Kernel.=== in case statements, even though I never called it

ruby-talk
1

I was profiling a script I wrote, and the method with the greatest
amount of time used was Kernel.=== . I was surprised, not only since I
didn't use this method in my code, but my code had nothing involving
modules. When I changed a case statement that was inside my loop to an
if and many elsifs, I saved about 2 seconds on profiling, and the call
was gone. Is there a reason such a slow call is used in case statements?

···

--
Posted via http://www.ruby-forum.com/.

2

case uses === (pattern match) not ==.

It is very useful:

case str
when /foo/
   ...
end

case num
when 3..5
   ...
end

···

On 9/5/06, Jesse Silliman <wierdbro@gmail.com> wrote:

I was profiling a script I wrote, and the method with the greatest
amount of time used was Kernel.=== . I was surprised, not only since I
didn't use this method in my code, but my code had nothing involving
modules. When I changed a case statement that was inside my loop to an
if and many elsifs, I saved about 2 seconds on profiling, and the call
was gone. Is there a reason such a slow call is used in case statements?

--
Tomasz Wegrzanowski [ http://t-a-w.blogspot.com/ ]

3

Tomasz Wegrzanowski wrote:

case uses === (pattern match) not ==.

It is very useful:

case str
when /foo/
   ...
end

case num
when 3..5
   ...
end

Oh, I'm quite aware of that. Oh... its calling the objects inherited
Kernel.===, not calling Kernel::===. Ok, well, in that case, a good
question would be "Why is it so much slower in simple cases than == is,
even though for string they do the exact same thing?"

···

--
Posted via http://www.ruby-forum.com/\.

4

Kernel#=== calls == method, not aliases it. So you have 2 method calls,
and two dynamic dispatches instead of one: Kernel#===, then String#==.

You can test this by alias_method'ing === to == in String class :wink:

Relevant code:
VALUE
rb_equal(VALUE obj1, VALUE obj2)
{
    VALUE result;

    if (obj1 == obj2) return Qtrue;
    result = rb_funcall(obj1, id_eq, 1, obj2);
    if (RTEST(result)) return Qtrue;
    return Qfalse;
}

···

On 9/5/06, Jesse Silliman <wierdbro@gmail.com> wrote:

Oh, I'm quite aware of that. Oh... its calling the objects inherited
Kernel.===, not calling Kernel::===. Ok, well, in that case, a good
question would be "Why is it so much slower in simple cases than == is,
even though for string they do the exact same thing?"

--
Tomasz Wegrzanowski [ http://t-a-w.blogspot.com/ ]