 '11 at 14:16
1

@Tim Pietzckerh, you can make it better! ;) http://ruby-doc.org/documentation-guidelines.html – Vasiliy Ermolovich May 25 '11 at 14:18
@Tim Pietzcker: the author of that paragraph would probably say the same about your Japanese ;-) – Jörg W Mittag May 25 '11 at 15:27
2

I just made an error report to ruby about the mistake in the doc (comment). – sawa May 25 '11 at 16:05

score 1 · Answer 2 · answered May 25 '11 at 13:52

1

I believe 003 isn't supported because it's a full blown Regexp object in Rubyland at that point, possibly with overridden methods and such. That makes the scope of assigned locals a lot more complicated.

answered May 25 '11 at 13:52

JEG2

21
2

Thanks, James. I had another idea about the intention from the Ruby creators, but you're probably right that it's a technical issue. – Staffan Nöteberg May 25 '11 at 14:53

Why does the =~ operator only sometimes have side effects?

2 Answers2