From: Dave Burt <dave@...>
Date: 2006-03-29T18:18:41+09:00
Subject: Re: strings and ruby style?

Kev Jackson wrote:
> My original goal was to allow the qc team to write
>
> text :id, :pg_401, :name
>
> So the arguments in theory should be symbols not strings, hence the need 
> for some kind of to_s/#{} hackery

That's fair enough, though I'd prefer to either use strings or change the 
$test_data hash to avoid having to convert. For instance, you could use a 
HashWithIndifferentAccess from Rails' Active Support module (require 
'active_support'), or you could just convert the keys to symbols as you load 
the data.

> ...
> However, coming from a Java background, "s1" + "s2" is a performance 
> (memory/speed) nightmare, so I shy away from S1 + S2 and try to use 
> #{S1}S2 where possible.  Perhaps string + string isn't so bad in Ruby, but 
> old habits die hard...

It's not so terrible in Java either; when you want to avoid it is where 
you're adding multiple times to get a single result string, especially in a 
big loop. Then you'd use a StringBuffer to avoid creating and 
garbage-collecting zillions of strings. Of course, Ruby's strings are 
mutable, so you don't need another class of object.

Compare:
s = ""; 50.times {|i| s += i.to_s + " "} # watch out: 50 strings are 
discarded

with:
s = ""; 50.times {|i| s << i.to_s << " " } # like Java's StringBuffer: one 
object, efficient

Anyway, in this case, "#{x}y" is about the same as x + "y", because you're 
only creating one extra string. You've just got the added method-call 
overhead for String#+.

> Thanks

You're welcome.

Cheers,
Dave