[Tutor] Optimization

[Karl Pflästerer]

On 25 Oct 2003, Daniel Ehrenberg <- littledanehren at yahoo.com wrote:

> I just read GvR's essay about optimization in Python
> (at http://python.org/doc/essays/list2str.html), and I

I read it after having read your e-mail. It was interesting.

> came up with a somewhat efficient algorithm for it. It

But GvR came up with nearly the same one.

> was too easy to come up with it and it placed second
> in efficiency compared to his other programs (he made
> 7 algorithms for it). The program converts a list of

Funktion 6 is the same as yours.
,----[ f6; Python Patterns - An Optimization Anecdote ]
|  import string
|     def f6(list):
|         return string.joinfields(map(chr, list), "")
`----

The only difference is he uses the strinfg module.  I think the time the
essay was written the `join' method of strings didn't exist.  The string
module was the only possibility.

> numerical ASCII values into a string. The best two he
> made (mine did better than one, worse than the other)
> imported different modules, but mine didn't have to.

The reason is written above.  The time the essay had been written you
also would have to import the string module.

> There's recently been some debate about programming
> styles, so I'm wondering if I did somehting wrong with
> that. Here's the best program, written by GvR:

>>>> import array
>>>> def f7(list):
> ...     return array.array('B', list).tostring()

> and here's mine:

>>>> f8 = lambda list: ''.join(map(chr, list))

As I wrote yours f8 is nearly the same as GvR's f6.  Furthermore you
cheat a bit; you don't give a name to you function but use just a
lambda.  That can be a bit faster.  f6 written with the join method of
strings instead of the function from the strings module should behave
like your function.

> On my computer, his was twice as fast as mine, and
> mine was in turn twice as fast as the normally
> programmed, most obvious one:

>>>> def f1(list):
> ...     string = ""
> ...     for item in list:
> ...         string = string + chr(item)
> ...     return string

I don't know for whom that would be the most obvious one; for me not.
The mapping solution seems more obvious if you know a bit functional
programming.

> My function seemed fairly obvious to me, and it was
> better than most of his, so I think there's something
> wrong with it that made him not use it. There's

No there isn't.  In fact he did use it as I showed you above.

Furthermore your solution uses lambda; it is well known that Guido is
not the gretaest friend on earth of lambda.

> of programming are good and which ones are bad, so is
> this a bad style that I used?

IMO not.

*But*: do not use *list* as a parameter.  It is a builtin.

If you prefer functional or imperative style or a mixture; that's up to
you.  Just use the same style in one programm and don't shadow builtins
unless you know why you do it.


   Karl
-- 
Please do *not* send copies of replies to me.
I read the list