[Tutor] Alternatives to append() for "growing" a list

[Steven D'Aprano]

On Sun, Dec 01, 2013 at 02:32:38PM +1000, Amit Saha wrote:
> Hello,
> 
> I was told by someone (as a comment) that a code snippet such as this
> "would make Pythonistas talk my ear off about how evil the append()"
> function is:

There is absolutely nothing wrong with append. Either you have 
misunderstood, or the person who told you this was smoking crack.

> >>> mylist = []
> >>> mylist.append(1)
> # a number of times over

However, growing a list one item at a time using append like this is too 
much hard work. Some better solutions:

# You have an existing list, and want to append a bunch of things to it
mylist.extend([x, y+1, z*2])

# You want a list consisting of the same objects repeated many times
mylist = [0, 1, 2]*100  # like [0, 1, 2, 0, 1, 2, 0, 1, 2, ...]

# You want to create a list containing known objects
mylist = [1, 2, 4, 8, 16, 32, 64]

# Like above, but using a list comprehension
mylist = [2**n for n in range(7)]

# You don't know how many things you want to add.
mylist = []
x = 1
while x < 100:
    mylist.append(x)
    x = 3*x-1

and so on. As you can see, append has its job to do.


> I have some ideas that on an append() the list's internal size
> increases (doubled?) since CPython over-allocates memory, and such.

Not just on append. Any operation that adds or deletes items from a list 
might trigger a re-size. If the list needs to increase, it will double 
in size up to some maximum, then it will grow by a smaller amount. The 
purpose of this is that *on average* appending to the list will take a 
fixed amount of time.


> So, the question is: what is the alternative to growing a list when I
> have no idea what items or how many may be there?

Don't worry about growing the list. That's all handled for you as part 
of the list interface. The whole point of using Python is to not have to 
worry about internal details like that.



-- 
Steven