Why it is so dramatical?

[Christopher A. Craig]

djw <dwelch91 at nospam.attbi.com> writes:

> Stupid question:
> 
> If ""join() is 100-1000 X faster, why doesn't the Python interperter
> translate s = s1 + s2 + ... into s = "".join( [ s1, s2, ... ] ) 
> automatically?

You won't see that kind of dramatic results if you did that.  In fact,
my testing shows it would be slower.  

s = "".join(s1, s2) is slower than s=s1+s2.  The reason you see that
kind of speed up is that list.append(s1) is a ton faster than s+=s1,
and "".join() is moved outside the loop.  Compare the following
(presume 'a' contains some string):

method a)
s=""
for x in xrange(10000):
   s += a

method b)
l = []
for x in xrange(10000):
   l.append(a)
s = "".join(l)

method c)
s=""
for x in xrange(10000):
   s = "".join([s, a])


method a creates 10,000 individual string objects
method b creates 1 list object (with a bunch of appends) and 1 string
    object
method c creates 10,000 list objects (each of two elements s and a) and
    10,000 string objects.

On my system, as I would expect, method a takes about .40 seconds,
method b takes .048, and method c takes .51.

-- 
Christopher A. Craig <list-python at ccraig.org>
"Imagination is more important than knowledge" Albert Einstein